k-layer.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "alleged-damage",
   "metadata": {},
   "source": [
    "# K-Layer Networks on CIFAR-10, AdaGrad, Batch Normalization, Dropout, Data Augmentation, and Hyperparamter Search"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "exceptional-niger",
   "metadata": {},
   "source": [
    "## Contents\n",
    "\n",
    "* [Introduction](#introduction)\n",
    "* [Imports](#imports)\n",
    "* [Import Datasets](#import-datasets)\n",
    "* [Data Pre-Processing](#data-pre-processing)\n",
    "* [Data Augmentation](#data-augmentation)\n",
    "* [Model Building](#model-building)\n",
    "* [Hyperparameter Search](#hyperparameter-search)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "nutritional-single",
   "metadata": {},
   "source": [
    "## Introduction <a class=\"anchor\" id=\"introduction\"></a>\n",
    "\n",
    "In this notebook, I built on top of the examples in ```one-layer.ipynb``` and ```two-layer.ipynb```. At the core is the ```Dense``` layer (fully-connected layer) that can have activations such as ```SoftmaxActivation```, and ```ReLUActivation```. The learnable (or trainable) parameters of the ```Dense``` layer can be initialized with the ```XavierInitializer``` and the ```NormalInitializer```, and can be regularized with the ```L2Regularizer```. Further to the \\texttt{Dense} layer, batch normalization and dropout were implemented in the ```BatchNormalization``` and the ```Dropout``` layers. Dropout and batch normalization are used to avoid overfitting. A ```Model``` comprises any number of layers, and its loss function is the ```CategoricalCrossEntropyLoss```. The trainable parameters are optimized with the mini-batch gradient descent algorithm with adaptive learning rate via the ```AdaGradOptimizer``` and the learning rate schedule of the optimizer is set to be [```LRCyclingSchedule```](https://arxiv.org/abs/1506.01186). The ```Model``` is first compiled with the loss function, some metrics such as the ```AccuracyMetrics```, and the optimizer, and then it is fit to the data with the ```Model.fit``` method.\n",
    "\n",
    "In this notebook, it is demonstrated that the library can be used to build a ```Model``` with k-layers. In this notebook, to make it feasible, I implemented an up to five-layer neural network for multi-class classification such that all of the ```Dense``` layers' trainable parameters are initialized with the ```XavierInitializer``` (or the ```NormalInitializer``` when specified) and regularized with the ```L2Regularizer```. The variants of the aforementioned networks also contain a ```BatchNormalization``` and a ```Dropout``` layer after each hidden layer. Note that there is no batch normalization or dropout after the last layer which outputs the scores, or predictions of the network. In the up to five-layer network, there are [50, 30, 20, 10] nodes in the hidden layers. The hidden ```Dense``` layers have ```ReLUActivation``` activations, and the last ```Dense``` layer (classifier) has ```SoftmaxActivation``` activation. The ```Model``` is then compiled with the ```CategoricalCrossEntropyLoss``` loss function, the ```AccuracyMetrics``` performance metric, and the ```SGDOptimizer``` optimizer that uses the ```LRCyclingSchedule``` learning rate schedule.\n",
    "\n",
    "Data augmentation is used with the aid of the [imgaug](https://github.com/aleju/imgaug) library to further avoid overfitting, on top of dropout and batch normalization. Hyperparamter search is carried out to find the best combination of the newtwork depth (number of layers) and the L2 regularization rate. As in ```two-layer.ipynb```, the [hyperopt](#https://github.com/hyperopt/hyperopt), a Bayesian, distributed hyperparameter optimization library is used. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "banned-piece",
   "metadata": {},
   "source": [
    "## Imports <a class=\"anchor\" id=\"imports\"></a>\n",
    "\n",
    "Import the library parts and the required packages."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "prostate-basement",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm.notebook import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "import pickle\n",
    "from copy import deepcopy\n",
    "from math import sqrt, ceil\n",
    "import datetime\n",
    "import time\n",
    "import sys\n",
    "from itertools import product\n",
    "import pandas as pd\n",
    "import json\n",
    "from hyperopt import hp, fmin, tpe, space_eval, STATUS_OK, Trials\n",
    "import imgaug as ia\n",
    "import imgaug.augmenters as iaa\n",
    "\n",
    "from data_utils import load_cfar10_batch, load_label_names\n",
    "from losses import CategoricalCrossEntropyLoss, LossSmootherConstant\n",
    "from activations import LinearActivation, ReLUActivation, SoftmaxActivation\n",
    "from initializers import NormalInitializer, XavierInitializer\n",
    "from layers import Dense, BatchNormalization, Dropout\n",
    "from regularizers import L2Regularizer\n",
    "from models import Model\n",
    "from metrics import AccuracyMetrics\n",
    "from optimizers import AdaGradOptimizer\n",
    "from opt_utils import GradClipperByNothing\n",
    "from lr_schedules import LRConstantSchedule, LRExponentialDecaySchedule, LRCyclingSchedule\n",
    "from viz_utils import plot_losses, plot_costs, plot_accuracies, plot_lrs"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bibliographic-arrangement",
   "metadata": {},
   "source": [
    "## Importing Dataset and Data Pre-Processing <a class=\"anchor\" id=\"importing-dataset-and-data-pre-processing\"></a>\n",
    "\n",
    "In this notebook, the training-validation set is all of ```data_batch_1```, ```data_batch_2```, ```data_batch_3```, ```data_batch_4```, and ```data_batch_5```, that is then randomly split into the  training and the validation sets. Stratified split is used for splitting the train-val set into training and validation sets, that is, the class balance in the two sets are preserved. The test set is ```test_batch```. \n",
    "\n",
    "Pre-process datasets by standardization of features (subtract mean of each feature and divide by its standard deviation). Use the training mean and standard deviation on the validation and the test datasets! After this, the mean of each feature in the subsets should be around 0 and the standard deviation should be around 1.\n",
    "\n",
    "For a finer breakdown of these steps, see ```one-layer.ipynb```."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "narrow-panic",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train set shape: (45000, 32, 32, 3), val set shape: (5000, 32, 32, 3), test set shape: (10000, 32, 32, 3)\n",
      "train labels shape: (45000,), val labels shape: (5000,), test labels shape: (10000,)\n",
      "The class ratios in the training set are: [0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1]\n",
      "The class ratios in the validation set are: [0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1]\n",
      "The class ratios in the test set are: [0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1]\n",
      "[1 1 6 ... 8 3 8]\n"
     ]
    }
   ],
   "source": [
    "# Load data\n",
    "# train and val set are batch 1, 2, 3, 4, and 5, test set is test\n",
    "path = os.path.join(\"data\", \"data_batch_1\")\n",
    "x_train_img_1, y_train_1 = load_cfar10_batch(path)\n",
    "img_shape = x_train_img_1.shape[1:]\n",
    "\n",
    "path = os.path.join(\"data\", \"data_batch_2\")\n",
    "x_train_img_2, y_train_2 = load_cfar10_batch(path)\n",
    "\n",
    "path = os.path.join(\"data\", \"data_batch_3\")\n",
    "x_train_img_3, y_train_3 = load_cfar10_batch(path)\n",
    "\n",
    "path = os.path.join(\"data\", \"data_batch_4\")\n",
    "x_train_img_4, y_train_4 = load_cfar10_batch(path)\n",
    "\n",
    "path = os.path.join(\"data\", \"data_batch_5\")\n",
    "x_train_img_5, y_train_5 = load_cfar10_batch(path)\n",
    "\n",
    "x_train_val_img = np.vstack([x_train_img_1, x_train_img_2, x_train_img_3, x_train_img_4, x_train_img_5])\n",
    "y_train_val = np.hstack([y_train_1, y_train_2, y_train_3, y_train_4, y_train_5])\n",
    "\n",
    "x_train_img, x_val_img, y_train, y_val = train_test_split(x_train_val_img, y_train_val,\n",
    "                                                          test_size=0.1, random_state=42,\n",
    "                                                          stratify=y_train_val)\n",
    "\n",
    "path = os.path.join(\"data\", \"test_batch\")\n",
    "x_test_img, y_test = load_cfar10_batch(path)\n",
    "\n",
    "# check counts in datasets\n",
    "print(f\"train set shape: {x_train_img.shape}, \"\n",
    "      f\"val set shape: {x_val_img.shape}, test set shape: {x_test_img.shape}\")\n",
    "print(f\"train labels shape: {y_train.shape},\"\n",
    "      f\" val labels shape: {y_val.shape}, test labels shape: {y_test.shape}\")\n",
    "\n",
    "# assert balanced dataset\n",
    "train_counts = np.unique(y_train, return_counts=True)[1]\n",
    "train_ratios = train_counts / train_counts.sum()\n",
    "\n",
    "val_counts = np.unique(y_val, return_counts=True)[1]\n",
    "val_ratios = val_counts / val_counts.sum()\n",
    "\n",
    "test_counts = np.unique(y_test, return_counts=True)[1]\n",
    "test_ratios = test_counts / test_counts.sum()\n",
    "\n",
    "np.testing.assert_array_equal(train_ratios, val_ratios)\n",
    "np.testing.assert_array_equal(val_ratios, test_ratios)\n",
    "\n",
    "np.testing.assert_allclose(train_ratios, val_ratios, rtol=1e-1, atol=0)\n",
    "np.testing.assert_allclose(val_ratios, test_ratios, rtol=1e-1, atol=0)\n",
    "\n",
    "print(f\"The class ratios in the training set are: {train_ratios}\")\n",
    "print(f\"The class ratios in the validation set are: {val_ratios}\")\n",
    "print(f\"The class ratios in the test set are: {test_ratios}\")\n",
    "\n",
    "# Pre-process data\n",
    "x_train_un = x_train_img.reshape(x_train_img.shape[0], -1)\n",
    "x_val_un = x_val_img.reshape(x_val_img.shape[0], -1)\n",
    "x_test_un = x_test_img.reshape(x_test_img.shape[0], -1)\n",
    "\n",
    "x_train = x_train_un / 255.\n",
    "x_val = x_val_un / 255.\n",
    "x_test = x_test_un / 255.\n",
    "\n",
    "mean = np.mean(x_train, axis=0).reshape(1, x_train.shape[1])\n",
    "std = np.std(x_train, axis=0).reshape(1, x_train.shape[1])\n",
    "\n",
    "x_train = (x_train - mean) / std\n",
    "x_val = (x_val - mean) / std\n",
    "x_test = (x_test - mean) / std\n",
    "\n",
    "print(y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "stainless-century",
   "metadata": {},
   "source": [
    "## Data Augmentation <a class=\"anchor\" id=\"data-augmentation\"></a>\n",
    "\n",
    "Data augmentation is used to further decrease the likelihood of overfitting, that is, it is used a means of regularization. For data augmentation, the [imgaug](https://github.com/aleju/imgaug) library is used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "spiritual-evidence",
   "metadata": {},
   "outputs": [],
   "source": [
    "seq = iaa.Sequential([\n",
    "        iaa.Fliplr(0.5),  # horizontally flip 50% of the images\n",
    "        # Small gaussian blur with random sigma between 0 and 0.5.\n",
    "        # But we only blur about 50% of all images.\n",
    "        iaa.Sometimes(\n",
    "            0.5,\n",
    "            iaa.GaussianBlur(sigma=(0, 0.5))\n",
    "        ),\n",
    "    ])\n",
    "\n",
    "def aug_f(x, seq, img_shape):\n",
    "    x_img = x.reshape(x.shape[0], *img_shape)\n",
    "    x_img_aug = seq(images=x_img)\n",
    "    x = x_img.reshape(x_img_aug.shape[0], -1)\n",
    "\n",
    "    return x\n",
    "\n",
    "aug_func = lambda x: aug_f(x, seq, img_shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "assumed-reverse",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "example_imgs = x_train_val_img[:3]\n",
    "example_imgs_aug = seq(images=example_imgs)\n",
    "\n",
    "for example_img, example_img_aug in zip(example_imgs, example_imgs_aug):\n",
    "    plt.imshow(example_img)\n",
    "    plt.title(\"original\")\n",
    "    plt.show()\n",
    "    \n",
    "    plt.imshow(example_img_aug)\n",
    "    plt.title(\"aug\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "closed-swiss",
   "metadata": {},
   "source": [
    "## Model Building <a class=\"anchor\" id=\"model-building\"></a>\n",
    "\n",
    "Define the model building functions and run an example model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "quantitative-lightweight",
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_model(dims, reg_rate, in_dim, out_dim, seed):\n",
    "    params = {\"coeff\": 1.0, \"mean\": 0.0, \"std\": None}\n",
    "\n",
    "    dims = list(reversed(list(reversed(dims)) + [in_dim])) + [out_dim]\n",
    "\n",
    "    layers = []\n",
    "\n",
    "    for n_layer in range(1, len(dims)):\n",
    "        if n_layer == len(dims) - 1:\n",
    "            dense = \\\n",
    "                Dense(in_dim=dims[n_layer - 1], out_dim=dims[n_layer],\n",
    "                      kernel_initializer=XavierInitializer(seed=seed + n_layer, **params),\n",
    "                      bias_initializer=XavierInitializer(seed=seed + n_layer, **params),\n",
    "                      kernel_regularizer=L2Regularizer(reg_rate),\n",
    "                      activation=SoftmaxActivation()\n",
    "                      )\n",
    "        else:\n",
    "            dense = \\\n",
    "                Dense(in_dim=dims[n_layer - 1], out_dim=dims[n_layer],\n",
    "                      kernel_initializer=XavierInitializer(seed=seed + n_layer, **params),\n",
    "                      bias_initializer=XavierInitializer(seed=seed + n_layer, **params),\n",
    "                      kernel_regularizer=L2Regularizer(reg_rate),\n",
    "                      activation=ReLUActivation()\n",
    "                      )\n",
    "            bn = BatchNormalization(momentum=0.9, epsilon=1e-5)\n",
    "            dropout = Dropout(p=0.9)\n",
    "\n",
    "        layers.append(dense)\n",
    "\n",
    "        if n_layer == len(dims) - 1:\n",
    "            pass\n",
    "        else:\n",
    "            layers.append(bn)\n",
    "            layers.append(dropout)\n",
    "\n",
    "    model = Model(layers)\n",
    "    loss = CategoricalCrossEntropyLoss(loss_smoother=LossSmootherConstant())\n",
    "    \n",
    "    return model, loss, dims\n",
    "\n",
    "\n",
    "def compile_model(model, loss, step_size):\n",
    "    lr_initial = 1e-5\n",
    "    lr_max = 1e-1\n",
    "    lr_schedule = LRCyclingSchedule(lr_initial, lr_max, step_size)\n",
    "\n",
    "    optimizer = AdaGradOptimizer(lr_schedule=lr_schedule, grad_clipper=GradClipperByNothing())\n",
    "\n",
    "    metrics = [AccuracyMetrics()]\n",
    "\n",
    "    model.compile_model(optimizer, loss, metrics)\n",
    "    print(f\"len(dims)-1 = {len(dims) - 1}\")\n",
    "    print(model)\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "sapphire-selection",
   "metadata": {},
   "source": [
    "Derive the number of epochs needed that is equivalent of the desired number of cylical learning rate schedule cylces. Here 20 epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "incorrect-partition",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step size of cyc. lr: 2250 update steps\n",
      "full cycle of cyc.lr : 4500 update steps\n",
      "10.0 epochs = 1 full cycle = 4500 update steps\n",
      "2 cycle = 20 epochs = 9000 update steps\n"
     ]
    }
   ],
   "source": [
    "batch_size = 100\n",
    "\n",
    "step_size = int(5*np.floor(x_train.shape[0] / batch_size))\n",
    "print(f\"step size of cyc. lr: {step_size} update steps\")\n",
    "\n",
    "cycle_steps = 2*step_size\n",
    "print(f\"full cycle of cyc.lr : {cycle_steps} update steps\")\n",
    "\n",
    "epochs_one_full_cycle = (cycle_steps * batch_size) / x_train.shape[0]\n",
    "print(f\"{epochs_one_full_cycle} epochs = 1 full cycle = {cycle_steps} update steps\")\n",
    "\n",
    "n_cycle = 2\n",
    "n_epochs = int(n_cycle*epochs_one_full_cycle)\n",
    "print(f\"{n_cycle} cycle = {n_epochs} epochs = {n_cycle*cycle_steps} update steps\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "adopted-midnight",
   "metadata": {},
   "source": [
    "Run the example model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "rental-tradition",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "len(dims)-1 = 5\n",
      "model summary: \n",
      "layer 0: dense: \n",
      "\t shape -- in: 3072, out: 50\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.018042^2), reg: l2 with 2.0000e-01\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 1: batch norm with momentum 0.9\n",
      "layer 2: dropout with p=0.9\n",
      "layer 3: dense: \n",
      "\t shape -- in: 50, out: 30\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.141421^2), reg: l2 with 2.0000e-01\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 4: batch norm with momentum 0.9\n",
      "layer 5: dropout with p=0.9\n",
      "layer 6: dense: \n",
      "\t shape -- in: 30, out: 20\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.182574^2), reg: l2 with 2.0000e-01\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 7: batch norm with momentum 0.9\n",
      "layer 8: dropout with p=0.9\n",
      "layer 9: dense: \n",
      "\t shape -- in: 20, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.223607^2), reg: l2 with 2.0000e-01\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 10: batch norm with momentum 0.9\n",
      "layer 11: dropout with p=0.9\n",
      "layer 12: dense: \n",
      "\t shape -- in: 10, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.316228^2), reg: l2 with 2.0000e-01\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: softmax\n",
      "\n",
      "categorical cross-entropy loss with loss smoother constant\n",
      "adagrad with cycling lr schedule and clipper who does nothing\n",
      "\n",
      "starting epoch: 1 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.36it/s]\n",
      "epoch 1/20 \n",
      " \t -- train loss = 2.0428550513061876 / train cost = 3.778553771748582\n",
      "\t -- {\"accuracy_train\": 0.2634888888888889} \n",
      "\t -- val loss = 2.0415007674543637 / val cost = 3.777199487896758\n",
      "\t -- {\"accuracy_val\": 0.2688} \n",
      "\n",
      "\n",
      "starting epoch: 2 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.00it/s]\n",
      "epoch 2/20 \n",
      " \t -- train loss = 1.9388341239950662 / train cost = 2.238677684920399\n",
      "\t -- {\"accuracy_train\": 0.3302222222222222} \n",
      "\t -- val loss = 1.9340262759225273 / val cost = 2.2338698368478602\n",
      "\t -- {\"accuracy_val\": 0.324} \n",
      "\n",
      "\n",
      "starting epoch: 3 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.30it/s]\n",
      "epoch 3/20 \n",
      " \t -- train loss = 1.9114352448923546 / train cost = 2.103559246155381\n",
      "\t -- {\"accuracy_train\": 0.33035555555555557} \n",
      "\t -- val loss = 1.9125333605239103 / val cost = 2.104657361786937\n",
      "\t -- {\"accuracy_val\": 0.325} \n",
      "\n",
      "\n",
      "starting epoch: 4 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.12it/s]\n",
      "epoch 4/20 \n",
      " \t -- train loss = 1.960421587576926 / train cost = 2.1522979897371184\n",
      "\t -- {\"accuracy_train\": 0.28702222222222223} \n",
      "\t -- val loss = 1.965590776010175 / val cost = 2.1574671781703674\n",
      "\t -- {\"accuracy_val\": 0.2836} \n",
      "\n",
      "\n",
      "starting epoch: 5 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.57it/s]\n",
      "epoch 5/20 \n",
      " \t -- train loss = 1.926362217624622 / train cost = 2.1124103888408285\n",
      "\t -- {\"accuracy_train\": 0.31195555555555554} \n",
      "\t -- val loss = 1.9249491531781866 / val cost = 2.1109973243943934\n",
      "\t -- {\"accuracy_val\": 0.3204} \n",
      "\n",
      "\n",
      "starting epoch: 6 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 31.20it/s]\n",
      "epoch 6/20 \n",
      " \t -- train loss = 1.9310810159858416 / train cost = 2.1118944000563245\n",
      "\t -- {\"accuracy_train\": 0.3191777777777778} \n",
      "\t -- val loss = 1.937497196183444 / val cost = 2.118310580253927\n",
      "\t -- {\"accuracy_val\": 0.3144} \n",
      "\n",
      "\n",
      "starting epoch: 7 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.67it/s]\n",
      "epoch 7/20 \n",
      " \t -- train loss = 1.8936473123116775 / train cost = 2.0679968441218066\n",
      "\t -- {\"accuracy_train\": 0.30924444444444443} \n",
      "\t -- val loss = 1.9060783593956956 / val cost = 2.080427891205825\n",
      "\t -- {\"accuracy_val\": 0.307} \n",
      "\n",
      "\n",
      "starting epoch: 8 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.57it/s]\n",
      "epoch 8/20 \n",
      " \t -- train loss = 1.838313085815509 / train cost = 2.009526174727621\n",
      "\t -- {\"accuracy_train\": 0.32897777777777776} \n",
      "\t -- val loss = 1.8425593829331604 / val cost = 2.0137724718452725\n",
      "\t -- {\"accuracy_val\": 0.3258} \n",
      "\n",
      "\n",
      "starting epoch: 9 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.41it/s]\n",
      "epoch 9/20 \n",
      " \t -- train loss = 1.7588215479880953 / train cost = 1.9105633288836605\n",
      "\t -- {\"accuracy_train\": 0.3685777777777778} \n",
      "\t -- val loss = 1.7673482427601201 / val cost = 1.9190900236556854\n",
      "\t -- {\"accuracy_val\": 0.362} \n",
      "\n",
      "\n",
      "starting epoch: 10 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.85it/s]\n",
      "epoch 10/20 \n",
      " \t -- train loss = 1.6435843069340255 / train cost = 1.7739280825909396\n",
      "\t -- {\"accuracy_train\": 0.42704444444444445} \n",
      "\t -- val loss = 1.6670904341251094 / val cost = 1.7974342097820233\n",
      "\t -- {\"accuracy_val\": 0.409} \n",
      "\n",
      "\n",
      "starting epoch: 11 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.39it/s]\n",
      "epoch 11/20 \n",
      " \t -- train loss = 1.724612860168114 / train cost = 1.88289037634019\n",
      "\t -- {\"accuracy_train\": 0.38297777777777775} \n",
      "\t -- val loss = 1.7468186343060974 / val cost = 1.9050961504781734\n",
      "\t -- {\"accuracy_val\": 0.3698} \n",
      "\n",
      "\n",
      "starting epoch: 12 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:14<00:00, 30.29it/s]\n",
      "epoch 12/20 \n",
      " \t -- train loss = 1.7436461332877753 / train cost = 1.9115475252466578\n",
      "\t -- {\"accuracy_train\": 0.3691333333333333} \n",
      "\t -- val loss = 1.7594378311637497 / val cost = 1.9273392231226323\n",
      "\t -- {\"accuracy_val\": 0.363} \n",
      "\n",
      "\n",
      "starting epoch: 13 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:15<00:00, 28.62it/s]\n",
      "epoch 13/20 \n",
      " \t -- train loss = 1.8103362935745035 / train cost = 1.9851018336522668\n",
      "\t -- {\"accuracy_train\": 0.3456} \n",
      "\t -- val loss = 1.802866634696361 / val cost = 1.9776321747741243\n",
      "\t -- {\"accuracy_val\": 0.3464} \n",
      "\n",
      "\n",
      "starting epoch: 14 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:16<00:00, 27.09it/s]\n",
      "epoch 14/20 \n",
      " \t -- train loss = 1.8430949164025747 / train cost = 2.019211299015379\n",
      "\t -- {\"accuracy_train\": 0.33673333333333333} \n",
      "\t -- val loss = 1.851087347293069 / val cost = 2.027203729905873\n",
      "\t -- {\"accuracy_val\": 0.3384} \n",
      "\n",
      "\n",
      "starting epoch: 15 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:15<00:00, 28.38it/s]\n",
      "epoch 15/20 \n",
      " \t -- train loss = 1.8525104775091348 / train cost = 2.0355735232638095\n",
      "\t -- {\"accuracy_train\": 0.33068888888888887} \n",
      "\t -- val loss = 1.8595514031593836 / val cost = 2.0426144489140583\n",
      "\t -- {\"accuracy_val\": 0.3226} \n",
      "\n",
      "\n",
      "starting epoch: 16 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:13<00:00, 34.58it/s]\n",
      "epoch 16/20 \n",
      " \t -- train loss = 1.8559690468590238 / train cost = 2.031787936195528\n",
      "\t -- {\"accuracy_train\": 0.3137111111111111} \n",
      "\t -- val loss = 1.8630743995034365 / val cost = 2.038893288839941\n",
      "\t -- {\"accuracy_val\": 0.3044} \n",
      "\n",
      "\n",
      "starting epoch: 17 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:11<00:00, 40.73it/s]\n",
      "epoch 17/20 \n",
      " \t -- train loss = 1.8249650426043127 / train cost = 1.990699195602216\n",
      "\t -- {\"accuracy_train\": 0.33553333333333335} \n",
      "\t -- val loss = 1.8276318549644515 / val cost = 1.9933660079623547\n",
      "\t -- {\"accuracy_val\": 0.3368} \n",
      "\n",
      "\n",
      "starting epoch: 18 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:11<00:00, 40.17it/s]\n",
      "epoch 18/20 \n",
      " \t -- train loss = 1.750123785370002 / train cost = 1.9063014683519697\n",
      "\t -- {\"accuracy_train\": 0.3771555555555556} \n",
      "\t -- val loss = 1.7630815576529808 / val cost = 1.9192592406349485\n",
      "\t -- {\"accuracy_val\": 0.363} \n",
      "\n",
      "\n",
      "starting epoch: 19 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:11<00:00, 40.53it/s]\n",
      "epoch 19/20 \n",
      " \t -- train loss = 1.681637140743597 / train cost = 1.8251622749961114\n",
      "\t -- {\"accuracy_train\": 0.4030666666666667} \n",
      "\t -- val loss = 1.6941670968265186 / val cost = 1.837692231079033\n",
      "\t -- {\"accuracy_val\": 0.3898} \n",
      "\n",
      "\n",
      "starting epoch: 20 ...\n",
      "batch 450/450: 100%|██████████| 450/450 [00:11<00:00, 40.69it/s]\n",
      "epoch 20/20 \n",
      " \t -- train loss = 1.6022755288335468 / train cost = 1.7254068973733478\n",
      "\t -- {\"accuracy_train\": 0.43593333333333334} \n",
      "\t -- val loss = 1.6284001895213198 / val cost = 1.7515315580611208\n",
      "\t -- {\"accuracy_val\": 0.4232} \n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test acc: {'accuracy': 0.4337}\n"
     ]
    }
   ],
   "source": [
    "in_dim = x_train.shape[1]\n",
    "out_dim = np.unique(y_train).size\n",
    "\n",
    "# 5 layer, max depth\n",
    "dims = [50, 30, 20, 10]\n",
    "\n",
    "reg_rate = 0.2\n",
    "seed = 10\n",
    "\n",
    "model, loss, dims = build_model(dims, reg_rate, in_dim, out_dim, seed)\n",
    "model = compile_model(model, loss, step_size)\n",
    "\n",
    "verbose = 2\n",
    "history = model.fit(x_train, y_train, x_val, y_val, n_epochs, batch_size, verbose, aug_func)\n",
    "\n",
    "plot_losses(history, None)\n",
    "plot_costs(history, None)\n",
    "plot_accuracies(history, None)\n",
    "plot_lrs(history, None)\n",
    "\n",
    "params_val = {\"mode\": \"test\"}\n",
    "scores_test = model.forward(x_test, **params_val)\n",
    "y_hat_test = np.argmax(scores_test, axis=1)\n",
    "metrics_test = model.compute_metrics(y_test, scores_test)\n",
    "print(f\"test acc: {metrics_test}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "objective-mileage",
   "metadata": {},
   "source": [
    "Let's see some of the correct and incorrect predicitions of the trained model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "neither-tracker",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAEICAYAAACZA4KlAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAgeUlEQVR4nO2deZDd1XXnv+dtvUut1r40EpIFlqAwEBlwiPeJC2+xPakspMohVXaIayAzrvIkRZGpMbOkhswMuDzlxBM5UCguDw4V4+C4mIkxDsPgxMRilYgACaG91Wr1ot5fv+XMH++nciPf7+1W9+vXLX7fT1VXv77n3d89v/t+p3/v3e8755q7Qwjx9iez2A4IIRqDgl2IlKBgFyIlKNiFSAkKdiFSgoJdiJSgYF8CmNmVZvaCmY2Y2b9u0Jg/NrPrGjFWvTCz+83sC4vtx6WKgn1p8IcAnnL3Dnf/Hws9mJl9EsCIu7+w0GPNFTP7HTN75oLm/wbgj8yssBg+Xeoo2JcGmwG8woxmlq3zeF8A8M16HtBqZC5oy9VzDHfvAfAqgF+p53HTgoJ9kTGzHwH4IICvmdmomV1hZg+Z2dfN7HEzGwPwQTPbYWZPmdmQmb1iZr8y7RgrzexvzWzYzH5qZv85cFc8/9wCgA8B+L/T2rJmdreZvZF8lHjOzLoT2y8mxzyX/P7Faf2eMrM/NrMfAxgHsNXM3MzuMLODAA4mz/uEmb2Y+P4PZnbNtGN0m9mjZtZnZv1m9jUz2wHgfwJ4TzInQ9NO4SkAH5/frKcUd9fPIv+gdgF/ftrfDwE4B+Bm1P4hdwA4BOBuAOeDdQTAlcnzv538tALYCeA4gGfIWFcBGLug7Q8A7ANwJQAD8C4AKwF0ARgE8FkAOQC3Jn+vnOb3seSYOQB5AA7giaRvC4DrAZwBcCOALIDbABwB0JT8/RKArwBoA9AM4JeSY/9O6BwA/EsAzy/2a3Yp/ujOvnR5zN1/7O5VANcCaAdwr7tPufuPAHwfwK3JW/xfBfBldx93938GsCdy3E7U/lFM5/MA/p27v+Y1XnL3ftTuoAfd/ZvuXnb3h1F7G/3JaX0fcvdXEnspafsv7j7g7hMAfhfAn7v7s+5ecfc9AIoAbgJwA4ANAP7A3cfcfdLdg+9IpjGSnIO4SOr6mUrUlePTHm8AcDwJ/PMcBbARwGrUXsfjpO+FDKL2TmE63QDeCDx3QzLOdM6PGxtrettmALeZ2e9Payskx64AOOru5Yi/F9IBYOgini8SdGdfukxPRzwFoPuCBbDLAJwE0AegDGDTNFt35LgHUVtPuzBgtwWeewq1YJ3O+XFDfobajgP4Y3fvnPbTmrxLOA7gMrKQx9Ixd6D21l9cJAr2S4NnAYwB+EMzy5vZB1B7K/1td68AeBTAPWbWambvBPDb7EDJW+0fAnj/tOa/APCfzGx7sqp+jZmtBPA4gCvM7LfMLGdmv4HamsD3L8L3bwD4gpndmBy7zcw+bmYdAP4JQA+Ae5P2ZjO7OenXC2BTQGZ7P4D/fRHjiwQF+yWAu0+hJjd9FMBZAH8G4Lfd/dXkKXcCWA7gNGqS2sOofS5m/Dlqi27nuR/AIwB+AGAYwAMAWpLP7Z8A8CUA/ah9H+AT7n72Inzfi9rn9q+h9hHiEGqLb0j+UX0SwDtQW+g7AeA3kq4/Qk2OPG1mZwHAzNaj9s/mb2Y7vvgZlqxwircRZvYnANa5+22R5zwD4Pd9CX+x5kLM7D4Ab7j7ny22L5ciCva3Aclb9wJq8tm7UXv7/Xl3/5vF9EssLbQa//agA7W37htQ07TvA/DYonoklhy6swuRErRAJ0RKaOjb+K6u5b5p49pGDtkQDDanfk6l5JnGqzeRI8YGi7ofNup95MJy4mQvBgbOBV+1eQW7md0C4Kuofcf5L9z93tjzN21ci8cf+9OgrVKpBtuTcS6qfaGotx9v/ULchTbeL5Op93nz48XOLf4RkAR7pE+1ym2xKb4g2W5BmevH3li/TIb5z/uww33803fwcahlBpLvZP8patrvTtS+p71zrscTQiws8/mXeAOAQ+5+OPnSx7cBfKo+bgkh6s18gn0j3prwcAJvTZAAAJjZ7Wa218z2Dgycm8dwQoj5MJ9gD32K+rlPEu6+2913ufuurq7l8xhOCDEf5hPsJ/DW7KpNqGVJCSGWIPNZjf8pgO1mdjlqKY+/CeC3ZurEVhFjq77M1ugvBLHxoqvxsdXsyOpzjCpfxJ8j9Z9H/ppdfJ9aP96xUuHp8Gyle64KykIoQOzc6n19zznY3b1sZncC+DvUpLcH3Z0WTRRCLC7z0tnd/XHUki6EEEscfV1WiJSgYBciJSjYhUgJCnYhUkLDi1fMRU5oZMLLXOSfcplLP7HzzeX49Nf7nBspGTXaj7m8Zjz5ZOn4ONfjMXRnFyIlKNiFSAkKdiFSgoJdiJSgYBciJSyZ1fh6r0jGjje3EkF8vFifeBkmntESs8VW8eeSrBPz/1Igm43MP0k2is3vXOdjrokrjSq7dmm/ykKIWaNgFyIlKNiFSAkKdiFSgoJdiJSgYBciJTRUejMYlY2mpqZov0qlEmyfiwQ1k20uRKWaiHriFqmrVorIclHp8OLnN5fnTuYiu8+Yc1uJmSJTlUGkuF6GJxvlSpFung22V3ORF6bKD5iPhIxFTq7i4WsY4BUAHWHfazbmf0RipRYhxNsKBbsQKUHBLkRKULALkRIU7EKkBAW7ECmhodKbw6P12mi/OdR+y+fzFz3OghDd7ohLTYVcLJOL9zOE5yRyOBSnitxYaKamfOReQS0x3yPKW8mauB8Z/lpXcmHJ8eS5o7TP0LmT1NY2yV/QlS1rqW3NusuprVwNh2ElltUZmyzCvILdzI4AGAFQAVB2913zOZ4QYuGox539g+5+tg7HEUIsIPrMLkRKmG+wO4AfmNlzZnZ76AlmdruZ7TWzvQMD5+Y5nBBirsw32G929+sBfBTAHWb2vguf4O673X2Xu+/q6lo+z+GEEHNlXsHu7qeS32cAfBfADfVwSghRf+a8QGdmbQAy7j6SPP4IgP8Y6+PutNBfNsszfFjhvbkWbGxkgUWLZCFly/ycI8lmqFQj200RSSYXUSItx+Wkso3zjlWeyTV1NvyRrTwySfvkIyc9VOznfqwqUNNwPuzH66eeo30q2SFqs+N8PqaOdVDbxz/xr6ita9WWYHu5Gpn7DJNLI4VK+dFmZC2A7yaBmAPwv9z9/8zjeEKIBWTOwe7uhwG8q46+CCEWEElvQqQEBbsQKUHBLkRKULALkRIamvWWMUOhENaAqmRPrujxFmCPtTntrxUrKhk5L/dItpZHZMVs5Nwy4WKJY2Uu44xM9lDbQKmX2qZGRqgNB88EmyeP8Qy7N147QW1btnKZb3wrz4gb7WoJtrd28Es/195Obc1buMzXM8xlxYEJnj7SYZuC7RWSwQgAiFwfDN3ZhUgJCnYhUoKCXYiUoGAXIiUo2IVICQ1djQcQrclGu9R5u6Z6j5WJbNOTiRR/K2f4Cvlkla90D46HV7oBoG/4VLC9d4CvdA8OhfsAwGSBr55XR/mWUsv2E//f4NLFs08forbcjRuo7ZytpLaJ0opg+/ZMN+0Tk1fGC7wmQ9u2CWrrLfG6dmsz4fp0mSpf+c+AKRDa/kmI1KNgFyIlKNiFSAkKdiFSgoJdiJSgYBciJTReemPKQHSbpItPTolJaGaxBJqIH2S2ppzLZP1ECgOAk+NcajozzJNTevv5MUcmBoPtk+Ux2ifrXDos57jNJ/hknTwxEGxv6+d12tZuuYzaeioROazCt6gqtIRtQ7SGG9Cc55JXS54nyfQOhs8ZAPp7D1Db5vVXBtu7Wvh8oMKSqOawJZcQ4u2Fgl2IlKBgFyIlKNiFSAkKdiFSgoJdiJTQeOltDsxFestE/o95zJbntb36JsOZSy8dfor2OTb0CrX1l4aprVjmslZxlNcmq5bIXFX5S52P1LTr6OLS2/K1fKPO3qtag+0HxyPbOLXzue+6mo+14crwWADQsTxcuy4HLkVmBrks13+S1+QbfJPXmVu38p18vNGw1Jdr4VmF1Rypd2e8Vt+Md3Yze9DMzpjZ/mltXWb2hJkdTH6H8wiFEEuG2byNfwjALRe03QXgSXffDuDJ5G8hxBJmxmB396cBXPjVoE8B2JM83gPg0/V1SwhRb+a6QLfW3XsAIPm9hj3RzG43s71mtrd/gFf5EEIsLAu+Gu/uu919l7vvWtnFF1mEEAvLXIO918zWA0DymxdFE0IsCeYqvX0PwG0A7k1+P1Y3j+pFTK7LcHni9OhBant6/98F248OHKZ9prJcJhuvcKmpEpHeMlW+bRTK4f/fFb4zEUqR7bAykSzAzDg/aGdn+F1cC5HkAABVXrCx/TIuAbaB+9F+LtwvO8xlrT6ydRUADJ3g0tuOndup7aZdV1Hb+vaw/1NjPCsSWZLpVw1v/wXMTnp7GMA/ArjSzE6Y2edQC/JfNrODAH45+VsIsYSZ8c7u7rcS04fr7IsQYgHR12WFSAkKdiFSgoJdiJSgYBciJTQ4680wp4qTlFh1SC4njZS4tLL39b+ntsMD+4LtU6wSJYBikRdDLJZ4v1KRZ15FEvOQmwrLcsNnuKw1fILb8lNcyimA9+vIhAszNjuXDXMR2aj0ehu1TbRxOe/kVF+w/dxZnn23aVUXtb1/R3hfNgC4/oZOaqu0PEttJ048E2wfGuT5ZS2tO4LtpSKXFHVnFyIlKNiFSAkKdiFSgoJdiJSgYBciJSjYhUgJDZXeHE73YIvuzUbkutiebeOlIWp7+Q0ugxw7e5zakA/LRs5VMmCCZ72VJriGVp6K9BvlWXs+HLYNnuR9Ro7xDMEOuqcY0N7BM9Ga8mGpjPcAxstcyuv7uWJJP6OQ50VRVuTD18jWDi7l/cLGTmpbm+fS1ks/foPaDgzwzLz8qs3B9hWrufS2dmNTsL0S2RNPd3YhUoKCXYiUoGAXIiUo2IVICQp2IVJCw7d/iiy68z4kqWWqzFdGXzv8ErWd6n+T2lrbOrkj1XDCxfJOnuxSHubbDB07w7cLqrAaYwAGB0ap7eArp4PtxX5e2TdfDSetAMA4RqhtqsJfzKlMWE2YBFcgPMfvPZlJ3m9ijCsNHSvCK9qDY1xl+H8vhJNnAKB7ZSRkWvhrVlm2kdry7WGbd3bQPtkVZKzYHFKLEOJthYJdiJSgYBciJSjYhUgJCnYhUoKCXYiU0PAadBUP/38pV7mMk2sKyzi9w3x7nJ7+k9Q21sclktZl66ht+PRQsH1FN09Y2LiWmlDo58kuk1meFFKM1N67fMt1wfbuG9/LjzfG67ENnHqe2soDR/kxi2HJcaLEz3l0ktvKRX5fqkSunVGSXJPN8cSUapVnNj1zlMueO3aEE1oAYMeWldTW2h4Ow1yBn3NTa1iWy2R4qtFstn960MzOmNn+aW33mNlJM3sx+fnYTMcRQiwus3kb/xCAWwLtX3H3a5Ofx+vrlhCi3swY7O7+NBBJJhZCXBLMZ4HuTjN7OXmbTz+0mtntZrbXzPb2DwzNYzghxHyYa7B/HcA2ANcC6AFwH3uiu+92913uvmtlV+cchxNCzJc5Bbu797p7xd2rAL4B4Ib6uiWEqDdzkt7MbL279yR/fgbA/tjzz1OtVjE5FpY8zo1wKaRiYdvAyDjtM3Ccn9pPfsS3f7L8MLWNj4clnkPtfEmjawU/r/Ikl0ku3xmW0ADgmiveQW25THhboN6TXE4qjg7y4+W4j2eH+fyf6w/PcbHKjzeJArVVInJjPuLjVCW8pVQukn6ZzUZkvhaeiXbkHM/CbO/hdfJ+YWN3sH39ijW0z4r2sB/ZDPd9xmA3s4cBfADAKjM7AeDLAD5gZteittnaEQC/N9NxhBCLy4zB7u63BpofWABfhBALiL4uK0RKULALkRIU7EKkBAW7ECmhoVlv1UoFY+fCEkTPMS5fHT/WE2zvO8sltFcPvEptp4/wbXrGyzzzKlMIF2Y8meEZatfeGJZVAOCDv/oeartsa1hCA4CTR/l4j+wJr52eeP0E7VMucumtOM4z4qzCpaZSJVwEcsp4oUePSGhZC0toAODgBSfPTYQlx2zkPteU5xJgbP+qsSL3EZFstEIuPCctOe5Hnm2JRtoB3dmFSA0KdiFSgoJdiJSgYBciJSjYhUgJCnYhUkLDC05mSEaRRfYAmxgNZ1cdO8SLSvYcC+95BgDtrVz+yUSKHvYNhqW+ajOXO4bHebZTX3UfteUm+R5r2c5OapvKvB5s7znNi3O2Zlr4WDmeHeZZPo8VhG2VyCVXjW0E6Pz6sEiWGqrh17NS5cerVLmElm/ir/UKsq8cAKxaxffaa2oOS2z5QhP3g9jMJL0JkXoU7EKkBAW7EClBwS5ESlCwC5ESGrsabzzZoZrjSRXVbHgrIcvzGmidq3mtsMkir8dWqPDV+M0bwkktw1O8ztybB7kqsPyf+fQPF/mq6ju3cdt7P3J5sL3vME80Gj4VWenOcJtH7hXuJDklMr+5yEpyNsttVuY+ZjLhFf7mZp6Y0rmsldou38q3B7vm6q3UdsU2nhDV2UXqybXw1XjPkmtHq/FCCAW7EClBwS5ESlCwC5ESFOxCpAQFuxApYTY7wnQD+EsA6wBUAex296+aWReAvwKwBbVdYX7d3XkxM9S2j6kQSSaT4/JJ1+pwosbmrStpn+7LV1FbJnLW1dIoteWz4YSFwWHeZ3CMJ0DkRrhM0nuES3YT1SFq27EjLL29+30baZ+XnzxCbSu7uqjNyXwAQG9vOJmkUuX14hx86632lmZqW9bOpTIjSS3r1vLzWh3ZgLRrTRu1dV/Gj9mxgvvY1BG2NbeHax4CgDHpbZ416MoAvuTuOwDcBOAOM9sJ4C4AT7r7dgBPJn8LIZYoMwa7u/e4+/PJ4xEABwBsBPApAHuSp+0B8OkF8lEIUQcu6jO7mW0BcB2AZwGsPb+Ta/KbbzkphFh0Zh3sZtYO4DsAvuju/MPVz/e73cz2mtneoaFZdxNC1JlZBbuZ5VEL9G+5+6NJc6+ZrU/s6wEEy7i4+2533+Xuuzo7l9XDZyHEHJgx2K1W5+YBAAfc/f5ppu8BuC15fBuAx+rvnhCiXswm6+1mAJ8FsM/MXkza7gZwL4BHzOxzAI4B+LWZDmRw5CychVSIbP3T3haW3pYt47XT2nJcgmjNcflnqswzjapENlzZyv0YGOByzNEzXKkcPscz+kYiNdLKE0eD7ZuvXk37vKM7LNcBQNcyvhRTiWX77QvX6ytM8q2rUOXbeRUL/F1hoZlLVFPFcMbkmtX8dZkqch+9ym0FUksOAIplfs2B1PLLRmr8ZVh2G7/sZw52d38mcogPz9RfCLE00DfohEgJCnYhUoKCXYiUoGAXIiUo2IVICQ0tOOnuKJXDhSXLFS4nZYmMlovIdSP9PdRmziWjyYhEUiLTNTLOC1i+eaKP2g6c4vLaRJYX4NwKntF3diJ83pWNXFK8evt11Fbt59tXYbyXmt79jvBrs77C7y/FMS6vHa7wrZVKBZ4RZ9lwMcfxiLyGyJZXLW1c5hsf49dBUwvXxAr58GvjkS2qDGEfI8qb7uxCpAUFuxApQcEuREpQsAuREhTsQqQEBbsQKaGxe73B6R5gpTKXLWp1Li+mPS6hocpPu2+I+1FlRf7yfF+5cePyGiu+CQA2xkWUvld4ttyanWEZamIFP69K9QS1rW3iPrYVRqhtdXNYTjp9nB9v30F+Xk2beEZZayvPDpuqhMdrIZmUAJCNZK9NcUUU42NcPl63IVIU08IypWViQhqTB7lsqDu7EClBwS5ESlCwC5ESFOxCpAQFuxApoaGr8WaGfIGsPBpfWWer7uvW8YSQQuTMRobCdckAIF/lq7SFpnASRFNkNXjVBl7DrWPZMWp7+YVD1DbcF6lB92rYduVyXoMO2/jcr93IEz8uy0a2GhocCrb/8AhPnnm1yl/PZSM8Iae1xBOb1q/bEGyvViOJKQW+xVO5VKa2DPgqfj7HE5GaCmFbIc+vK2M16Oa5/ZMQ4m2Agl2IlKBgFyIlKNiFSAkKdiFSgoJdiJQwo/RmZt0A/hLAOtQ0sN3u/lUzuwfA7wI4X2Ttbnd/fIZjIZ8LD1maikgamXCf1au5rHX27AC1lap8rNWrO6mtKR9OMslmeXJHbIuqXGkdtQ2c5lshvX6YJ9ecOx0+t2N7h2if1ZdxP65cy+d4+Oyr1NZ+bjTYfvVmnjRUKvJkkb4iT/DonwiPBQBTZ04H20cHuPy6rovLlF0r11NboTmW7MJDrVJl58ZltGyW119kzEZnLwP4krs/b2YdAJ4zsycS21fc/b9f9KhCiIYzm73eegD0JI9HzOwAgI0L7ZgQor5c1Gd2M9sC4DoAzyZNd5rZy2b2oJnxWr9CiEVn1sFuZu0AvgPgi+4+DODrALYBuBa1O/99pN/tZrbXzPYODvJiB0KIhWVWwW5medQC/Vvu/igAuHuvu1fcvQrgGwBuCPV1993uvsvdd61YwRdnhBALy4zBbrVv3D8A4IC73z+tffqy5GcA7K+/e0KIejGb1fibAXwWwD4zezFpuxvArWZ2LWpFr44A+L2ZDlStVDE6FJaNiuNcDjt79my4vW+Y9hmNHG95F8+uKhd5RlmR1MlriWQ0jY1GZLmWTmr70HvfQ21trS9R24FD4Uy6nqP8vH7yg+PUdmV7F7VlB/m59ZwKz39rK393d0s3H+vIJPf/hSPc/2FyHazfsIn2KY3w8+of4JJu5zouYSJaTy5sq5R5NmJk5zPeZ6YnuPszxJuopi6EWFroG3RCpAQFuxApQcEuREpQsAuREhTsQqSEBm//BJiF/7+sWMFllxIp8ufOM6FWreSZS+3N4ew1ABgf5RlUk+MT5HhcTsqzLaMAjDkvlFiNSDU7r76C2nIt4WKJrx06SftM9PNvNv7kh/9AbcXIFkpr27uD7a++GZZRAaC1hxfgvHojzyi7rp0Xxdw/FJbssohsGUUKiwJAponLrM2R66q5hc9VEzlm7PquVmMFWsPozi5ESlCwC5ESFOxCpAQFuxApQcEuREpQsAuREhoqvWWyWXR0hGWqTIb/32HSRLEYzkIDAFR55pKVI5lobVxGy5D9tSZGefHC0hTP1vIcl9dKER87VnBpaMdV24LtmTzfh2xw5BS1bdvEJdHhviFq2/dKuGBm3xSXvEpjfB6PHudFNjdv30xtJ0jhzuwYz15bFi04yeejtZXLgzFZLk/2dMuR4qxAbK83ju7sQqQEBbsQKUHBLkRKULALkRIU7EKkBAW7ECmhodJbNpNBO8lQKhS4NMQkjZj0FpPyYqJFaWqK2lgxylyByypTRX68lizPaiqWeUacGe83NBTO2itXeJbUZHE5tW2OFIHsz/RQW3VsMNi+IlJks7WFZ5T1neHFRY+e5LKce1hKXWPh7EAAyFT5tdPWzLPXWiKZbUxeA/i1Gr2GJb0JIRgKdiFSgoJdiJSgYBciJSjYhUgJM67Gm1kzgKcBNCXP/2t3/7KZdQH4KwBbUNv+6dfdPbwE+7Nj0VXJTJbvZ1MgiTCtZb7Fk0dWMiuR2l6lyOp5hSSntCxbRvtUIz7mWvgKbcW50jA+xmvGda4I75y9fv1G2mdomB/Pi9y2ZgtP/LiqK3zekyWuCmQj59zRxVe6B4e4j83Z8LXTUuC+r169htq6t6yntlWr+LZibW189Z+trMdq0C3UanwRwIfc/V2obc98i5ndBOAuAE+6+3YATyZ/CyGWKDMGu9c4L97mkx8H8CkAe5L2PQA+vRAOCiHqw2z3Z88mO7ieAfCEuz8LYK279wBA8pu/9xFCLDqzCnZ3r7j7tQA2AbjBzK6e7QBmdruZ7TWzvf0D5+bophBivlzUary7DwF4CsAtAHrNbD0AJL+DJUHcfbe773L3XSu7+NcyhRALy4zBbmarzawzedwC4F8AeBXA9wDcljztNgCPLZCPQog6MJtEmPUA9phZFrV/Do+4+/fN7B8BPGJmnwNwDMCvzXQgB8CEF48oCdlcWJaziLyGiDSRA5c08mQsAGDdYlvxRAWSiPvVaqSeWYHLONwXPtiaSELR6AhPQJkc5wkoy1kJvciEVErcj4lJLomOjfGkIXZhtUZqDS5fHpYvAWDV6k5qW7aMy4OxGnSVSniyotfVHKS3GYPd3V8GcF2gvR/Ahy96RCHEoqBv0AmREhTsQqQEBbsQKUHBLkRKULALkRIslllT98HM+gAcTf5cBeBswwbnyI+3Ij/eyqXmx2Z3D+5f1dBgf8vAZnvdfdeiDC4/5EcK/dDbeCFSgoJdiJSwmMG+exHHno78eCvy4628bfxYtM/sQojGorfxQqQEBbsQKWFRgt3MbjGz18zskJktWqFKMztiZvvM7EUz29vAcR80szNmtn9aW5eZPWFmB5PfPM9yYf24x8xOJnPyopl9rAF+dJvZ35vZATN7xcz+TdLe0DmJ+NHQOTGzZjP7JzN7KfHjPyTt85sPd2/oD4AsgDcAbAVQAPASgJ2N9iPx5QiAVYsw7vsAXA9g/7S2/wrgruTxXQD+ZJH8uAfAv23wfKwHcH3yuAPA6wB2NnpOIn40dE5Qy/pvTx7nATwL4Kb5zsdi3NlvAHDI3Q+7+xSAb6NWqTY1uPvTAAYuaG54tV7iR8Nx9x53fz55PALgAICNaPCcRPxoKF6j7hWdFyPYNwI4Pu3vE1iECU1wAD8ws+fM7PZF8uE8S6la751m9nLyNn/BP05Mx8y2oFYsZVErGF/gB9DgOVmIis6LEeyhejqLpf/d7O7XA/gogDvM7H2L5MdS4usAtqG2IUgPgPsaNbCZtQP4DoAvujuvh9V4Pxo+Jz6Pis6MxQj2EwC6p/29CcCpRfAD7n4q+X0GwHdR+4ixWMyqWu9C4+69yYVWBfANNGhOzCyPWoB9y90fTZobPichPxZrTpKxh3CRFZ0ZixHsPwWw3cwuN7MCgN9ErVJtQzGzNjPrOP8YwEcA7I/3WlCWRLXe8xdTwmfQgDmxWvXEBwAccPf7p5kaOifMj0bPyYJVdG7UCuMFq40fQ22l8w0Af7RIPmxFTQl4CcArjfQDwMOovR0sofZO53MAVqK2Z97B5HfXIvnxTQD7ALycXFzrG+DHL6H2Ue5lAC8mPx9r9JxE/GjonAC4BsALyXj7Afz7pH1e86GvywqREvQNOiFSgoJdiJSgYBciJSjYhUgJCnYhUoKCXYiUoGAXIiX8f6ISIwEbtEgNAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAEICAYAAACZA4KlAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAhrElEQVR4nO2de5BlV3Xev3Xfj77dPd3zfugtMMLRqyYytiiDLWxkYQdIFbblKkpUyZapgBNV2U6psCtWknKFPIDClYRkMCoERbBxgEAITiwrUYhsl9AI9IyQBknz0sz0TM9Mz/Tz9n2s/HHPhNZ4f7t7prvvHWl/v6que+9ed++zzz5n9Tl3f2etbe4OIcQbn9ygOyCE6A9ydiESQc4uRCLI2YVIBDm7EIkgZxciEeTsb1DMzM3smgv4/l+Z2U3r2ae1xsw+aWYfHnQ/Xi/I2QeEme03s3cNuh8AYGa/BGDa3b8/6L4wzOxDZvboecX/GsDvmVlpEH16vSFnv0Qxs0IfN/dhAF9cywatR+68sjXdJ3c/CuAHAP7eWrb7RkXOPgDM7IsALgPwX81sxsz+sZldkd16321mBwH8TzN7p5kdPq/u/78jMLO8mX3MzF4ys2kze8LMdgW293YzO2RmPxOwlQD8LID/vaSMtmtmP2Vmj5vZmez1p5bUe8TM/tDM/grAHICrsn36iJntA7Av+94vmtmTZjZlZn9tZtcvaWOXmX3NzE6Y2Ukz+7dm9hYA/wHAT2bjNbVkFx4B8J4LOwKJ4u76G8AfgP0A3rXk8xUAHMAXANQBVAG8E8BhVg/A7wJ4BsCbARiAGwCMZzYHcA2AdwM4BOAW0o+3Apg9ryzYLoAxAKcBfBBAAcCd2edz23wEwMGszQKAYtaPh7K6VQA3AzgO4CcA5AHcle1TOfv8FIBPZWNQAfD2rO0PAXg00P+/D+B7gz6er4c/XdkvPe5391l3n1/Bd38dwO+7+wve4yl3P7nE/gEAewDc4e7fJW2MApheYbvvAbDP3b/o7m13/zJ6t9G/tKTu5939uczeysr+hbufyvbpNwD8R3d/zN077v4ggCaAtwG4BcB2AL+bjcGCu5//O/18prN9EMvQz9+FYmUcuoDv7gLwUsR+L4AvuPszke+cBtBYYbvbARw4r+wAgB1LPof6v7TscgB3mdlvLSkrZW13ABxw93akv+fTADB1Ad9PFl3ZBwcLN1xaPgugdu6DmeUBbFpiPwTg6sg2PgDgfWZ2b+Q7+3pN2/kOG2r3CHrOupTLALy65HNov5aWHQLwh+4+uuSvlt0lHAJwGZnIY+P1FvRu/cUyyNkHxwSAq5b5zosAKmb2HjMrAvh99H7bnuOPAfxzM7s2m/2+3szGl9iPALgNwD80s38Q2kB2q/2XAN6xgna/DeBNZvZrZlYws18BcB2Ab618t/FZAB82s5/I2q5n+9cA8F0ARwF8PCuvmNmtWb0JADsDMts7APz5BWw/XQY9aZDqH4D3ojeZNQXgd/CjCbrCed/7EHoOcDz73n78aIIuj94/gFfQ++36OICdmc0BXJO9vxK92+1fJ315D4A/X/I51u7bATwB4Ez2+vYl9R45fxtL+7Gk7Paszals3/4MQCOzXQbgvwA4CWASwB9l5SUA/w3AKQCTWdk2AIcBlAZ9PF8Pf5YNmkic7IGV3/JL+MGa8zGzTwB4yd3//aD78npAzi5EIug3uxCJIGcXIhHk7EIkQl8fqqmWy96oDwVtZrGaZF4hWidivOhpinDF2LSHRXeM21rtFrXF5lnyuXy4zkXudC7Hrwe5HO9/Ph/uR7FYvKhtdbtdamu1+DM4bKxiYxg9YpE+srFfrh4bk0q1FiwHgHI5HOh34NBBnDx5MrgLq3J2M7sdwKfRk2r+2N0/Hvt+oz6EX3n3u0lbsQ2FD0w+0nsDH3g4H3h3flJ1Oh1Sh28ql4/cPOX4Dhw7PkFti4uL1DYyPBIsjzmLg9vq9Sq1VSoV3o+RcD+2bt1K61QjJ/f8HH96+PjxE9S2sLAQLO9ExiMXORkrZb7PjaHzH0RcUq/ObRs3h8fkx6+/gda57Oorg+XvuO0dwXJgFbfx2dNc/w7AL6D3YMWdZnbdxbYnhFhfVvOb/RYAP3T3l919EcCfoPegiBDiEmQ1zr4Drw1wOIzXBkQAAMzsHjPba2Z755vhWyohxPqzGmcP/bD5W79e3X2Pu+92993VyO8dIcT6shpnP4xeKOQ5dqIXeCGEuARZzWz84wCuNbMr0Qtx/FUAvxav4nRG23J8SpupFq3FcFsA4M5nVC3yPy4qyZBZWiYzZQ1GTLz/G8fHqS2mGOQL4UMaVSkjsly3y/vYbDapbWpqKlh+5ky4HAAKBZ43spDnkl2rxWVKphg0Gnx2vFrlCsRQrU5ttYitUOZt1mphFaLbiSgoTFKkNVbh7O7eNrOPAvgf6ElvD7j7cxfbnhBifVmVzu7u30YvxlkIcYmjx2WFSAQ5uxCJIGcXIhHk7EIkQl+j3swMhSKLhuL/d4aGwtJEtcof0jl69Bi1zc/zJ/likUtdIncsti5crgMARLZF9UYA+YjNSNBQgUhyvU1FthWRFeMRfWFiMllsW/U6l7ViATnlcjlYHhsPVgcAykUuDxYjNiaJAkCeBEt1o0FZJNIvIvXqyi5EIsjZhUgEObsQiSBnFyIR5OxCJEJfZ+NzuRyGG+FZ1e07ttF6bDZ+euYsrXNiMpI7rclnkbuR4BSe2ikWfhDJqxYJdOi0eZsscAIACoXw9iIT7nQ2uGfjM+SFmI3MPhciASiF2Ex3JEgm1sdSKVwvNhsfy5MXs8XGIxexsTYLkbxrtkxoU7APF1xDCPG6RM4uRCLI2YVIBDm7EIkgZxciEeTsQiRCn6U3Q7UWDlqI5To7cGB/sHzy5HFaZ36OB7t4ZEWYWGxHoRCWT2IyTqHIbZ1InrxmZNWX2LJLzBTbr1hAS9QWC8ghcl48IOTigm7iS1SRfkS2FZUbY5JdictypUheOxbIc7G5DRm6sguRCHJ2IRJBzi5EIsjZhUgEObsQiSBnFyIR+iq9tdotHD0WXg7uwIGXaL12J5y3rFTmkVC5fEyquUhppUCikyJ1YppXN6KexPKPxaLsmBxmEbnOYhFxsbGK2YisaLHltXg3onJjLGqPHRsWDQcA1QqXyaqRfHe1Ks+TV20MU1uFLBsVk94uQnlbnbOb2X4A0wA6ANruvns17Qkh1o+1uLL/jLtPrkE7Qoh1RL/ZhUiE1Tq7A/gLM3vCzO4JfcHM7jGzvWa2d36BL/ErhFhfVnsbf6u7HzGzzQAeMrMfuPt3ln7B3fcA2AMAm8c3XMS0ghBiLVjVld3dj2SvxwF8HcAta9EpIcTac9FXdjOrA8i5+3T2/ucB/LN4rS5g4Wi0UpFHeVVKYSmkEJE6YlJTPnJ/EYvkotFVsZCySGSbdckSPgDaXf6Tp7XIo6tq1fDSRe0C3+lugS/JVGnzbQ3P8UjFIRLt14nIdZ0KH/vY8mDliBxWrYUlr3ptiNZp1Lj0FpPsupHTYIhEewJAbSR8Hndi98Fkma8Yq7mN3wLg61noYQHAf3L3/76K9oQQ68hFO7u7vwzghjXsixBiHZH0JkQiyNmFSAQ5uxCJIGcXIhH6GvXW7TrmFuaDtpEql2Qa9fD6YPkql0/aHS5rLcyG+wDEo7yYLOeRCLVcjg9xtB5JbgkA1Q6XhkY9LL01W5F15QpcQitGrgdnj/GQiPFN4bX7hkb5Wm9nW1xubAxx6aoyPEJt5dGNwfJSncu25TI/ZrXIWm9Tp05TWyeyhmC5Ej5mcwtcju4i3J5HIiJ1ZRciEeTsQiSCnF2IRJCzC5EIcnYhEqGvs/HuhnYnPKtaKvMcXaMbNoTbi2xrfpFHJbSKfGa6GJltZYEwHkkIlnM+xN3I/9rhCg+4qOTCs7cAUDozHSwfyvMZ/AIJFgGAa266ntperu6jtkNHjoX7cZArIZ15vmRXaftmait3+Uz90EYyVpH8hfk6P2YWW+qryfsxPc/3u9EiM/WRIKrYOcfQlV2IRJCzC5EIcnYhEkHOLkQiyNmFSAQ5uxCJ0FfpzXJ5FIrh4JVKdROtV63VwoYuz53mziW0do3bukwGAdAmtlhASzEikXRmeOBHdYgHjGwcD0uRAIDc4WDxtrHtvB8VLvG86U0/Rm0zkf3+v4f2B8u38/gkdCbP8PbOTFHb2Dwfx/lSWHq7/MeuoXXqdS7LlSvcZcpbx6ht8vhZauuS5HWWj8i2kt6EEAw5uxCJIGcXIhHk7EIkgpxdiESQswuRCH2V3tDtotsK59U6c5bLJ/lCuE57YY7WOTUVkWM6XF7LgctQnWZY6stHln8qtLltLM8lwJbxyLbq5Ty6asebrgqWD1WvpHVefvEAtcUi81oV3v8zFtbYaiQHIQCU8nysjg3zfsw3eL2FuZPhbc2N0zqFBt9WocW1w2KRSMQAjKuUmJ0Oj0l9mOfJyxfI2EfOxWWv7Gb2gJkdN7Nnl5SNmdlDZrYve40Iv0KIS4GV3MZ/HsDt55XdB+Bhd78WwMPZZyHEJcyyzp6tt37qvOL3Angwe/8ggPetbbeEEGvNxU7QbXH3owCQvdI0ImZ2j5ntNbO9C4s8D7YQYn1Z99l4d9/j7rvdfXclsra1EGJ9uVhnnzCzbQCQvR5fuy4JIdaDi5XevgngLgAfz16/sZJKuRxQr4Rlr/nIUjcvHZgI15nh0luzyWUhK3F5YrjG5Y6hclhaidXpzHIJcPsGvmxRZ4gngZxr8Qiqse3hyKtTx6ZonZNT3LbY4pGFlQZffmtk+9ZgeWcinBATAPJVLl3l67Hlq/hSWbOtcBLLg68epHW6C3x884sz1GYd7k7Tc9y2dVc4Aq82xM+rbpdEvUWC4VYivX0ZwN8AeLOZHTazu9Fz8p8zs30Afi77LIS4hFn2yu7udxLTbWvcFyHEOqLHZYVIBDm7EIkgZxciEeTsQiRCX6Pe8nlglEQozS7ytchytbBENTK6jdaplnlUE0gUHQAUI9JFoxKWhnZu5f3oRqS36ROHqG2uxaPDNpW5rLhw+miwfGb/+U88/4juDN/WyanT1NYBl8OGR8LHrNuKRGWd4v34OzORtdmaPBKtPBy2jbb4cRld4IkvOwVer1vi5/Bim7taZyEsb7IEpwCwsBjuRzeSBFRXdiESQc4uRCLI2YVIBDm7EIkgZxciEeTsQiRCX6W3Qr6A8ZGwJLZrhK8pNufhSKlCgUdJVao8KWO7OUttuRbX3vIW/t84OszX+MoVuVTTmuORwblIgsLOPDceOhKWmjpTXJ7yE1yWm3j2OWqbbIcjygBgjEQkDud4Is3cDI+Iu67F65VLPOqN5V8sdLjMV4uMb2GIbwu1iKw4zCMcrTEaLF+InAP8Or2KhJNCiDcGcnYhEkHOLkQiyNmFSAQ5uxCJ0NfZeEMO+Vx4lrxS4F1ZJPnkWvN8ptud5wpzEngAAD7PZ61bZHmifJn3vdDkQTftSLLdWpHPPk+1eR/brbAyUJ2fpHWqx7ht9pHHqa1c533cWA/nT9uMyGz8HFdCylVuq47wgeyWwvWKkbE3EmQCALnDXIGolrmttJMrR4Urw9fcEx6bWQ+rAryGruxCJIOcXYhEkLMLkQhydiESQc4uRCLI2YVIhL5Kb+32IiZPvhq0HXiJ5/1qeVi+yhe4jJM3vjRUMRLo0InIebP5cL3JMR4I05rhEuCWq3nuuvKGDdQ2GcmfNtYJS02NKV5nc5fntCvMcumw0+XjWGQBRc7lqSJXRNEpcbmx2eRBLdYJi1GlSK62AoueAZBf5P3Ib+bBLoUrdlJbsxy+5sZy/BVj6zwRVrL80wNmdtzMnl1Sdr+ZvWpmT2Z/d1zwloUQfWUlt/GfB3B7oPxT7n5j9vftte2WEGKtWdbZ3f07AHjAsxDidcFqJug+amZPZ7f59Aemmd1jZnvNbO9sZFlmIcT6crHO/hkAVwO4EcBRAJ9gX3T3Pe6+29131yuRB5KFEOvKRTm7u0+4e8fduwA+C+CWte2WEGKtuSjpzcy2ufu5dYbeD+DZ2PfPkcsB9Vr4/0vzNJfKcu2wbFQp8HxguQKXTzzHl9UZr/PcddPzYRltg3GJpDwUjv4CgJpz6XCWRDUBwMw8l69G8+FDWu7w/+ulXOQ0iFwOFtpcK+sSpa9c4ONrBS4BlvN8PEpF3ma1Em6z5JEousjYg7QHAHM53manwse4VQ3f8Xacn6fWIedcRJFb1tnN7MsA3glgo5kdBvAHAN5pZjdmTe8H8JvLtSOEGCzLOru73xko/tw69EUIsY7ocVkhEkHOLkQiyNmFSAQ5uxCJ0NeoN4ej2wk/RbcwzZdkatTC0kquzaWJ023eXqVW5dviqhbGSBLIjSSyCgCu2bSd2iw/RG1Pn4z0vx2JyiIpB7s1LvN1KtxW7PIxzne4LUe62I0ky7TxUWrb8K6fpLbqZh4hmCORebG+t8/yse8cPkptI1ddSW2LW7dS2/HZsKSbK3OZb5hE7VlEe9OVXYhEkLMLkQhydiESQc4uRCLI2YVIBDm7EInQ57XeDHkLR/iMDfNY95FGOHLs9DSPNpuPJFi8emQXtdVOTVFbeSYc5VWeOkvrFOZ4RNZMiUs87VF+aIpFHpXVZgkRt3B5atfNN1BbObIomkWSNrKoLI8k9Jw4cozavr//ALUVJk9QW6kQvp7Vi1zWqrb5flU7PNKvFFlordPl19VTk1PB8vHt47SORyItGbqyC5EIcnYhEkHOLkQiyNmFSAQ5uxCJ0NfZ+EKhhE1j4cCQ6tAorbfYDgfPPPXC07TOqSbPadfcyWdi3/pWPjNN0udhbnaK1hnavJnacpGca7MT+6ltocVTclfJrG+zxmfVp3fx5asmSWANEM8LV8mH962W4/u8f4rPqv/lXz9KbYuRZZIK+XD/62U+Hg3SdwD4u9t4QMvIlTwQZuHwBLW1p8PRV7kdtAq6+XDAi0cUAV3ZhUgEObsQiSBnFyIR5OxCJIKcXYhEkLMLkQgrWRFmF4AvANgKoAtgj7t/2szGAPwpgCvQWxXml939dKytXM5QqYfznT351D5aL18MdzMXiQUoRAIdTuZ4HjRsHKamEpGaapfzgIWNV19FbfMzXELz04eorRvJnwayzFM3cqRfePmH1PbKoVepzbo831kxFx6rQqTrExNcejth/Jg5kdcAoEYCYVpEugKAySYPULIjh6lt52m+snmpw2XKhW44uKbDlngCkCPLfJnxsVjJlb0N4Lfd/S0A3gbgI2Z2HYD7ADzs7tcCeDj7LIS4RFnW2d39qLt/L3s/DeB5ADsAvBfAg9nXHgTwvnXqoxBiDbig3+xmdgWAmwA8BmDLuZVcs1f+qJgQYuCs2NnNbAjAVwHc6+48W8PfrnePme01s73Ts/MX00chxBqwImc3syJ6jv4ld/9aVjxhZtsy+zYAx0N13X2Pu+92992NOl+cQQixvizr7Nab3vscgOfd/ZNLTN8EcFf2/i4A31j77gkh1oqVRL3dCuCDAJ4xsyezso8B+DiAr5jZ3QAOAvjAcg2ZAflKWE5oFnhX2iQn2EiZ53erR6SV4iLPI3b8BS5DnTgSvHnB6HiD1rGjJ6ltbiGSwy2Sqy0Prl91yL7V63ysqsO8/xvOhpcmAoBCJOqtRHLXzc7w9rqnIhJamS9R1c7zcSyVw+dVvc73ueI8Im5+hkdT/p8XfkBtO3ZyOW9s+5ZgeSeylBMNb4tUWdbZ3f1RgMY53rZcfSHEpYGeoBMiEeTsQiSCnF2IRJCzC5EIcnYhEqGvCSdhDiuGZZJKo0arNVtEWpmNJBo0Lk9NneYPAB6q8jbLCEdenT7F5bUFEnUFAMUal3+6FS5rFQsRGS0XFk7KVV4HVf6w08jICLV5NyZ5haWyoYjMNxN5wvLMcX7MFtmSVwDyRrSoLpfQNm7mUYyVsY3UVi7zSMtdV11BbWNbw0+axyI3W+3wPntEe9OVXYhEkLMLkQhydiESQc4uRCLI2YVIBDm7EInQX+ktZ8iTKCTjKgPcwjLUbGSNr3YkauzMXHhtLQA4Oc+jsqwVjiirRdbXurHIE/hUYokvI7H/GzcMUds4idhadD5WrTkuQ1VqvB/z81wqO3k6nHt0uMGlt9k2P2YLHW671urUtrkQHqtW5JidWOTnR3UbXxdv2zg/1o1RLmHOkEjA0jDfrwJJOMlj1nRlFyIZ5OxCJIKcXYhEkLMLkQhydiESob+z8QA6ZLbQcvwBfi+G63gtkisskt+tHVku6PQin5meITP1tsBniifm+Ox+e5HXaw5xeWLbZbuo7dptO4Plr7z4Eq1z+Dhfdml04yZqa0Rm1icnJ4PlYxs20DqLOX7tKVT4sX5ziQenXLUpPEM+PM5n1b81dZDajkyF9wsALiPbAoDpOZ6Dbm4+fM5t3LGN1jELj1VEZNCVXYhUkLMLkQhydiESQc4uRCLI2YVIBDm7EImwrPRmZrsAfAHAVgBdAHvc/dNmdj+A3wBwTrf5mLt/O9aWd4FuMyyJtVuLtF43H/6fVB7medWGR3dQW4sE1gDAbIf3Y5rIaM1I7rT2HLfNdvm2WnNclnvlxZep7fknnw23FwkManW4TDlxKhzQAgDl2JJMJGhoamqK1tmygx+zjZFTNZb7rXxtWL7aXOd55rZMvkptR1pcSj09M01tQ3UevFSphM/jVsQn3MNSdWT1pxXp7G0Av+3u3zOzBoAnzOyhzPYpd/83K2hDCDFgVrLW21EAR7P302b2PAD+L1gIcUlyQb/ZzewKADcBeCwr+qiZPW1mD5gZfzRKCDFwVuzsZjYE4KsA7nX3swA+A+BqADeid+X/BKl3j5ntNbO90zP896sQYn1ZkbObWRE9R/+Su38NANx9wt077t4F8FkAt4Tquvsed9/t7rsbQzzriRBifVnW2c3MAHwOwPPu/skl5UunOd8PIDwNLIS4JFjJbPytAD4I4BkzezIr+xiAO83sRvRm+/cD+M1lW3KgS9J7NSqRqz7J1daM6AzGlv0B0ChzyW57lUsyrcWwnNTs8KWmZiL5zCZO8GgzW+Rt5gtcOjw1F5Z/Zs9yWWhxgUs8lYhkVCzw02dmNixRFQtcJmtGpKZOl0e9WYlfs87mwsfszNkp3t40j3z0Cj8uJyLLgB14icul42PhCLzKEM9Bt+XycHRjTHxbyWz8owhHzkU1dSHEpYWeoBMiEeTsQiSCnF2IRJCzC5EIcnYhEqGvCSe7XWBuLhxhVSxySaZeCnezmI8kqezw1Ht547ZyJOnhcD0shTSdyzGlDh/ioUaN2gptvm+5HJfeDp88Hiw/FpEiOzUe9dYY5ssWlUpcDtvQGg2Wx2S+2U5YJgOA020+xpORJbtm9oUlr9k2P98mmrw9r/PjmSfRmQCQj8iUyIePJz8qAOg5rOWfhEgeObsQiSBnFyIR5OxCJIKcXYhEkLMLkQj9XevNcvBiWL6yMo/KYhFxkUCoqDxVzPOKEfUEXSINWZfLQkNF3o/qEE/uMzw0TG1zczwqa24xPI7dJm9vZoa3N1TjEtWmyDpwi81wpCLJkwgAmGvypJinJ/gaa0+dOkJtI/NhebNc5OfApHN5sFHdQm2FyEJrI2NcwhwhUW/lGpdmi8Vwsk+LyMq6sguRCHJ2IRJBzi5EIsjZhUgEObsQiSBnFyIR+iq95XI51EgCw4Uml6/g4f9JnuMyA5MmAKASSW4ZS1TZaoclmXyb96PTCUtQAGARHaodiQDL5fn2tm0NS0M7tofXPAOAU6f5em6dLu/jZbsup7ZSKTzGrUU+HtMRSfFAJPnisVcOUtt8M9ymGZf5cpFoxO0jDWobbvA+ViOJOxsjYemtFFlLLxLcRtGVXYhEkLMLkQhydiESQc4uRCLI2YVIhGVn482sAuA7AMrZ9/+zu/+BmY0B+FMAV6C3/NMvuzuf1gWQz+cxMhIOCGi3ecatZpMFJvApyVhAQKnEg1Ny0Rl+ogqAB4t0Ony/LB8Z/kguvHJklpbl8otN3m7YwANyhho8gGb7tu3UVqmEZ6ZbrdjySfz0qW8Kz1gDwK5dbCkkYG7qbLB8fpHPxg9t5OOxc/NWahur85n6asRWIMuRlUjuxR6RiCLCSq7sTQA/6+43oLc88+1m9jYA9wF42N2vBfBw9lkIcYmyrLN7j3PpNovZnwN4L4AHs/IHAbxvPToohFgbVro+ez5bwfU4gIfc/TEAW9z9KABkr5vXrZdCiFWzImd394673whgJ4BbzOzHV7oBM7vHzPaa2d4z0zwftxBifbmg2Xh3nwLwCIDbAUyY2TYAyF6DqxO4+x533+3uu0ca/JFBIcT6sqyzm9kmMxvN3lcBvAvADwB8E8Bd2dfuAvCNdeqjEGINWEkgzDYAD5pZHr1/Dl9x92+Z2d8A+IqZ3Q3gIIAPLNeQmaFAlsGp1XhwCpPD3Lms5R5Z/imy14UCl+VYm+1IIExM5itVecBFNxIks7jIc6Q5qddu8wCUYolLhzH5Z745T21tIjl6N5IjLRKEtHVsnNqu3MIlwBwJ5GlHzp1iNSyFAUAtIntWI3ntKjV+V5sjculCkwdDdbvsePIxXNbZ3f1pADcFyk8CuG25+kKISwM9QSdEIsjZhUgEObsQiSBnFyIR5OxCJIIxqWZdNmZ2AsCB7ONGAHxNn/6hfrwW9eO1vN76cbm7B9fl6quzv2bDZnvdffdANq5+qB8J9kO38UIkgpxdiEQYpLPvGeC2l6J+vBb147W8YfoxsN/sQoj+ott4IRJBzi5EIgzE2c3sdjN7wcx+aGYDS1RpZvvN7Bkze9LM9vZxuw+Y2XEze3ZJ2ZiZPWRm+7JXnuJ0fftxv5m9mo3Jk2Z2Rx/6scvM/peZPW9mz5nZP8rK+zomkX70dUzMrGJm3zWzp7J+/NOsfHXj4e59/QOQB/ASgKsAlAA8BeC6fvcj68t+ABsHsN2fBnAzgGeXlP0rAPdl7+8D8C8H1I/7AfxOn8djG4Cbs/cNAC8CuK7fYxLpR1/HBL3M30PZ+yKAxwC8bbXjMYgr+y0AfujuL7v7IoA/QS9TbTK4+3cAnDqvuO/Zekk/+o67H3X372XvpwE8D2AH+jwmkX70Fe+x5hmdB+HsOwAcWvL5MAYwoBkO4C/M7Akzu2dAfTjHpZSt96Nm9nR2m7/uPyeWYmZXoJcsZaAZjM/rB9DnMVmPjM6DcPZQXqJB6X+3uvvNAH4BwEfM7KcH1I9Lic8AuBq9BUGOAvhEvzZsZkMAvgrgXncPL+UymH70fUx8FRmdGYNw9sMAdi35vBPAkQH0A+5+JHs9DuDr6P3EGBQryta73rj7RHaidQF8Fn0aEzMroudgX3L3r2XFfR+TUD8GNSbZtqdwgRmdGYNw9scBXGtmV5pZCcCvopeptq+YWd3MGufeA/h5AM/Ga60rl0S23nMnU8b70Ycxsd7CfJ8D8Ly7f3KJqa9jwvrR7zFZt4zO/ZphPG+28Q70ZjpfAvB7A+rDVegpAU8BeK6f/QDwZfRuB1vo3encDWAcvTXz9mWvYwPqxxcBPAPg6ezk2taHfrwdvZ9yTwN4Mvu7o99jEulHX8cEwPUAvp9t71kA/yQrX9V46HFZIRJBT9AJkQhydiESQc4uRCLI2YVIBDm7EIkgZxciEeTsQiTC/wNcH+gL5phUYwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "label_names = load_label_names()\n",
    "correct_preds_indices = np.argwhere(y_hat_test == y_test).flatten()\n",
    "incorrect_preds_indices = np.argwhere(y_hat_test != y_test).flatten()\n",
    "np.random.seed(123)\n",
    "np.random.shuffle(correct_preds_indices)\n",
    "np.random.seed(123)\n",
    "np.random.shuffle(incorrect_preds_indices)\n",
    "\n",
    "for preds_idx in np.hstack([correct_preds_indices[:5], incorrect_preds_indices[:5]]):\n",
    "    plt.imshow(x_test_img[preds_idx])\n",
    "    \n",
    "    pred_label = label_names[y_hat_test[preds_idx]]\n",
    "    true_label = label_names[y_test[preds_idx]]\n",
    "    title = \\\n",
    "        f\"{pred_label} (correct)\" \\\n",
    "        if pred_label == true_label \\\n",
    "        else f\"{pred_label} (incorrect, true: {true_label})\"\n",
    "    plt.title(title) \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "employed-texture",
   "metadata": {},
   "source": [
    "## Hyperparameter Search <a class=\"anchor\" id=\"hyperparameter-search\"></a>\n",
    "\n",
    "Hyperparamter search is carried out to find the best combination of the newtwork depth (number of layers) and the L2 regularization rate. As in ```two-layer.ipynb```, the [hyperopt](#https://github.com/hyperopt/hyperopt), a Bayesian, distributed hyperparameter optimization library is used.\n",
    "\n",
    "Each run in the search goes on for 20 epochs. The best model is then re-trained for 30 epochs on the combination of the training and the validation (train-val) set, the test metrics of which are reported as a measure of the model performance on unseen data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "dirty-despite",
   "metadata": {},
   "outputs": [],
   "source": [
    "def objective(x_train, y_train, x_val, y_val, aug_func, batch_size, n_epochs, step_size, **kwargs):\n",
    "    seed = 12\n",
    "\n",
    "    in_dim = x_train.shape[1]\n",
    "    out_dim = 10\n",
    "\n",
    "    reg_rate = kwargs[\"reg_rate\"]\n",
    "    dims = kwargs[\"dims\"]\n",
    "\n",
    "    model, loss, dims = build_model(dims, reg_rate, in_dim, out_dim, seed)\n",
    "    model = compile_model(model, loss, step_size)\n",
    "\n",
    "    verbose = 1\n",
    "    history = model.fit(x_train, y_train, x_val, y_val, n_epochs, batch_size, verbose, aug_func)\n",
    "\n",
    "    params_val = {\"mode\": \"test\"}\n",
    "    scores_val = model.forward(x_val, **params_val)\n",
    "    y_hat_val = np.argmax(scores_val, axis=1)\n",
    "    metrics_val = model.compute_metrics(y_val, scores_val)\n",
    "    print(f\"val acc: {metrics_val}\")\n",
    "    val_acc = metrics_val['accuracy']\n",
    "\n",
    "    return {\n",
    "        'loss': -val_acc,\n",
    "        'status': STATUS_OK,\n",
    "        # -- store other results like this\n",
    "        'eval_time': time.time(),\n",
    "        'val_acc': val_acc,\n",
    "        'reg_rate': reg_rate,\n",
    "        'n_layers': len(dims) - 1,\n",
    "        'dims': dims\n",
    "    }\n",
    "\n",
    "\n",
    "def run_trials(x_train, y_train, x_val, y_val, aug_func, batch_size, n_epcohs, step_size, k_layer_hyperopt_path):\n",
    "    # how many additional trials to do after loading saved trials. 1 = save after iteration\n",
    "    trials_step = 1\n",
    "    \n",
    "    # initial max_trials. put something small to not have to wait\n",
    "    max_trials = 2\n",
    "\n",
    "    # try to load an already saved trials object, and increase the max\n",
    "    try:\n",
    "        trials = pickle.load(open(k_layer_hyperopt_path, \"rb\"))\n",
    "        print(\"Found saved Trials! Loading...\")\n",
    "        max_trials = len(trials.trials) + trials_step\n",
    "        print(\"Rerunning from {} trials to {} (+{}) trials\".format(len(trials.trials), max_trials, trials_step))\n",
    "        \n",
    "    # create a new trials object and start searching\n",
    "    except:\n",
    "        trials = Trials()\n",
    "\n",
    "    objective_lambda = lambda kwargs: objective(x_train, y_train, x_val, y_val, aug_func, \n",
    "                                                batch_size, n_epochs, step_size, **kwargs)\n",
    "\n",
    "    dims_options = [\n",
    "        (50, 30, 20, 10),\n",
    "        (50, 30, 20),\n",
    "        (50, 30)\n",
    "    ]\n",
    "\n",
    "    reg_limits = (0.01, 0.2)\n",
    "\n",
    "    space = {\n",
    "        \"reg_rate\": hp.uniform(\"reg_rate\", reg_limits[0], reg_limits[1]),\n",
    "        \"dims\": hp.choice(\"dims\", dims_options)\n",
    "    }\n",
    "\n",
    "    best = fmin(objective_lambda,\n",
    "                space=space,\n",
    "                algo=tpe.suggest,\n",
    "                max_evals=max_trials,\n",
    "                trials=trials)\n",
    "\n",
    "    print(\"Best:\", best)\n",
    "\n",
    "    # save the trials object\n",
    "    with open(k_layer_hyperopt_path, \"wb\") as f:\n",
    "        pickle.dump(trials, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "artistic-crime",
   "metadata": {},
   "source": [
    "Derive the number of epochs needed that is equivalent of the desired number of cylical learning rate schedule cylces. Here 20 epochs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "sporting-japan",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step size of cyc. lr: 2250 update steps\n",
      "full cycle of cyc.lr : 4500 update steps\n",
      "10.0 epochs = 1 full cycle = 4500 update steps\n",
      "2 cycle = 20 epochs = 9000 update steps\n"
     ]
    }
   ],
   "source": [
    "batch_size = 100\n",
    "\n",
    "step_size = int(5*np.floor(x_train.shape[0] / batch_size))\n",
    "print(f\"step size of cyc. lr: {step_size} update steps\")\n",
    "\n",
    "cycle_steps = 2*step_size\n",
    "print(f\"full cycle of cyc.lr : {cycle_steps} update steps\")\n",
    "\n",
    "epochs_one_full_cycle = (cycle_steps * batch_size) / x_train.shape[0]\n",
    "print(f\"{epochs_one_full_cycle} epochs = 1 full cycle = {cycle_steps} update steps\")\n",
    "\n",
    "n_cycle = 2\n",
    "n_epochs = int(n_cycle*epochs_one_full_cycle)\n",
    "print(f\"{n_cycle} cycle = {n_epochs} epochs = {n_cycle*cycle_steps} update steps\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "distant-wagon",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found saved Trials! Loading...\n",
      "Rerunning from 4 trials to 5 (+1) trials\n",
      "len(dims)-1 = 5                                      \n",
      "model summary:                                       \n",
      "layer 0: dense: \n",
      "\t shape -- in: 3072, out: 50\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.018042^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 1: batch norm with momentum 0.9\n",
      "layer 2: dropout with p=0.9\n",
      "layer 3: dense: \n",
      "\t shape -- in: 50, out: 30\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.141421^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 4: batch norm with momentum 0.9\n",
      "layer 5: dropout with p=0.9\n",
      "layer 6: dense: \n",
      "\t shape -- in: 30, out: 20\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.182574^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 7: batch norm with momentum 0.9\n",
      "layer 8: dropout with p=0.9\n",
      "layer 9: dense: \n",
      "\t shape -- in: 20, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.223607^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: softmax\n",
      "\n",
      "categorical cross-entropy loss with loss smoother constant\n",
      "adagrad with cycling lr schedule and clipper who does nothing\n",
      "\n",
      "starting epoch: 1 ...                                \n",
      "epoch 1/20                                           \n",
      " \t -- train loss = 1.774405111157633 / train cost = 2.4813961828990774\n",
      "\t -- {\"accuracy_train\": 0.3791333333333333} \n",
      "\t -- val loss = 1.7888701867640973 / val cost = 2.495861258505541\n",
      "\t -- {\"accuracy_val\": 0.3666} \n",
      "\n",
      "\n",
      "starting epoch: 2 ...                                \n",
      "epoch 2/20                                           \n",
      " \t -- train loss = 1.6267222218947366 / train cost = 1.9992487019996046\n",
      "\t -- {\"accuracy_train\": 0.42375555555555555} \n",
      "\t -- val loss = 1.6429013162116957 / val cost = 2.0154277963165637\n",
      "\t -- {\"accuracy_val\": 0.4164} \n",
      "\n",
      "\n",
      "starting epoch: 3 ...                                \n",
      "epoch 3/20                                           \n",
      " \t -- train loss = 1.5781315450590465 / train cost = 1.8141427410142472\n",
      "\t -- {\"accuracy_train\": 0.43924444444444444} \n",
      "\t -- val loss = 1.6032873845591562 / val cost = 1.839298580514357\n",
      "\t -- {\"accuracy_val\": 0.42} \n",
      "\n",
      "\n",
      "starting epoch: 4 ...                                \n",
      "epoch 4/20                                           \n",
      " \t -- train loss = 1.6194392845721424 / train cost = 1.8106219322057733\n",
      "\t -- {\"accuracy_train\": 0.41806666666666664} \n",
      "\t -- val loss = 1.6470920279100951 / val cost = 1.838274675543726\n",
      "\t -- {\"accuracy_val\": 0.4044} \n",
      "\n",
      "\n",
      "starting epoch: 5 ...                                \n",
      "epoch 5/20                                           \n",
      " \t -- train loss = 1.590886205131146 / train cost = 1.765022119875212\n",
      "\t -- {\"accuracy_train\": 0.43693333333333334} \n",
      "\t -- val loss = 1.6217096289323985 / val cost = 1.7958455436764644\n",
      "\t -- {\"accuracy_val\": 0.4284} \n",
      "\n",
      "\n",
      "starting epoch: 6 ...                                \n",
      "epoch 6/20                                           \n",
      " \t -- train loss = 1.580061367450204 / train cost = 1.7430916370256297\n",
      "\t -- {\"accuracy_train\": 0.4384222222222222} \n",
      "\t -- val loss = 1.6160944767440077 / val cost = 1.7791247463194335\n",
      "\t -- {\"accuracy_val\": 0.4184} \n",
      "\n",
      "\n",
      "starting epoch: 7 ...                                \n",
      "epoch 7/20                                           \n",
      " \t -- train loss = 1.4951630671801268 / train cost = 1.646233654054863\n",
      "\t -- {\"accuracy_train\": 0.4726444444444444} \n",
      "\t -- val loss = 1.5383790173517466 / val cost = 1.6894496042264828\n",
      "\t -- {\"accuracy_val\": 0.4512} \n",
      "\n",
      "\n",
      "starting epoch: 8 ...                                \n",
      "epoch 8/20                                           \n",
      " \t -- train loss = 1.4646525154289114 / train cost = 1.605496281862598\n",
      "\t -- {\"accuracy_train\": 0.47573333333333334} \n",
      "\t -- val loss = 1.518995387507456 / val cost = 1.6598391539411426\n",
      "\t -- {\"accuracy_val\": 0.4516} \n",
      "\n",
      "\n",
      "starting epoch: 9 ...                                \n",
      "epoch 9/20                                           \n",
      " \t -- train loss = 1.3919110125176435 / train cost = 1.522008397481196\n",
      "\t -- {\"accuracy_train\": 0.5089777777777778} \n",
      "\t -- val loss = 1.4624992361874265 / val cost = 1.592596621150979\n",
      "\t -- {\"accuracy_val\": 0.4778} \n",
      "\n",
      "\n",
      "starting epoch: 10 ...                               \n",
      "epoch 10/20                                          \n",
      " \t -- train loss = 1.3317672140149577 / train cost = 1.4537382962436178\n",
      "\t -- {\"accuracy_train\": 0.5356} \n",
      "\t -- val loss = 1.4204283959391435 / val cost = 1.5423994781678037\n",
      "\t -- {\"accuracy_val\": 0.4946} \n",
      "\n",
      "\n",
      "starting epoch: 11 ...                               \n",
      "epoch 11/20                                          \n",
      " \t -- train loss = 1.377146036183437 / train cost = 1.4985213935806763\n",
      "\t -- {\"accuracy_train\": 0.5107333333333334} \n",
      "\t -- val loss = 1.467185117153948 / val cost = 1.5885604745511874\n",
      "\t -- {\"accuracy_val\": 0.4768} \n",
      "\n",
      "\n",
      "starting epoch: 12 ...                               \n",
      "epoch 12/20                                          \n",
      " \t -- train loss = 1.4806419889373381 / train cost = 1.614534068282174\n",
      "\t -- {\"accuracy_train\": 0.47615555555555555} \n",
      "\t -- val loss = 1.5599129423920344 / val cost = 1.6938050217368703\n",
      "\t -- {\"accuracy_val\": 0.4446} \n",
      "\n",
      "\n",
      "starting epoch: 13 ...                               \n",
      "epoch 13/20                                          \n",
      " \t -- train loss = 1.4658285224142455 / train cost = 1.610929632223554\n",
      "\t -- {\"accuracy_train\": 0.47802222222222224} \n",
      "\t -- val loss = 1.5275650713666884 / val cost = 1.6726661811759969\n",
      "\t -- {\"accuracy_val\": 0.455} \n",
      "\n",
      "\n",
      "starting epoch: 14 ...                               \n",
      "epoch 14/20                                          \n",
      " \t -- train loss = 1.472532449531246 / train cost = 1.6259820458118244\n",
      "\t -- {\"accuracy_train\": 0.4802444444444444} \n",
      "\t -- val loss = 1.5263064145615879 / val cost = 1.6797560108421663\n",
      "\t -- {\"accuracy_val\": 0.4548} \n",
      "\n",
      "\n",
      "starting epoch: 15 ...                               \n",
      "epoch 15/20                                          \n",
      " \t -- train loss = 1.5199752726400984 / train cost = 1.67432826892833\n",
      "\t -- {\"accuracy_train\": 0.45768888888888887} \n",
      "\t -- val loss = 1.5651119987491882 / val cost = 1.71946499503742\n",
      "\t -- {\"accuracy_val\": 0.433} \n",
      "\n",
      "\n",
      "starting epoch: 16 ...                               \n",
      "epoch 16/20                                          \n",
      " \t -- train loss = 1.472022248023878 / train cost = 1.6211602070961189\n",
      "\t -- {\"accuracy_train\": 0.4767111111111111} \n",
      "\t -- val loss = 1.5175863572199682 / val cost = 1.666724316292209\n",
      "\t -- {\"accuracy_val\": 0.455} \n",
      "\n",
      "\n",
      "starting epoch: 17 ...                               \n",
      "epoch 17/20                                          \n",
      " \t -- train loss = 1.4615313413682811 / train cost = 1.6029453128509281\n",
      "\t -- {\"accuracy_train\": 0.48464444444444443} \n",
      "\t -- val loss = 1.5183380723850572 / val cost = 1.6597520438677043\n",
      "\t -- {\"accuracy_val\": 0.4608} \n",
      "\n",
      "\n",
      "starting epoch: 18 ...                               \n",
      "epoch 18/20                                          \n",
      " \t -- train loss = 1.406385744727522 / train cost = 1.5422200231479715\n",
      "\t -- {\"accuracy_train\": 0.5030222222222223} \n",
      "\t -- val loss = 1.4760485889472215 / val cost = 1.6118828673676708\n",
      "\t -- {\"accuracy_val\": 0.4732} \n",
      "\n",
      "\n",
      "starting epoch: 19 ...                               \n",
      "epoch 19/20                                          \n",
      " \t -- train loss = 1.3318787460614612 / train cost = 1.4571390915312785\n",
      "\t -- {\"accuracy_train\": 0.5346222222222222} \n",
      "\t -- val loss = 1.4156201597666138 / val cost = 1.5408805052364312\n",
      "\t -- {\"accuracy_val\": 0.4962} \n",
      "\n",
      "\n",
      "starting epoch: 20 ...                               \n",
      "epoch 20/20                                          \n",
      " \t -- train loss = 1.2776771996223197 / train cost = 1.3955971064446726\n",
      "\t -- {\"accuracy_train\": 0.5536666666666666} \n",
      "\t -- val loss = 1.3778240890332003 / val cost = 1.4957439958555532\n",
      "\t -- {\"accuracy_val\": 0.506} \n",
      "\n",
      "\n",
      "val acc: {'accuracy': 0.506}                         \n",
      "100%|██████████| 5/5 [03:39<00:00, 219.46s/trial, best loss: -0.506]\n",
      "Best: {'dims': 1, 'reg_rate': 0.018442837925266218}\n",
      "Found saved Trials! Loading...\n",
      "Rerunning from 5 trials to 6 (+1) trials\n",
      "len(dims)-1 = 5                                      \n",
      "model summary:                                       \n",
      "layer 0: dense: \n",
      "\t shape -- in: 3072, out: 50\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.018042^2), reg: l2 with 1.8654e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 1: batch norm with momentum 0.9\n",
      "layer 2: dropout with p=0.9\n",
      "layer 3: dense: \n",
      "\t shape -- in: 50, out: 30\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.141421^2), reg: l2 with 1.8654e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 4: batch norm with momentum 0.9\n",
      "layer 5: dropout with p=0.9\n",
      "layer 6: dense: \n",
      "\t shape -- in: 30, out: 20\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.182574^2), reg: l2 with 1.8654e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 7: batch norm with momentum 0.9\n",
      "layer 8: dropout with p=0.9\n",
      "layer 9: dense: \n",
      "\t shape -- in: 20, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.223607^2), reg: l2 with 1.8654e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 10: batch norm with momentum 0.9\n",
      "layer 11: dropout with p=0.9\n",
      "layer 12: dense: \n",
      "\t shape -- in: 10, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.316228^2), reg: l2 with 1.8654e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: softmax\n",
      "\n",
      "categorical cross-entropy loss with loss smoother constant\n",
      "adagrad with cycling lr schedule and clipper who does nothing\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "starting epoch: 1 ...                                \n",
      "epoch 1/20                                           \n",
      " \t -- train loss = 2.1021013615327493 / train cost = 3.0582299457081477\n",
      "\t -- {\"accuracy_train\": 0.23457777777777777} \n",
      "\t -- val loss = 2.106742619084735 / val cost = 3.0628712032601335\n",
      "\t -- {\"accuracy_val\": 0.2358} \n",
      "\n",
      "\n",
      "starting epoch: 2 ...                                \n",
      "epoch 2/20                                           \n",
      " \t -- train loss = 1.7836528298871188 / train cost = 2.3952806592901648\n",
      "\t -- {\"accuracy_train\": 0.35402222222222224} \n",
      "\t -- val loss = 1.8023084626336503 / val cost = 2.4139362920366962\n",
      "\t -- {\"accuracy_val\": 0.347} \n",
      "\n",
      "\n",
      "starting epoch: 3 ...                                \n",
      "epoch 3/20                                           \n",
      " \t -- train loss = 1.7134337045425287 / train cost = 2.0895901009301285\n",
      "\t -- {\"accuracy_train\": 0.38008888888888887} \n",
      "\t -- val loss = 1.737523955099472 / val cost = 2.113680351487072\n",
      "\t -- {\"accuracy_val\": 0.3622} \n",
      "\n",
      "\n",
      "starting epoch: 4 ...                                \n",
      "epoch 4/20                                           \n",
      " \t -- train loss = 1.6557199816404828 / train cost = 1.9060999489963175\n",
      "\t -- {\"accuracy_train\": 0.40644444444444444} \n",
      "\t -- val loss = 1.6798484111549252 / val cost = 1.93022837851076\n",
      "\t -- {\"accuracy_val\": 0.398} \n",
      "\n",
      "\n",
      "starting epoch: 5 ...                                \n",
      "epoch 5/20                                           \n",
      " \t -- train loss = 1.6798116384601978 / train cost = 2.2352282915827897\n",
      "\t -- {\"accuracy_train\": 0.39595555555555556} \n",
      "\t -- val loss = 1.6984114057529318 / val cost = 2.2538280588755235\n",
      "\t -- {\"accuracy_val\": 0.3928} \n",
      "\n",
      "\n",
      "starting epoch: 6 ...                                \n",
      "epoch 6/20                                           \n",
      " \t -- train loss = 1.6161775954598923 / train cost = 1.879445159850754\n",
      "\t -- {\"accuracy_train\": 0.42715555555555557} \n",
      "\t -- val loss = 1.6406837928350044 / val cost = 1.9039513572258662\n",
      "\t -- {\"accuracy_val\": 0.4198} \n",
      "\n",
      "\n",
      "starting epoch: 7 ...                                \n",
      "epoch 7/20                                           \n",
      " \t -- train loss = 1.5620041629178352 / train cost = 1.7495724753315036\n",
      "\t -- {\"accuracy_train\": 0.4477333333333333} \n",
      "\t -- val loss = 1.6003585255170931 / val cost = 1.7879268379307616\n",
      "\t -- {\"accuracy_val\": 0.4348} \n",
      "\n",
      "\n",
      "starting epoch: 8 ...                                \n",
      "epoch 8/20                                           \n",
      " \t -- train loss = 1.5148819314460014 / train cost = 1.674080069411678\n",
      "\t -- {\"accuracy_train\": 0.4582} \n",
      "\t -- val loss = 1.5532772613468733 / val cost = 1.7124753993125499\n",
      "\t -- {\"accuracy_val\": 0.4444} \n",
      "\n",
      "\n",
      "starting epoch: 9 ...                                \n",
      "epoch 9/20                                           \n",
      " \t -- train loss = 1.4727661471645284 / train cost = 1.617067955661569\n",
      "\t -- {\"accuracy_train\": 0.4829777777777778} \n",
      "\t -- val loss = 1.544998092436698 / val cost = 1.6892999009337386\n",
      "\t -- {\"accuracy_val\": 0.461} \n",
      "\n",
      "\n",
      "starting epoch: 10 ...                               \n",
      "epoch 10/20                                          \n",
      " \t -- train loss = 1.4098741802783805 / train cost = 1.547664162134176\n",
      "\t -- {\"accuracy_train\": 0.5017555555555555} \n",
      "\t -- val loss = 1.4771409143819734 / val cost = 1.6149308962377689\n",
      "\t -- {\"accuracy_val\": 0.4744} \n",
      "\n",
      "\n",
      "starting epoch: 11 ...                               \n",
      "epoch 11/20                                          \n",
      " \t -- train loss = 1.4678873615945986 / train cost = 1.6021100605561298\n",
      "\t -- {\"accuracy_train\": 0.4892444444444444} \n",
      "\t -- val loss = 1.5440627465145031 / val cost = 1.6782854454760343\n",
      "\t -- {\"accuracy_val\": 0.4644} \n",
      "\n",
      "\n",
      "starting epoch: 12 ...                               \n",
      "epoch 12/20                                          \n",
      " \t -- train loss = 1.4860542353009862 / train cost = 1.6217626381227233\n",
      "\t -- {\"accuracy_train\": 0.4734888888888889} \n",
      "\t -- val loss = 1.5645719954330983 / val cost = 1.7002803982548353\n",
      "\t -- {\"accuracy_val\": 0.4488} \n",
      "\n",
      "\n",
      "starting epoch: 13 ...                               \n",
      "epoch 13/20                                          \n",
      " \t -- train loss = 1.5129081814577023 / train cost = 1.6526109490719458\n",
      "\t -- {\"accuracy_train\": 0.464} \n",
      "\t -- val loss = 1.6217781484734723 / val cost = 1.7614809160877158\n",
      "\t -- {\"accuracy_val\": 0.4484} \n",
      "\n",
      "\n",
      "starting epoch: 14 ...                               \n",
      "epoch 14/20                                          \n",
      " \t -- train loss = 1.5167694842511432 / train cost = 1.6593498185666735\n",
      "\t -- {\"accuracy_train\": 0.4573111111111111} \n",
      "\t -- val loss = 1.5663651479384146 / val cost = 1.708945482253945\n",
      "\t -- {\"accuracy_val\": 0.4412} \n",
      "\n",
      "\n",
      "starting epoch: 15 ...                               \n",
      "epoch 15/20                                          \n",
      " \t -- train loss = 1.5291729774425837 / train cost = 1.6756532949046734\n",
      "\t -- {\"accuracy_train\": 0.46197777777777776} \n",
      "\t -- val loss = 1.569090049182069 / val cost = 1.7155703666441586\n",
      "\t -- {\"accuracy_val\": 0.4388} \n",
      "\n",
      "\n",
      "starting epoch: 16 ...                               \n",
      "epoch 16/20                                          \n",
      " \t -- train loss = 1.4712759439107808 / train cost = 1.613140923042416\n",
      "\t -- {\"accuracy_train\": 0.48257777777777777} \n",
      "\t -- val loss = 1.5116289629823765 / val cost = 1.6534939421140118\n",
      "\t -- {\"accuracy_val\": 0.4672} \n",
      "\n",
      "\n",
      "starting epoch: 17 ...                               \n",
      "epoch 17/20                                          \n",
      " \t -- train loss = 1.448323380859467 / train cost = 1.5838397235717676\n",
      "\t -- {\"accuracy_train\": 0.4890888888888889} \n",
      "\t -- val loss = 1.4897061235566051 / val cost = 1.6252224662689057\n",
      "\t -- {\"accuracy_val\": 0.4696} \n",
      "\n",
      "\n",
      "starting epoch: 18 ...                               \n",
      "epoch 18/20                                          \n",
      " \t -- train loss = 1.4305232538693167 / train cost = 1.5596709753925295\n",
      "\t -- {\"accuracy_train\": 0.49033333333333334} \n",
      "\t -- val loss = 1.4912367151211319 / val cost = 1.6203844366443447\n",
      "\t -- {\"accuracy_val\": 0.4594} \n",
      "\n",
      "\n",
      "starting epoch: 19 ...                               \n",
      "epoch 19/20                                          \n",
      " \t -- train loss = 1.3653877761547288 / train cost = 1.4875560144243414\n",
      "\t -- {\"accuracy_train\": 0.5169333333333334} \n",
      "\t -- val loss = 1.4406237763968455 / val cost = 1.562792014666458\n",
      "\t -- {\"accuracy_val\": 0.4822} \n",
      "\n",
      "\n",
      "starting epoch: 20 ...                               \n",
      "epoch 20/20                                          \n",
      " \t -- train loss = 1.3117294655039649 / train cost = 1.4292029851426236\n",
      "\t -- {\"accuracy_train\": 0.5388888888888889} \n",
      "\t -- val loss = 1.4014413833044579 / val cost = 1.5189149029431166\n",
      "\t -- {\"accuracy_val\": 0.4994} \n",
      "\n",
      "\n",
      "val acc: {'accuracy': 0.4994}                        \n",
      "100%|██████████| 6/6 [03:47<00:00, 227.19s/trial, best loss: -0.506]\n",
      "Best: {'dims': 1, 'reg_rate': 0.018442837925266218}\n",
      "Found saved Trials! Loading...\n",
      "Rerunning from 6 trials to 7 (+1) trials\n",
      "len(dims)-1 = 5                                      \n",
      "model summary:                                       \n",
      "layer 0: dense: \n",
      "\t shape -- in: 3072, out: 50\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.018042^2), reg: l2 with 5.5519e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 1: batch norm with momentum 0.9\n",
      "layer 2: dropout with p=0.9\n",
      "layer 3: dense: \n",
      "\t shape -- in: 50, out: 30\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.141421^2), reg: l2 with 5.5519e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 4: batch norm with momentum 0.9\n",
      "layer 5: dropout with p=0.9\n",
      "layer 6: dense: \n",
      "\t shape -- in: 30, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.182574^2), reg: l2 with 5.5519e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: softmax\n",
      "\n",
      "categorical cross-entropy loss with loss smoother constant\n",
      "adagrad with cycling lr schedule and clipper who does nothing\n",
      "\n",
      "starting epoch: 1 ...                                \n",
      "epoch 1/20                                           \n",
      " \t -- train loss = 1.7142594452179516 / train cost = 2.4172014470401564\n",
      "\t -- {\"accuracy_train\": 0.3977777777777778} \n",
      "\t -- val loss = 1.7340901127250181 / val cost = 2.437032114547223\n",
      "\t -- {\"accuracy_val\": 0.3888} \n",
      "\n",
      "\n",
      "starting epoch: 2 ...                                \n",
      "epoch 2/20                                           \n",
      " \t -- train loss = 1.6821017089137025 / train cost = 1.9323606165330272\n",
      "\t -- {\"accuracy_train\": 0.4086444444444444} \n",
      "\t -- val loss = 1.7011519602470584 / val cost = 1.951410867866383\n",
      "\t -- {\"accuracy_val\": 0.4038} \n",
      "\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "starting epoch: 3 ...                                \n",
      "epoch 3/20                                           \n",
      " \t -- train loss = 1.6907396040693707 / train cost = 1.889546929998687\n",
      "\t -- {\"accuracy_train\": 0.39797777777777776} \n",
      "\t -- val loss = 1.70294566444503 / val cost = 1.9017529903743462\n",
      "\t -- {\"accuracy_val\": 0.3906} \n",
      "\n",
      "\n",
      "starting epoch: 4 ...                                \n",
      "epoch 4/20                                           \n",
      " \t -- train loss = 1.7102590267220654 / train cost = 1.9011752484954587\n",
      "\t -- {\"accuracy_train\": 0.39613333333333334} \n",
      "\t -- val loss = 1.7228192430526914 / val cost = 1.9137354648260847\n",
      "\t -- {\"accuracy_val\": 0.3872} \n",
      "\n",
      "\n",
      "starting epoch: 5 ...                                \n",
      "epoch 5/20                                           \n",
      " \t -- train loss = 1.7285709641612816 / train cost = 1.9186808677627272\n",
      "\t -- {\"accuracy_train\": 0.3796888888888889} \n",
      "\t -- val loss = 1.747420915898851 / val cost = 1.9375308195002965\n",
      "\t -- {\"accuracy_val\": 0.3666} \n",
      "\n",
      "\n",
      "starting epoch: 6 ...                                \n",
      "epoch 6/20                                           \n",
      " \t -- train loss = 1.683267882151922 / train cost = 1.8633916523100893\n",
      "\t -- {\"accuracy_train\": 0.3950666666666667} \n",
      "\t -- val loss = 1.700260240717747 / val cost = 1.8803840108759142\n",
      "\t -- {\"accuracy_val\": 0.3902} \n",
      "\n",
      "\n",
      "starting epoch: 7 ...                                \n",
      "epoch 7/20                                           \n",
      " \t -- train loss = 1.634655988840947 / train cost = 1.802039488695656\n",
      "\t -- {\"accuracy_train\": 0.41631111111111113} \n",
      "\t -- val loss = 1.6478749484816904 / val cost = 1.8152584483363996\n",
      "\t -- {\"accuracy_val\": 0.4194} \n",
      "\n",
      "\n",
      "starting epoch: 8 ...                                \n",
      "epoch 8/20                                           \n",
      " \t -- train loss = 1.5835279931773802 / train cost = 1.7445582095704246\n",
      "\t -- {\"accuracy_train\": 0.43777777777777777} \n",
      "\t -- val loss = 1.6096054878337323 / val cost = 1.7706357042267766\n",
      "\t -- {\"accuracy_val\": 0.425} \n",
      "\n",
      "\n",
      "starting epoch: 9 ...                                \n",
      "epoch 9/20                                           \n",
      " \t -- train loss = 1.5445385137878322 / train cost = 1.6857010462272684\n",
      "\t -- {\"accuracy_train\": 0.4570222222222222} \n",
      "\t -- val loss = 1.576253073080107 / val cost = 1.717415605519543\n",
      "\t -- {\"accuracy_val\": 0.4444} \n",
      "\n",
      "\n",
      "starting epoch: 10 ...                               \n",
      "epoch 10/20                                          \n",
      " \t -- train loss = 1.429311326314984 / train cost = 1.5538636837216018\n",
      "\t -- {\"accuracy_train\": 0.5001333333333333} \n",
      "\t -- val loss = 1.4763436917249853 / val cost = 1.600896049131603\n",
      "\t -- {\"accuracy_val\": 0.4778} \n",
      "\n",
      "\n",
      "starting epoch: 11 ...                               \n",
      "epoch 11/20                                          \n",
      " \t -- train loss = 1.5124418390867018 / train cost = 1.6515619899002727\n",
      "\t -- {\"accuracy_train\": 0.4636222222222222} \n",
      "\t -- val loss = 1.5598448885138108 / val cost = 1.6989650393273816\n",
      "\t -- {\"accuracy_val\": 0.4466} \n",
      "\n",
      "\n",
      "starting epoch: 12 ...                               \n",
      "epoch 12/20                                          \n",
      " \t -- train loss = 1.5415757372075987 / train cost = 1.7022738960413772\n",
      "\t -- {\"accuracy_train\": 0.4519111111111111} \n",
      "\t -- val loss = 1.5761715406511498 / val cost = 1.7368696994849284\n",
      "\t -- {\"accuracy_val\": 0.4378} \n",
      "\n",
      "\n",
      "starting epoch: 13 ...                               \n",
      "epoch 13/20                                          \n",
      " \t -- train loss = 1.5807527950906555 / train cost = 1.7511526678614249\n",
      "\t -- {\"accuracy_train\": 0.4326888888888889} \n",
      "\t -- val loss = 1.6098360121192852 / val cost = 1.7802358848900546\n",
      "\t -- {\"accuracy_val\": 0.4204} \n",
      "\n",
      "\n",
      "starting epoch: 14 ...                               \n",
      "epoch 14/20                                          \n",
      " \t -- train loss = 1.6170514279126973 / train cost = 1.7942149553740998\n",
      "\t -- {\"accuracy_train\": 0.4239111111111111} \n",
      "\t -- val loss = 1.6394551266722577 / val cost = 1.8166186541336602\n",
      "\t -- {\"accuracy_val\": 0.4148} \n",
      "\n",
      "\n",
      "starting epoch: 15 ...                               \n",
      "epoch 15/20                                          \n",
      " \t -- train loss = 1.6566832791266601 / train cost = 1.837815982745443\n",
      "\t -- {\"accuracy_train\": 0.4082222222222222} \n",
      "\t -- val loss = 1.6865635829964016 / val cost = 1.8676962866151845\n",
      "\t -- {\"accuracy_val\": 0.3942} \n",
      "\n",
      "\n",
      "starting epoch: 16 ...                               \n",
      "epoch 16/20                                          \n",
      " \t -- train loss = 1.608256946419352 / train cost = 1.778592641566303\n",
      "\t -- {\"accuracy_train\": 0.42146666666666666} \n",
      "\t -- val loss = 1.6439345880653542 / val cost = 1.8142702832123052\n",
      "\t -- {\"accuracy_val\": 0.4062} \n",
      "\n",
      "\n",
      "starting epoch: 17 ...                               \n",
      "epoch 17/20                                          \n",
      " \t -- train loss = 1.5558541752350623 / train cost = 1.7162429136953248\n",
      "\t -- {\"accuracy_train\": 0.45197777777777776} \n",
      "\t -- val loss = 1.5882206952853233 / val cost = 1.7486094337455858\n",
      "\t -- {\"accuracy_val\": 0.4306} \n",
      "\n",
      "\n",
      "starting epoch: 18 ...                               \n",
      "epoch 18/20                                          \n",
      " \t -- train loss = 1.534604727761409 / train cost = 1.6828166761740933\n",
      "\t -- {\"accuracy_train\": 0.4560666666666667} \n",
      "\t -- val loss = 1.5687310189404802 / val cost = 1.7169429673531644\n",
      "\t -- {\"accuracy_val\": 0.4378} \n",
      "\n",
      "\n",
      "starting epoch: 19 ...                               \n",
      "epoch 19/20                                          \n",
      " \t -- train loss = 1.4636207023157952 / train cost = 1.5992163077725923\n",
      "\t -- {\"accuracy_train\": 0.4820888888888889} \n",
      "\t -- val loss = 1.5033494727102872 / val cost = 1.6389450781670842\n",
      "\t -- {\"accuracy_val\": 0.4582} \n",
      "\n",
      "\n",
      "starting epoch: 20 ...                               \n",
      "epoch 20/20                                          \n",
      " \t -- train loss = 1.3831506985688062 / train cost = 1.5038485595985105\n",
      "\t -- {\"accuracy_train\": 0.5146666666666667} \n",
      "\t -- val loss = 1.4452773153284815 / val cost = 1.5659751763581857\n",
      "\t -- {\"accuracy_val\": 0.488} \n",
      "\n",
      "\n",
      "val acc: {'accuracy': 0.488}                         \n",
      "100%|██████████| 7/7 [03:29<00:00, 209.31s/trial, best loss: -0.506]\n",
      "Best: {'dims': 1, 'reg_rate': 0.018442837925266218}\n"
     ]
    }
   ],
   "source": [
    "# loop indefinitely and stop whenever you like\n",
    "run_for = 3\n",
    "k_layer_hyperopt_path = \"assets/k_layer/klayer.hyperopt\"\n",
    "\n",
    "for i in range(run_for):\n",
    "    run_trials(x_train, y_train, x_val, y_val, aug_func, batch_size, n_epochs, step_size, k_layer_hyperopt_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fitting-egyptian",
   "metadata": {},
   "source": [
    "Load the best performing model with the network depth and L2 regularization stregth found from the hyperparameter search. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "plain-understanding",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7\n",
      "[{'loss': -0.5004, 'status': 'ok', 'eval_time': 1629915934.5592592, 'val_acc': 0.5004, 'reg_rate': 0.019240294501572188, 'n_layers': 5, 'dims': [3072, 50, 30, 20, 10, 10]}, {'loss': -0.455, 'status': 'ok', 'eval_time': 1629916160.4795947, 'val_acc': 0.455, 'reg_rate': 0.13214421822496275, 'n_layers': 4, 'dims': [3072, 50, 30, 20, 10]}, {'loss': -0.4572, 'status': 'ok', 'eval_time': 1629916392.15205, 'val_acc': 0.4572, 'reg_rate': 0.18615281251620688, 'n_layers': 3, 'dims': [3072, 50, 30, 10]}, {'loss': -0.4862, 'status': 'ok', 'eval_time': 1629916637.427237, 'val_acc': 0.4862, 'reg_rate': 0.06244968837267639, 'n_layers': 3, 'dims': [3072, 50, 30, 10]}, {'loss': -0.506, 'status': 'ok', 'eval_time': 1631284162.8050916, 'val_acc': 0.506, 'reg_rate': 0.018442837925266218, 'n_layers': 4, 'dims': [3072, 50, 30, 20, 10]}, {'loss': -0.4994, 'status': 'ok', 'eval_time': 1631284389.9962392, 'val_acc': 0.4994, 'reg_rate': 0.01865445340668206, 'n_layers': 5, 'dims': [3072, 50, 30, 20, 10, 10]}, {'loss': -0.488, 'status': 'ok', 'eval_time': 1631284599.3187609, 'val_acc': 0.488, 'reg_rate': 0.05551860506740719, 'n_layers': 3, 'dims': [3072, 50, 30, 10]}]\n",
      "best_trial\n",
      "{'loss': -0.506, 'status': 'ok', 'eval_time': 1631284162.8050916, 'val_acc': 0.506, 'reg_rate': 0.018442837925266218, 'n_layers': 4, 'dims': [3072, 50, 30, 20, 10]}\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>val_acc</th>\n",
       "      <th>reg_rate</th>\n",
       "      <th>n_layers</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.5004</td>\n",
       "      <td>0.019240</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.4550</td>\n",
       "      <td>0.132144</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.4572</td>\n",
       "      <td>0.186153</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.4862</td>\n",
       "      <td>0.062450</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.5060</td>\n",
       "      <td>0.018443</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.4994</td>\n",
       "      <td>0.018654</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.4880</td>\n",
       "      <td>0.055519</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   val_acc  reg_rate  n_layers\n",
       "0   0.5004  0.019240         5\n",
       "1   0.4550  0.132144         4\n",
       "2   0.4572  0.186153         3\n",
       "3   0.4862  0.062450         3\n",
       "4   0.5060  0.018443         4\n",
       "5   0.4994  0.018654         5\n",
       "6   0.4880  0.055519         3"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "k_layer_hyperopt_csv_path = \"assets/k_layer/klayer.csv\"\n",
    "\n",
    "trials = pickle.load(open(k_layer_hyperopt_path, \"rb\"))\n",
    "print(len(trials.trials))\n",
    "print(trials.results)\n",
    "\n",
    "best_trial = max(trials.results, key=lambda x: x['val_acc'])\n",
    "print(\"best_trial\")\n",
    "print(best_trial)\n",
    "\n",
    "df = pd.DataFrame(trials.results)\n",
    "df[\"val_acc\"] = df[\"loss\"] * -1\n",
    "df = df.drop(['status', 'eval_time', 'loss', 'dims'], 1)\n",
    "df.to_csv(k_layer_hyperopt_csv_path)\n",
    "\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "black-south",
   "metadata": {},
   "source": [
    "Derive the number of epochs needed that is equivalent of the desired number of cylical learning rate schedule cylces. Here 30 epochs and on the combination of the train-val dataset. accuracy_val is irrelevant, as we only test on the unseen test set after training on the combined train-val dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "sonic-fight",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "trainval: x_train_val.shape = (50000, 3072), y_train_val.shape = (50000,)\n",
      "step size of cyc. lr: 2500 update steps\n",
      "full cycle of cyc.lr : 5000 update steps\n",
      "10.0 epochs = 1 full cycle = 5000 update steps\n",
      "3 cycle = 30 epochs = 15000 update steps\n"
     ]
    }
   ],
   "source": [
    "batch_size = 100\n",
    "\n",
    "# train best model on train-val\n",
    "x_train_val = np.vstack([x_train, x_val])\n",
    "y_train_val = np.hstack([y_train, y_val])\n",
    "print(f\"trainval: x_train_val.shape = {x_train_val.shape}, y_train_val.shape = {y_train_val.shape}\")\n",
    "\n",
    "step_size = int(5*np.floor(x_train_val.shape[0] / batch_size))\n",
    "print(f\"step size of cyc. lr: {step_size} update steps\")\n",
    "\n",
    "cycle_steps = 2*step_size\n",
    "print(f\"full cycle of cyc.lr : {cycle_steps} update steps\")\n",
    "\n",
    "epochs_one_full_cycle = (cycle_steps * batch_size) / x_train_val.shape[0]\n",
    "print(f\"{epochs_one_full_cycle} epochs = 1 full cycle = {cycle_steps} update steps\")\n",
    "\n",
    "n_cycle = 3\n",
    "n_epochs = int(n_cycle*epochs_one_full_cycle)\n",
    "print(f\"{n_cycle} cycle = {n_epochs} epochs = {n_cycle*cycle_steps} update steps\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "spiritual-catalog",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "len(dims)-1 = 4\n",
      "model summary: \n",
      "layer 0: dense: \n",
      "\t shape -- in: 3072, out: 50\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.018042^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 1: batch norm with momentum 0.9\n",
      "layer 2: dropout with p=0.9\n",
      "layer 3: dense: \n",
      "\t shape -- in: 50, out: 30\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.141421^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 4: batch norm with momentum 0.9\n",
      "layer 5: dropout with p=0.9\n",
      "layer 6: dense: \n",
      "\t shape -- in: 30, out: 20\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.182574^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: relu\n",
      "\n",
      "layer 7: batch norm with momentum 0.9\n",
      "layer 8: dropout with p=0.9\n",
      "layer 9: dense: \n",
      "\t shape -- in: 20, out: 10\n",
      "\t w -- init: Xavier ~ 1.000000 x N(0.000000, 0.223607^2), reg: l2 with 1.8443e-02\n",
      "\t b -- init: Xavier ~ 1.000000 x N(0.000000, 1.000000^2)\n",
      "\t activation: softmax\n",
      "\n",
      "categorical cross-entropy loss with loss smoother constant\n",
      "adagrad with cycling lr schedule and clipper who does nothing\n",
      "\n",
      "starting epoch: 1 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.39it/s]\n",
      "epoch 1/30 \n",
      " \t -- train loss = 1.757624619067456 / train cost = 2.4286688258051843\n",
      "\t -- {\"accuracy_train\": 0.38288} \n",
      "\t -- val loss = 2.231411155472021 / val cost = 2.9024553622097495\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 2 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.44it/s]\n",
      "epoch 2/30 \n",
      " \t -- train loss = 1.633041066748714 / train cost = 1.9671275408516875\n",
      "\t -- {\"accuracy_train\": 0.41976} \n",
      "\t -- val loss = 2.056145506586061 / val cost = 2.3902319806890344\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 3 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.00it/s]\n",
      "epoch 3/30 \n",
      " \t -- train loss = 1.6035331700790423 / train cost = 1.8212300726789237\n",
      "\t -- {\"accuracy_train\": 0.4316} \n",
      "\t -- val loss = 1.8346852035467864 / val cost = 2.0523821061466676\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 4 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.30it/s]\n",
      "epoch 4/30 \n",
      " \t -- train loss = 1.6488385312126352 / train cost = 1.8607579425056362\n",
      "\t -- {\"accuracy_train\": 0.41736} \n",
      "\t -- val loss = 2.251073153558953 / val cost = 2.462992564851954\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 5 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.80it/s]\n",
      "epoch 5/30 \n",
      " \t -- train loss = 1.618034829205762 / train cost = 1.79115690931404\n",
      "\t -- {\"accuracy_train\": 0.4181} \n",
      "\t -- val loss = 2.539710429538113 / val cost = 2.712832509646391\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 6 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.62it/s]\n",
      "epoch 6/30 \n",
      " \t -- train loss = 1.5833933234582926 / train cost = 1.7441961596980629\n",
      "\t -- {\"accuracy_train\": 0.43286} \n",
      "\t -- val loss = 2.2645144612904953 / val cost = 2.4253172975302655\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 7 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.76it/s]\n",
      "epoch 7/30 \n",
      " \t -- train loss = 1.5339264621354658 / train cost = 1.6847510026953558\n",
      "\t -- {\"accuracy_train\": 0.45104} \n",
      "\t -- val loss = 2.324648581234165 / val cost = 2.475473121794055\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 8 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.51it/s]\n",
      "epoch 8/30 \n",
      " \t -- train loss = 1.4768486786888604 / train cost = 1.618781263661484\n",
      "\t -- {\"accuracy_train\": 0.47674} \n",
      "\t -- val loss = 2.1078714977031123 / val cost = 2.249804082675736\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 9 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.08it/s]\n",
      "epoch 9/30 \n",
      " \t -- train loss = 1.4180109373487244 / train cost = 1.5488295296671446\n",
      "\t -- {\"accuracy_train\": 0.4973} \n",
      "\t -- val loss = 1.5846584412904607 / val cost = 1.715477033608881\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 10 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.05it/s]\n",
      "epoch 10/30 \n",
      " \t -- train loss = 1.3451126426636852 / train cost = 1.465901141646505\n",
      "\t -- {\"accuracy_train\": 0.5281} \n",
      "\t -- val loss = 2.0409969328376376 / val cost = 2.1617854318204572\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 11 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.84it/s]\n",
      "epoch 11/30 \n",
      " \t -- train loss = 1.3872292518255431 / train cost = 1.5082480274619101\n",
      "\t -- {\"accuracy_train\": 0.50996} \n",
      "\t -- val loss = 2.1565192209627497 / val cost = 2.2775379965991167\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 12 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.07it/s]\n",
      "epoch 12/30 \n",
      " \t -- train loss = 1.4178750960441784 / train cost = 1.5559596456644649\n",
      "\t -- {\"accuracy_train\": 0.49416} \n",
      "\t -- val loss = 2.357837472359487 / val cost = 2.4959220219797733\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 13 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.01it/s]\n",
      "epoch 13/30 \n",
      " \t -- train loss = 1.4981212187441688 / train cost = 1.647128301406613\n",
      "\t -- {\"accuracy_train\": 0.46716} \n",
      "\t -- val loss = 2.188934263792056 / val cost = 2.3379413464545\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 14 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.05it/s]\n",
      "epoch 14/30 \n",
      " \t -- train loss = 1.4880660596608342 / train cost = 1.6438638893239066\n",
      "\t -- {\"accuracy_train\": 0.47572} \n",
      "\t -- val loss = 1.8759834950956553 / val cost = 2.031781324758728\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 15 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.10it/s]\n",
      "epoch 15/30 \n",
      " \t -- train loss = 1.52785354729825 / train cost = 1.687672845198337\n",
      "\t -- {\"accuracy_train\": 0.4561} \n",
      "\t -- val loss = 2.6713635938564564 / val cost = 2.8311828917565434\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 16 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.30it/s]\n",
      "epoch 16/30 \n",
      " \t -- train loss = 1.5129793399621023 / train cost = 1.6635781396199643\n",
      "\t -- {\"accuracy_train\": 0.45912} \n",
      "\t -- val loss = 2.2901310113833877 / val cost = 2.4407298110412494\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 17 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.88it/s]\n",
      "epoch 17/30 \n",
      " \t -- train loss = 1.450773980078197 / train cost = 1.5950421564144903\n",
      "\t -- {\"accuracy_train\": 0.48926} \n",
      "\t -- val loss = 1.915634852626395 / val cost = 2.0599030289626885\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 18 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.86it/s]\n",
      "epoch 18/30 \n",
      " \t -- train loss = 1.4024311880464237 / train cost = 1.537661396508202\n",
      "\t -- {\"accuracy_train\": 0.50176} \n",
      "\t -- val loss = 1.7615446505426602 / val cost = 1.8967748590044384\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 19 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.23it/s]\n",
      "epoch 19/30 \n",
      " \t -- train loss = 1.3596458825760043 / train cost = 1.4840130778060519\n",
      "\t -- {\"accuracy_train\": 0.51738} \n",
      "\t -- val loss = 1.6353688227566492 / val cost = 1.7597360179866968\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 20 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:12<00:00, 41.66it/s]\n",
      "epoch 20/30 \n",
      " \t -- train loss = 1.2889067989376846 / train cost = 1.404624157700589\n",
      "\t -- {\"accuracy_train\": 0.54906} \n",
      "\t -- val loss = 1.588134907972171 / val cost = 1.7038522667350753\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 21 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.87it/s]\n",
      "epoch 21/30 \n",
      " \t -- train loss = 1.3494240933107575 / train cost = 1.4633508911229978\n",
      "\t -- {\"accuracy_train\": 0.52704} \n",
      "\t -- val loss = 1.9907136218157588 / val cost = 2.104640419627999\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 22 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.30it/s]\n",
      "epoch 22/30 \n",
      " \t -- train loss = 1.4035062060700403 / train cost = 1.5325287428957932\n",
      "\t -- {\"accuracy_train\": 0.50194} \n",
      "\t -- val loss = 2.0211008358483036 / val cost = 2.1501233726740567\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 23 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.02it/s]\n",
      "epoch 23/30 \n",
      " \t -- train loss = 1.4524111948503775 / train cost = 1.5950102998289424\n",
      "\t -- {\"accuracy_train\": 0.48398} \n",
      "\t -- val loss = 1.7810450091375283 / val cost = 1.923644114116093\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 24 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.80it/s]\n",
      "epoch 24/30 \n",
      " \t -- train loss = 1.438997555965257 / train cost = 1.5879982334980274\n",
      "\t -- {\"accuracy_train\": 0.4877} \n",
      "\t -- val loss = 1.708976443193205 / val cost = 1.8579771207259754\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 25 ...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.12it/s]\n",
      "epoch 25/30 \n",
      " \t -- train loss = 1.5292335404678405 / train cost = 1.6827918864046234\n",
      "\t -- {\"accuracy_train\": 0.45796} \n",
      "\t -- val loss = 2.4428601588432866 / val cost = 2.596418504780069\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 26 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.20it/s]\n",
      "epoch 26/30 \n",
      " \t -- train loss = 1.44706081395088 / train cost = 1.5938857772285524\n",
      "\t -- {\"accuracy_train\": 0.48566} \n",
      "\t -- val loss = 2.1686460859906997 / val cost = 2.315471049268372\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 27 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.91it/s]\n",
      "epoch 27/30 \n",
      " \t -- train loss = 1.4121978575164231 / train cost = 1.5501312872170163\n",
      "\t -- {\"accuracy_train\": 0.50058} \n",
      "\t -- val loss = 1.23853205009653 / val cost = 1.376465479797123\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 28 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 42.15it/s]\n",
      "epoch 28/30 \n",
      " \t -- train loss = 1.3738910070892971 / train cost = 1.5038540384625902\n",
      "\t -- {\"accuracy_train\": 0.51424} \n",
      "\t -- val loss = 2.4920709859284758 / val cost = 2.6220340173017687\n",
      "\t -- {\"accuracy_val\": 0.0} \n",
      "\n",
      "\n",
      "starting epoch: 29 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.80it/s]\n",
      "epoch 29/30 \n",
      " \t -- train loss = 1.3210794977154507 / train cost = 1.4408857118516687\n",
      "\t -- {\"accuracy_train\": 0.53194} \n",
      "\t -- val loss = 1.552910574779808 / val cost = 1.672716788916026\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n",
      "starting epoch: 30 ...\n",
      "batch 500/500: 100%|██████████| 500/500 [00:11<00:00, 41.76it/s]\n",
      "epoch 30/30 \n",
      " \t -- train loss = 1.2661206210107299 / train cost = 1.3783438222801123\n",
      "\t -- {\"accuracy_train\": 0.55774} \n",
      "\t -- val loss = 1.7396032914746578 / val cost = 1.8518264927440402\n",
      "\t -- {\"accuracy_val\": 0.5} \n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test acc: {'accuracy': 0.5207}\n"
     ]
    }
   ],
   "source": [
    "# best hyperparams\n",
    "dims = best_trial[\"dims\"][1:-1]\n",
    "reg_rate = best_trial[\"reg_rate\"]\n",
    "\n",
    "in_dim = x_train.shape[1]\n",
    "out_dim = np.unique(y_train).size\n",
    "\n",
    "seed = 10\n",
    "model, loss, dims = build_model(dims, reg_rate, in_dim, out_dim, seed)\n",
    "model = compile_model(model, loss, step_size)\n",
    "\n",
    "verbose = 2\n",
    "history = model.fit(x_train_val, y_train_val, x_val[:2], y_val[:2], n_epochs, batch_size, verbose, aug_func)\n",
    "\n",
    "path_save_losses = \"assets/k_layer/losses.png\"\n",
    "path_save_costs = \"assets/k_layer/costs.png\"\n",
    "path_save_accuracies = \"assets/k_layer/accuracies.png\"\n",
    "path_save_lrs = \"assets/k_layer/lrs.png\"\n",
    "plot_losses(history, path_save_losses, show_val=False)\n",
    "plot_costs(history, path_save_costs, show_val=False)\n",
    "plot_accuracies(history, path_save_accuracies, show_val=False)\n",
    "plot_lrs(history, path_save_lrs)\n",
    "\n",
    "params_val = {\"mode\": \"test\"}\n",
    "scores_test = model.forward(x_test, **params_val)\n",
    "y_hat_test = np.argmax(scores_test, axis=1)\n",
    "metrics_test = model.compute_metrics(y_test, scores_test)\n",
    "print(f\"test acc: {metrics_test}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ordered-momentum",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "nn_blocks_env",
   "language": "python",
   "name": "nn_blocks_env"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}