misc: Update SDK version. Update project description. Clean up some c…

…ode.

CHANGELOG.md

@@ -1,3 +1,9 @@
+# 0.4.4
+
+- Update SDK version from `2.12.0` to `2.17.0`.
+- Updated project description to make it more accurate in describing what
+  `synadart` actually is.
+
 # 0.4.3
 
 - Bumped version of `sprint` from `1.0.2+3` to `1.0.3`.

example/example.dart

@@ -2,33 +2,36 @@ import 'package:synadart/src/layers/core/dense.dart';
 import 'package:synadart/synadart.dart';
 
 void main() {
-  final network = Sequential(learningRate: 0.2, layers: [
-    Dense(
-      size: 15,
-      activation: ActivationAlgorithm.sigmoid,
-    ),
-    Dense(
-      size: 5,
-      activation: ActivationAlgorithm.sigmoid,
-    ),
-    Dense(
-      size: 1,
-      activation: ActivationAlgorithm.sigmoid,
-    )
-  ]);
+  final network = Sequential(
+    learningRate: 0.2,
+    layers: [
+      Dense(
+        size: 15,
+        activation: ActivationAlgorithm.sigmoid,
+      ),
+      Dense(
+        size: 5,
+        activation: ActivationAlgorithm.sigmoid,
+      ),
+      Dense(
+        size: 1,
+        activation: ActivationAlgorithm.sigmoid,
+      )
+    ],
+  );
 
   // We are expecting to get the number '5'.
   final expected = [
-    [0.01],
-    [0.01],
-    [0.01],
-    [0.01],
-    [0.01],
-    [0.99],
-    [0.01],
-    [0.01],
-    [0.01],
-    [0.01],
+    [0.01], // 0
+    [0.01], // 1
+    [0.01], // 2
+    [0.01], // 3
+    [0.01], // 4
+    [0.99], // 5
+    [0.01], // 6
+    [0.01], // 7
+    [0.01], // 8
+    [0.01], // 9
   ];
 
   // Training data contains different number patterns.
@@ -38,8 +41,7 @@ void main() {
     '111001111100111'.split('').map(double.parse).toList(),
     '111001111001111'.split('').map(double.parse).toList(),
     '101101111001001'.split('').map(double.parse).toList(),
-    // This is the number 5
-    '111100111001111'.split('').map(double.parse).toList(),
+    '111100111001111'.split('').map(double.parse).toList(), // 5
     '111100111101111'.split('').map(double.parse).toList(),
     '111001001001001'.split('').map(double.parse).toList(),
     '111101111101111'.split('').map(double.parse).toList(),
@@ -60,7 +62,7 @@ void main() {
   final numberFive = trainingData[5];
 
   // Train the network using the training and expected data.
-  network.train(inputs: trainingData, expected: expected, iterations: 5000);
+  network.train(inputs: trainingData, expected: expected, iterations: 20000);
 
   print('Confidence in recognising a 5: ${network.process(numberFive)}');
   for (final test in testData) {

lib/src/activation.dart

@@ -28,14 +28,14 @@ const algorithms = <ActivationAlgorithm, List<ActivationFunctionSignature>>{
 ActivationFunction resolveActivationAlgorithm(
   ActivationAlgorithm activationAlgorithm,
 ) =>
-    (weightedSum) => algorithms[activationAlgorithm]![0](weightedSum());
+    (weightedSum) => algorithms[activationAlgorithm]!.first(weightedSum());
 
 /// Resolves an `ActivationAlgorithm` to the derivative of the mathematical
 /// function in the form of an `ActivationFunction`
 ActivationFunction resolveActivationDerivative(
   ActivationAlgorithm activationAlgorithm,
 ) =>
-    (weightedSum) => algorithms[activationAlgorithm]![1](weightedSum());
+    (weightedSum) => algorithms[activationAlgorithm]!.last(weightedSum());
 
 /// Shrinks the range of values to inbetween 0 and 1 using exponentials. Results
 /// can be driven into saturation, which makes the sigmoid function unsuited for

lib/src/layers/core/dense.dart

@@ -6,7 +6,7 @@ import 'package:synadart/src/layers/layer.dart';
 class Dense extends Layer {
   /// Construct a dense layer using the [activation] algorithm and [size].
   Dense({
-    required int size,
-    required ActivationAlgorithm activation,
-  }) : super(size: size, activation: activation);
+    required super.size,
+    required super.activation,
+  });
 }

lib/src/layers/layer.dart

@@ -60,14 +60,16 @@ class Layer {
   }) {
     isInput = parentLayerSize == 0;
 
-    neurons.addAll(Iterable.generate(
-      size,
-      (_) => Neuron(
-        activationAlgorithm: activation,
-        parentLayerSize: parentLayerSize,
-        learningRate: learningRate,
+    neurons.addAll(
+      Iterable.generate(
+        size,
+        (_) => Neuron(
+          activationAlgorithm: activation,
+          parentLayerSize: parentLayerSize,
+          learningRate: learningRate,
+        ),
       ),
-    ));
+    );
   }
 
   /// Accept a single input or multiple [inputs] by assigning them sequentially

lib/src/layers/recurrent/lstm.dart

@@ -22,16 +22,10 @@ class LSTM extends Layer {
   /// [recurrenceActivation] - Algorithm used to activate recurrence
   /// connections.
   LSTM({
-    required int size,
-    required ActivationAlgorithm activation,
+    required super.size,
+    required super.activation,
     required ActivationAlgorithm recurrenceActivation,
-  }) : super(
-          size: size,
-          activation: activation,
-        ) {
-    this.recurrenceActivation =
-        resolveActivationAlgorithm(recurrenceActivation);
-  }
+  }) : recurrenceActivation = resolveActivationAlgorithm(recurrenceActivation);
 
   /// Obtain the output by applying the recurrent memory algorithm.
   @override

lib/src/networks/network.dart

@@ -51,8 +51,9 @@ class Network {
   /// Adds a `Layer` to this `Network`.
   void addLayer(Layer layer) {
     layer.initialise(
-        parentLayerSize: layers.isEmpty ? 0 : layers[layers.length - 1].size,
-        learningRate: learningRate);
+      parentLayerSize: layers.isEmpty ? 0 : layers.last.size,
+      learningRate: learningRate,
+    );
 
     layers.add(layer);
 

lib/src/networks/sequential.dart

@@ -1,4 +1,3 @@
-import 'package:synadart/src/layers/layer.dart';
 import 'package:synadart/src/networks/network.dart';
 import 'package:synadart/src/networks/training/backpropagation.dart';
 
@@ -7,7 +6,7 @@ import 'package:synadart/src/networks/training/backpropagation.dart';
 class Sequential extends Network with Backpropagation {
   /// Creates a `Sequential` model network.
   Sequential({
-    required double learningRate,
-    List<Layer>? layers,
-  }) : super(learningRate: learningRate, layers: layers);
+    required super.learningRate,
+    super.layers,
+  });
 }

lib/src/networks/training/backpropagation.dart

@@ -42,7 +42,8 @@ mixin Backpropagation on Network {
 
     if (inputs.length != expected.length) {
       log.severe(
-          'Inputs and expected result lists must be of the same length.');
+        'Inputs and expected result lists must be of the same length.',
+      );
       return;
     }
 
@@ -66,14 +67,19 @@ mixin Backpropagation on Network {
 
     for (var iteration = 0; iteration < iterations; iteration++) {
       stopwatch.start();
+
       for (var index = 0; index < inputs.length; index++) {
         propagateBackwards(inputs[index], expected[index]);
       }
+
       stopwatch.stop();
+
       if (iteration % 500 == 0) {
         log.info(
-            'Iterations: $iteration/$iterations ~ ETA: ${secondsToETA((stopwatch.elapsedMicroseconds * (iterations - iteration)) ~/ 1000000)}');
+          'Iterations: $iteration/$iterations ~ ETA: ${secondsToETA((stopwatch.elapsedMicroseconds * (iterations - iteration)) ~/ 1000000)}',
+        );
       }
+
       stopwatch.reset();
     }
   }

lib/src/neurons/neuron.dart

@@ -116,7 +116,7 @@ class Neuron {
     }
 
     if (this.inputs.isNotEmpty) {
-      this.inputs[0] = input!;
+      this.inputs.first = input!;
     } else {
       this.inputs.add(input!);
     }
@@ -142,5 +142,5 @@ class Neuron {
   /// it will output the weighted sum of the [inputs] and [weights], passed
   /// through the activation function.
   double get output =>
-      weights.isEmpty ? inputs[0] : activation(() => dot(inputs, weights));
+      weights.isEmpty ? inputs.first : activation(() => dot(inputs, weights));
 }

lib/src/utils/mathematical_operations.dart

@@ -1,27 +1,33 @@
 /// Calculates the dot product of two lists
 double dot(List<double> a, List<double> b) {
   var result = 0.0;
+
   for (var index = 0; index < a.length; index++) {
     result += a[index] * b[index];
   }
+
   return result;
 }
 
 /// Adds values in list [b] to list [a]
 List<double> add(List<double> a, List<double> b) {
   final result = <double>[];
+
   for (var index = 0; index < a.length; index++) {
     result.add(a[index] + b[index]);
   }
+
   return result;
 }
 
 /// Subtracts values in list [b] from list [a]
 List<double> subtract(List<double> a, List<double> b) {
   final result = <double>[];
+
   for (var index = 0; index < a.length; index++) {
     result.add(a[index] - b[index]);
   }
+
   return result;
 }
 

lib/src/utils/value_generator.dart

@@ -16,6 +16,9 @@ Iterable<double> doubleIterableSync({double from = 0, double to = 0}) sync* {
 }
 
 /// Generates a list of size [size], filled with random `double` values.
-List<double> generateListWithRandomDoubles(
-        {required int size, double from = 0, double to = 0}) =>
+List<double> generateListWithRandomDoubles({
+  required int size,
+  double from = 0,
+  double to = 0,
+}) =>
     doubleIterableSync(from: from, to: to).take(size).toList();

lib/synadart.dart

@@ -1,3 +1,5 @@
+/// A limited but fully documented neural network library created for
+/// educational purposes.
 library synadart;
 
 export 'src/activation.dart';

pubspec.yaml

@@ -1,20 +1,20 @@
 name: synadart
-version: 0.4.3
+version: 0.4.4
 
 description: >-
-  A simple-to-grasp, complete and fully documented Neural Network library, 
-  written from scratch in Dart.
+  A limited but fully documented neural network library created for educational
+  purposes.
 
 homepage: https://github.com/wordcollector/synadart
 repository: https://github.com/wordcollector/synadart
 issue_tracker: https://github.com/wordcollector/synadart/issues
 
 environment:
-  sdk: '>=2.12.0 <3.0.0'
+  sdk: '>=2.17.0 <3.0.0'
 
 dependencies:
   # Logging
-  sprint: ^1.0.3
+  sprint: ^1.0.4
 
 dev_dependencies:
-  words: ^0.0.2+1
+  words: ^0.1.1

test/synadart_test.dart

@@ -1,4 +1 @@
-import 'package:synadart/synadart.dart';
-
-void main() {
-}
+void main() {}