negative-learning

This repository contains the accompanying code for paper "Classification Based on missing features in Deep Convolutional Neural Networks" submitted to NNW journal in December 2018.

Dependencies

• pytorch 1.2.0 (CUDA supported)
• torchvision 0.4.0 (for MNIST and EMNIST datasets)
• tqdm

Running

python mnist.py [--emnist] [--emnist-mnist]

Example outputs

Example outputs are in .txt files in this repository.

Contact

nmilosev [at] dmi.rs