Robust-Classification-with-Convolutional-Prototype-Learning-Pytorch

Unofficial PyTorch Implementation of Robust Classification with Convolutional Prototype Learning

The original paper is available at: https://arxiv.org/abs/1805.03438

Architecture and DCE loss is defined in Models.py

The dataset is downloaded from the source (http://www.iro.umontreal.ca/~lisa/deep/data/mnist/mnist.pkl.gz), mentioned in the official Tensorflow repository (https://github.com/YangHM/Convolutional-Prototype-Learning).

1. Run main.py to train the model.
2. Run extract_features.py to extract the hidden layer features
3. Run visualize_scatter.py to generate the scatter plot

Features with DCE loss:

Features with GCPL: