Pytorch BSconv layer class codes to substitute your ordinary conv layer. BSnet neurons can be substituted into convolution, fully connected, attention and transformer layers neurons.
Boolean Structured Convolutional Deep Learning Network (BSconvnet)
- Our model has only 17000+ parameters instead of 3.4 millions parameters in SmoothNet
- Use separable convolutional deep learning network, so it does not overfit
- Achieved state of the art accuracy on CIFAR10 dataset
- Able to be trained on online using Google Colab with GPU
- No data augmentations, no regularization such as weight decay and dropout
Download the jupyter notebook
Open using Google Colab Colab It can also be run using a jupyter notebook
Follow the steps in the notebook, run each block of codes starting from the top to the bottom
Leaderboard of model accuracies on CIFAR10 dataset Leaderboard
Discrete Markov Random Field Relaxation
That's it. Have a Nice Day!!!