Sahil Singla

sortpool2d. A new kind of pooling layer for faster and sharper convergence

salient_imagenet. Code for the ICLR 2022 paper. Salient Imagenet: How to discover spurious features in deep learning?

barlow. Code for the CVPR 2021 paper: Understanding Failures of Deep Networks via Robust Feature Extraction

style-transfer. style transfer code for generating stylized images using content and style loss from different layers, from different loss networks (vgg-16, vgg-19, inception-v1, inception-v2, inception-v3 trained on imagenet, inception-v3 trained on openimages, inception-v4

orthogonal-convolution. Github repo for my experiments with the orthogonal convolution idea

loss-exps. Experiments with different loss functions for image classification

SOC. Code for the ICML 2021 and ICLR 2022 papers: Skew Orthogonal Convolutions, Improved deterministic l2 robustness on CIFAR-10 and CIFAR-100

CASO. code for the paper https://arxiv.org/abs/1902.00407

CONV-SV. A fast and efficient way to compute a differentiable bound on the singular values of convolution layers

resnet-exps. Experiments with center loss function for image classification on resnet architecture

word-language-model. word-language-model imported and modified from pytorch-examples

improved_l2_robustness. Python

fantastic-four. Code for the ICLR paper: https://openreview.net/forum?id=JCRblSgs34Z

so-robust. Code for ICML paper "Second-Order Provable Defenses against Adversarial Attacks"

sparse-convolution. Code for sparse convolution written in C++

char-rnn. Python

data-augmentation. Code for fast data augmentation

singlasahil14.github.io. HTML