(ICLR 2022 Spotlight) Official PyTorch implementation of "How Do Vision Transformers Work?"

Create animations for the optimization trajectory of neural nets

Explores the ideas presented in Deep Ensembles: A Loss Landscape Perspective (https://arxiv.org/abs/1912.02757) by Stanislav Fort, Huiyi Hu, and Balaji Lakshminarayanan.

Implements sharpness-aware minimization (https://arxiv.org/abs/2010.01412) in TensorFlow 2.

This repository contains the code and models for our paper "Investigating and Mitigating Failure Modes in Physics-informed Neural Networks(PINNs)"

Landscaper is a comprehensive Python framework designed for exploring the loss landscapes of deep learning models.

[TMLR] "Can You Win Everything with Lottery Ticket?" by Tianlong Chen, Zhenyu Zhang, Jun Wu, Randy Huang, Sijia Liu, Shiyu Chang, Zhangyang Wang

Source code for NeurIPS-2024 paper "Where Do Large Learning Rates Lead Us"

analytic solution to the git-merge algorithm, derived from "Git Re-Basin: Merging Models modulo Permutation Symmetries"

[NeurIPS 2021] Towards Better Understanding of Training Certifiably Robust Models against Adversarial Examples | ⛰️⚠️

[Int. J. Comput. Vis. 2024] Revisiting Deep Ensemble for Out-of-Distribution Detection: A Loss Landscape Perspective

Surrogate Gap Guided Sharpness-Aware Minimization (GSAM) implementation for keras/tensorflow 2

Worth-reading papers and related awesome resources on deep learning optimization algorithms. 值得一读的深度学习优化器论文与相关资源。

Visualize loss landscape

a web-based interactive 3D visualization tool for model optimization in loss-space, non-linearity effect on node boundaries, and an abstract overview of pytorch's framework for the optimization framework

The project explores CNN architectures (ResNet vs VGG, Deformable CNNs) and 3D scene reconstruction with Tiny NeRF, including ray sampling and volume rendering. Project Year: 2025

This project builds on recent research that explores the phenomenon of Grokking. The goal is to investigate when, why, and how grokking occurs, focusing on transformers under various batch sizes.

code and difference of resolution for visualizing the loss landscape of a GAN and understanding what a loss landscape is

Code for NeurIPS 2024 paper "Only Strict Saddles in the Energy Landscape of Predictive Coding Networks?"