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.

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

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

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

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。

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

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

Visualize loss landscape

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

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

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.

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

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

Create animations for the optimization trajectory of neural nets

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