deep-learning-theory 开源项目推荐（29 个）

MinghuiChen43 /awesome-deep-phenomena

一份精选的深度学习有趣实证研究和见解论文列表。持续更新......

james-simon /eigenlearning

codebase for "A Theory of the Inductive Bias and Generalization of Kernel Regression and Wide Neural Networks"

SeanJia /InfoMCR

Code for "Information-Theoretic Local Minima Characterization and Regularization"

Dimundel /Jacobian-Analysis-of-a-Recurrent-Transformer-Block

Jacobian stability bounds for a looped Transformer block with MSA, FFN and RMSNorm (AINL 2026)

nabeelshan78 /First-Principles-Deep-Learning

From-scratch NumPy implementations of core deep learning architectures (DNN, CNN, RNN). A journey into the first principles of AI.

gszfwsb /Unveiling-Induction-Heads

PyTorch implementation for "Unveiling Induction Heads: Provable Training Dynamics and Feature Learning in Transformers", NeurIPS 2024

pragyaangaur /Viscous-Conservation-Laws

Momentum breaks the conservation laws of neural network training, and friction decides by how much. Exact drift laws for heavy ball and Nesterov, proved in Lean 4.

umutonuryasar /CS229-Machine-Learning

Technical portfolio for Stanford CS229 (Summer 2020). First-principles approach to ML: Math-heavy derivations, NumPy-from-scratch implementations, and research project.

avysogorets /goldilocks-zone

Code to reproduce the paper "Deconstructing the Goldilocks Zone of Neural Network Initialization"

AEjonanonymous /The-Ghost-In-Yalls-Machine

Resolving the Black Box Interpretability Problem by Proving the Mathematical Necessity of J-space via Hessian Rank Deficiency in Lean 4.

deepthink-umich /deep-lora-transformers

[ICML'24] Compressible Dynamics in Deep Overparameterized Low-Rank Learning & Adaptation

deepthink-umich /neural-collapse-MSE

[ICML'22] On the Optimization Landscape of Neural Collapse under MSE Loss: Global Optimality with Unconstrained Features

deepthink-umich /normalized-neural-collapse

[NeurIPS'22] Neural Collapse with Normalized Features: A Geometric Analysis over the Riemannian Manifold

deepthink-umich /law-of-parsimony

[arXiv 2023] The Law of Parsimony in Gradient Descent for Learning Deep Linear Networks

drwjkirkpatrick-web /penrose-aperiodicity-barrier

Proof 4: What neural networks cannot see — the Penrose aperiodicity barrier. 3/3 theorems empirically verified.

joelneeser /master-thesis

Approximation-Theoretic Properties of Finite-Precision Neural Networks – MSc thesis, ETH Zurich (2024)

deepthink-umich /Neural-Collapse

[NeurIPS'21] A Geometric Analysis of Neural Collapse with Unconstrained Features

AleczO /cyclic-group-representations

Theoretical and empirical analysis of cyclic group representations in neural networks for modular arithmetic.

DissipativeAI /LIFE

LIFE 框架：基于动态同构的自指耗散智能模型 LIFE Framework: Life Is Full Effective — A Dynamic Isomorphic Dissipative Intelligence

deepthink-umich /PNC_DLN

[JMLR 2025] Understanding Deep Representation Learning via Layerwise Feature Compression and Discrimination