Paper To Read

• [2410.19637] A distributional simplicity bias in the learning dynamics of transformers

• Belrose: Neural networks learn statistics of increasing… - Google Scholar

  • 当該論文を引用している論文調べ
• [2603.12901] A theory of learning data statistics in diffusion models, from easy to hard
• [2510.04285] Probing Geometry of Next Token Prediction Using Cumulant Expansion of the Softmax Entropy
• [2602.12257] On the implicit regularization of Langevin dynamics with projected noise
• Neural networks trained with SGD learn distributions of increasing complexity | OpenReview
• SGD on Neural Networks Learns Functions of Increasing Complexity
• [1805.08522] Deep learning generalizes because the parameter-function map is biased towards simple functions

• You're Measuring Model Complexity Wrong — LessWrong
• [2410.02984] Differentiation and Specialization of Attention Heads via the Refined Local Learning Coefficient
• [2511.04564] Uncertainties in Physics-informed Inverse Problems: The Hidden Risk in Scientific AI
• [2406.10234] Review and Prospect of Algebraic Research in Equivalent Framework between Statistical Mechanics and Machine Learning Theory

• Computational Complexity of Learning Neural Networks: Smoothness and Degeneracy

  • https://neurips.cc/virtual/2023/poster/72775
• Singular Learning Theory for Dummies — LessWrong
• Timaeus | Compressibility Measures Complexity: Minimum Description Length Meets Singular Learning Theory
• https://neurips.cc/virtual/2024/poster/95029

https://arxiv.org/search/cs?searchtype=author&query=Murfet,+D

• Search | arXiv e-print repository

• [2603.05228] The Geometric Inductive Bias of Grokking: Bypassing Phase Transitions via Architectural Topology

• [2301.00808] ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders
• [2602.15029] Symmetry in language statistics shapes the geometry of model representations

• ベイズ推定に基づくスペクトルデータからの情報抽出

  • https://mns.k.u-tokyo.ac.jp/pdf/2023jpsj_nabika.pdf
• https://www.ai-gakkai.or.jp/jsai2017/webprogram/2017/pdf/733.pdf
• [2303.14151] Double Descent Demystified: Identifying, Interpreting & Ablating the Sources of a Deep Learning Puzzle
• [1801.05894] Deep Learning: An Introduction for Applied Mathematicians
• [2603.05228] The Geometric Inductive Bias of Grokking: Bypassing Phase Transitions via Architectural Topology

https://github.com/Hiroki11x/Papers これを参考にしたい．

読まなきゃと思ってる Paper List | Hiroki Naganuma