Hiroki Naganuma

既存コンテンツの引用棚卸しと対応表

総合索引

抽出範囲と読み方

既存のcontent/**/*.mdを対象に、学術ドメインと研究関連PDFへのリンクを機械抽出した。54ページから412候補を得て、URL系列で224件を今回の文献台帳に照合した。残る188件は未再検証候補としてそのまま提示する。この件数は論文数ではなく、会議ページ、著者ページ、書籍等も含む候補数である。

中心対象の37研究トピック内に限ると248候補中219件を照合した。残る29候補には講義資料、プロフィール、匿名旧投稿、別URL対応の未確定資料が含まれる。会議参加記・周辺ページにのみある引用は、すべてを同じ深さで再調査したわけではない。

ホスト名とPDF拡張子による抽出であり、すべての文献の発見を保証しない。コメント・コード内、文献名だけの記載、相対リンク、通常のブログ等は抽出外。arXivの版番号とPDF/HTML差、OpenReview ID等は統合したが、arXivと会議版の異なるURLの対応が残る可能性がある。未再検証は、無関係・誤り・未発表という判定ではない。 台帳への対応が未確定という意味で、要旨の一部を回収できたが著者や版が不明な匿名投稿も含む。下の旧ページ由来ラベルは題名として未検証である。

棚卸しJSON · 確認済み登録の参考文献

37トピックの対応

元のページ	再サーベイ
Attention	基盤モデル
GAN: Generative Adversarial Networks	基盤モデル
LLM / Implementation に関して	基盤モデル
MultiModal	基盤モデル
Reasoning	基盤モデル
RLHF: Reinforcement Learning from Human Feedback	基盤モデル
Scaling Laws	基盤モデル
Transformer	基盤モデル
Calibration	汎化・分布シフト・校正・公平性
Compositional Generalization	汎化・分布シフト・校正・公平性
Distribution Shift入門: Covariate・Label・Concept Shift	汎化・分布シフト・校正・公平性
ECE: Expected Calibration Error	汎化・分布シフト・校正・公平性
機械学習のFairness: Group Robustnessとの違いと評価指標	汎化・分布シフト・校正・公平性
IC: Information Criteria	汎化・分布シフト・校正・公平性
OOD入門: Distribution Shift・Domain Generalization・OOD Detection	汎化・分布シフト・校正・公平性
Double Descent	数理手法・学習ダイナミクス
FIM: Fisher Information Matrix	数理手法・学習ダイナミクス
NTK: Neural Tangent Kernel	数理手法・学習ダイナミクス
AD: Automatic Differentiation	数理手法・学習ダイナミクス
Flow Matching入門: 連続の式・条件付き目的・ODE生成	数理手法・学習ダイナミクス
Hessian	数理手法・学習ダイナミクス
MCMC	数理手法・学習ダイナミクス
Misc	数理手法・学習ダイナミクス
Modular Manifolds入門: Stiefel多様体とニューラルネット最適化	数理手法・学習ダイナミクス
Bi-Level Optimization	最適化
CGD: Conjugate Gradient	最適化
Damping	最適化
Large-Batch Training Survey: スケーリング則・限界・最適化手法	最適化
Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化	最適化
Learning Rate Schedule Survey: Shape・Scale・適応化	最適化
Muonの数理: Frobenius・スペクトル・核ノルムと最急降下	最適化
Muon Optimizer: Newton–Schulz直交化の定義・理論・実装上の注意	最適化
NGD: Natural Gradient	最適化
Optimizer	最適化
SAM: Sharpness Aware Minimization	最適化
Shampoo: Shampoo Optimizer	最適化
Parallel Training	並列学習・計算基盤

既存引用候補の照合結果

登録済みと未再検証を分け、各候補の出現元を示す。旧メモの行番号は今回の導線追加後に移動しうるため、JSON内の行番号は抽出時点の参考値。

今回のサーベイに登録あり

• Benchmarking Neural Network Training Algorithms — OP005。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録

• On Empirical Comparisons of Optimizers for Deep Learning — OP004。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録、Optimizer

• Which Algorithmic Choices Matter at Which Batch Sizes? Insights From a Noisy Quadratic Model — OP023。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録、Optimizer

• LoRA: Low-Rank Adaptation of Large Language Models — FM088。出現元：Microsoft Research Intern 2024 備忘録

• Editing Models with Task Arithmetic — SY010。出現元：Microsoft Research Intern 2024 備忘録

• VisDA: The Visual Domain Adaptation Challenge — GE087。出現元：Datasets、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• SGDR: Stochastic Gradient Descent with Warm Restarts — OP039。出現元：AlgoPerf Workshop 2025、Learning Rate Schedule Survey: Shape・Scale・適応化

• Prodigy: An Expeditiously Adaptive Parameter-Free Learner — OP089。出現元：AlgoPerf Workshop 2025

• On Calibration and Out-of-domain Generalization — GE40。出現元：読まなきゃと思ってる Paper List、Calibration、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• A Fine-Grained Analysis on Distribution Shift — GE122。出現元：読まなきゃと思ってる Paper List、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• The Evolution of Out-of-Distribution Robustness Throughout Fine-Tuning — GE088。出現元：読まなきゃと思ってる Paper List、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Towards a Theoretical Framework of Out-of-Distribution Generalization — GE089。出現元：読まなきゃと思ってる Paper List、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Can You Trust Your Model’s Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift — GE39。出現元：読まなきゃと思ってる Paper List、Calibration

• Scalable Marginal Likelihood Estimation for Model Selection in Deep Learning — GE090。出現元：読まなきゃと思ってる Paper List、Calibration、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• An Investigation into Neural Net Optimization via Hessian Eigenvalue Density — MD06。出現元：読まなきゃと思ってる Paper List

• StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks — FM101。出現元：GAN: Generative Adversarial Networks

• RGBD-GAN: Unsupervised 3D Representation Learning From Natural Image Datasets via RGBD Image Synthesis — FM102。出現元：GAN: Generative Adversarial Networks

• Training language GANs from Scratch — FM103。出現元：GAN: Generative Adversarial Networks

• Social-BiGAT: Multimodal Trajectory Forecasting using Bicycle-GAN and Graph Attention Networks — FM104。出現元：GAN: Generative Adversarial Networks

• 深層生成モデルによる統計的推論 — FM110。出現元：GAN: Generative Adversarial Networks

• Spectral Normalization for Generative Adversarial Networks — FM073, MD79。出現元：GAN: Generative Adversarial Networks

• The Numerics of GANs — FM074。出現元：GAN: Generative Adversarial Networks

• Large Scale GAN Training for High Fidelity Natural Image Synthesis — FM077。出現元：GAN: Generative Adversarial Networks

• NIPS 2016 Tutorial: Generative Adversarial Networks — FM107。出現元：GAN: Generative Adversarial Networks

• DeepSpeed: 深層学習の訓練と推論を劇的に高速化するフレームワーク — FM111。出現元：LLM / Implementation に関して

• Multimodal Machine Learning: A Survey and Taxonomy — FM028。出現元：MultiModal

• Multimodal Deep Learning — FM095。出現元：MultiModal

• Recent Advances and Trends in Multimodal Deep Learning: A Review — FM097。出現元：MultiModal

• Information content and analysis methods for Multi-Modal High-Throughput Biomedical Data — FM108。出現元：MultiModal

• Multimodal deep learning for biomedical data fusion: a review — FM096。出現元：MultiModal

• Multimodal deep learning models for early detection of Alzheimer’s disease stage — FM109。出現元：MultiModal

• On the Benefits of Early Fusion in Multimodal Representation Learning — FM036。出現元：MultiModal

• Revisit Multimodal Meta-Learning through the Lens of Multi-Task Learning — FM039。出現元：MultiModal

• LLM Reasoning: Key Ideas and Limitations — FM112。出現元：Reasoning

• Scaling Laws for Neural Language Models — FM018。出現元：Scaling Laws

• An Empirical Model of Large-Batch Training — OP022。出現元：Scaling Laws、Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Scaling Vision Transformers — FM020。出現元：Scaling Laws

• Explaining Neural Scaling Laws — FM021。出現元：Scaling Laws

• Attention Is All You Need — FM002。出現元：Transformer

• On Calibration of Modern Neural Networks — GE28。出現元：Calibration

• Unified Uncertainty Calibration — GE83。出現元：Calibration

• On Fairness and Calibration — GE50。出現元：Calibration

• Sample Complexity of Uniform Convergence for Multicalibration — GE091。出現元：Calibration

• Trainable Calibration Measures for Neural Networks from Kernel Mean Embeddings — GE31。出現元：Calibration

• Preventing Failures Due to Dataset Shift: Learning Predictive Models That Transport — GE092。出現元：Calibration

• Calibrating Deep Neural Networks using Focal Loss — GE35。出現元：Calibration

• Intra Order-preserving Functions for Calibration of Multi-Class Neural Networks — GE093。出現元：Calibration

• Beyond temperature scaling: Obtaining well-calibrated multiclass probabilities with Dirichlet calibration — GE34。出現元：Calibration

• Beyond temperature scaling: Obtaining well-calibrated multiclass probabilities with Dirichlet calibration — GE34。出現元：Calibration

• Measuring Calibration in Deep Learning — GE30。出現元：Calibration

• Detecting and Correcting for Label Shift with Black Box Predictors — GE02。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• Dataset Shift in Machine Learning — GE138。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• Causality matters in medical imaging — GE131。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• Causality matters in medical imaging — GE131。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• Understanding Dataset Shift and Potential Remedies — GE139。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• Equality of Opportunity in Supervised Learning — GE48。出現元：機械学習のFairness: Group Robustnessとの違いと評価指標

• Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization — GE07。出現元：機械学習のFairness: Group Robustnessとの違いと評価指標、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Towards a Standard for Identifying and Managing Bias in Artificial Intelligence — GE140。出現元：機械学習のFairness: Group Robustnessとの違いと評価指標

• Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification — GE53。出現元：機械学習のFairness: Group Robustnessとの違いと評価指標

• Tilted Empirical Risk Minimization — GE123。出現元：IC: Information Criteria

• An Online Method for A Class of Distributionally Robust Optimization with Non-convex Objectives — GE141。出現元：IC: Information Criteria

• Generalized Out-of-Distribution Detection: A Survey — GE27。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• In Search of Lost Domain Generalization — GE08。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• WILDS: A Benchmark of in-the-Wild Distribution Shifts — GE09。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• OpenOOD: Benchmarking Generalized Out-of-Distribution Detection — GE25。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Out of Distribution Generalization in Machine Learning — GE094。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Shortcut Learning in Deep Neural Networks — GE86。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Invariant Risk Minimization — GE05。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• An Information-theoretic Approach to Distribution Shifts — GE095。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Can Subnetwork Structure be the Key to Out-of-Distribution Generalization? — GE096。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Test Sample Accuracy Scales with Training Sample Density in Neural Networks — GE097。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Symmetric Cross Entropy for Robust Learning with Noisy Labels — GE098。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Adaptive Risk Minimization: Learning to Adapt to Domain Shift — GE099。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Domain Generalization Guided by Gradient Signal to Noise Ratio of Parameters — GE132。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Identifiability Conditions for Domain Adaptation — GE84。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Domain-invariant Feature Exploration for Domain Generalization — GE124。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• A Note on “Assessing Generalization of SGD via Disagreement” — GE125。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Assessing Generalization of SGD via Disagreement — GE82。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Generalization in Supervised Learning Through Riemannian Contraction — GE130。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Predicting Out-of-Domain Generalization with Neighborhood Invariance — GE127。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• A Closer Look at Accuracy vs. Robustness — GE144。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Noise or Signal: The Role of Image Backgrounds in Object Recognition — GE100。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• On the Impact of Spurious Correlation for Out-of-distribution Detection — GE101。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• The Auto Arborist Dataset: A Large-Scale Benchmark for Multiview Urban Forest Monitoring Under Domain Shift — GE146。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• OoD-Bench: Quantifying and Understanding Two Dimensions of Out-of-Distribution Generalization — GE85。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Adaptive versus Standard Descent Methods and Robustness Against Adversarial Examples — GE102。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Towards Shape Biased Unsupervised Representation Learning for Domain Generalization — GE103。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Understanding and Testing Generalization of Deep Networks on Out-of-Distribution Data — GE104。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Understanding Robustness of Transformers for Image Classification — GE105。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Pretrained Transformers Improve Out-of-Distribution Robustness — GE106。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• An Empirical Investigation of Domain Generalization with Empirical Risk Minimizers — GE143。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Uncertainty Baselines: Benchmarks for Uncertainty & Robustness in Deep Learning — GE107。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• A Winning Hand: Compressing Deep Networks Can Improve Out-Of-Distribution Robustness — GE108。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• The Pitfalls of Simplicity Bias in Neural Networks — GE109。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Domain Generalization for Object Recognition with Multi-task Autoencoders — GE110。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Unbiased Metric Learning: On the Utilization of Multiple Datasets and Web Images for Softening Bias — GE145。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Deeper, Broader and Artier Domain Generalization — GE111。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Deep Hashing Network for Unsupervised Domain Adaptation — GE133。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Recognition in Terra Incognita — GE112。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Moment Matching for Multi-Source Domain Adaptation — GE113。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Adaptiope: A Modern Benchmark for Unsupervised Domain Adaptation — GE147。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness — GE114。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Concept Bottleneck Models — GE115。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Do Image Classifiers Generalize Across Time? — GE116。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples — GE117。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• A Large-scale Study of Representation Learning with the Visual Task Adaptation Benchmark — GE118。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Don’t Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering — GE119。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Adversarial NLI: A New Benchmark for Natural Language Understanding — GE120。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Benchmarking Adversarial Robustness — GE134。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Natural Adversarial Examples — GE121。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• On the interplay between noise and curvature and its effect on optimization and generalization — GE70, MD08。出現元：FIM: Fisher Information Matrix

• Neural Tangent Kernel: Convergence and Generalization in Neural Networks — MD10。出現元：NTK: Neural Tangent Kernel

• A Recipe for Global Convergence Guarantee in Deep Neural Networks — MD21。出現元：NTK: Neural Tangent Kernel

• Any Target Function Exists in a Neighborhood of Any Sufficiently Wide Random Network: A Geometrical Perspective — MD22。出現元：NTK: Neural Tangent Kernel

• Automatic Differentiation in Machine Learning: a Survey — MD01。出現元：AD: Automatic Differentiation

• Flow Matching for Generative Modeling — MD28。出現元：Flow Matching入門: 連続の式・条件付き目的・ODE生成

• Improving and generalizing flow-based generative models with minibatch optimal transport — MD30。出現元：Flow Matching入門: 連続の式・条件付き目的・ODE生成

• Stochastic Interpolants: A Unifying Framework for Flows and Diffusions — MD31。出現元：Flow Matching入門: 連続の式・条件付き目的・ODE生成

• Flow Matching Guide and Code — MD35。出現元：Flow Matching入門: 連続の式・条件付き目的・ODE生成

• Causal de Finetti: On the Identification of Invariant Causal Structure in Exchangeable Data — MD68。出現元：Misc

• 独立成分解析の信号処理への応用 — MD92。出現元：Misc

• BOME! Bilevel Optimization Made Easy: A Simple First-Order Approach — OP071。出現元：Bi-Level Optimization

• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium — OP072, FM076。出現元：Bi-Level Optimization

• Stability and Generalization of Bilevel Programming in Hyperparameter Optimization — OB021。出現元：Bi-Level Optimization

• DARTS: Differentiable Architecture Search — OB022。出現元：Bi-Level Optimization

• Model-based Reinforcement Learning: A Survey — OB023。出現元：Bi-Level Optimization

• High-Resolution Image Synthesis with Latent Diffusion Models — MD93。出現元：Bi-Level Optimization

• The Missing Invariance Principle Found – the Reciprocal Twin of Invariant Risk Minimization — OB024。出現元：Bi-Level Optimization

• What Is Missing in IRM Training and Evaluation? Challenges and Solutions — OB025。出現元：Bi-Level Optimization

• A Gradient Method for Multilevel Optimization — OB026。出現元：Bi-Level Optimization

• Investigating Bi-Level Optimization for Learning and Vision from a Unified Perspective: A Survey and Beyond — OP073。出現元：Bi-Level Optimization

• 共役勾配法 — OB028。出現元：CGD: Conjugate Gradient

• 非線形計画法（3）—無制約最適化問題— — OB029。出現元：Damping

• Optimizing Neural Networks with Kronecker-factored Approximate Curvature — OP015。出現元：Damping、NGD: Natural Gradient

• ニューラルネットワークにおける適応的二次最適化手法 — OB027。出現元：Damping

• Measuring the Effects of Data Parallelism on Neural Network Training — OB002。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour — OP021。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法、Learning Rate Schedule Survey: Shape・Scale・適応化

• Large Batch Training of Convolutional Networks — OP024。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Large Batch Optimization for Deep Learning: Training BERT in 76 minutes — OP025。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Don’t Use Large Mini-Batches, Use Local SGD — OP031。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法、Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Accelerated Large Batch Optimization of BERT Pretraining in 54 minutes — OB003。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Adaptive Sampling Strategies for Stochastic Optimization — OB004。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• AdAdaGrad: Adaptive Batch Size Schemes for Adaptive Gradient Methods — OB005。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Online Evolutionary Batch Size Orchestration for Scheduling Deep Learning Workloads in GPU Clusters — OB006。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• A Large Batch Optimizer Reality Check: Traditional, Generic Optimizers Suffice Across Batch Sizes — OP026。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Don’t Decay the Learning Rate, Increase the Batch Size — OP040。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Large-Scale Deep Learning Optimizations: A Comprehensive Survey — OB007。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• The Geometry of Sign Gradient Descent — OB008。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Optimization-Induced Dynamics of Lipschitz Continuity in Neural Networks — OB009。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Exact Risk Curves of signSGD in High-Dimensions: Quantifying Preconditioning and Noise-Compression Effects — OB010。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Scaling Collapse Reveals Universal Dynamics in Compute-Optimally Trained Neural Networks — OB011。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Local SGD Converges Fast and Communicates Little — OP030。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Communication-Efficient Learning of Deep Networks from Decentralized Data — OP036。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Can Decentralized Algorithms Outperform Centralized Algorithms? A Case Study for Decentralized Parallel Stochastic Gradient Descent — OB012。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Extrapolation for Large-batch Training in Deep Learning — OB013。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• DiLoCo: Distributed Low-Communication Training of Language Models — OP033。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Asynchronous Local-SGD Training for Language Modeling — OP034。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Distributed Deep Learning in Open Collaborations — OB014。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• SlowMo: Improving Communication-Efficient Distributed SGD with Slow Momentum — OP032。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Locally Adaptive Federated Learning — OB015。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Layer-wise and Dimension-wise Locally Adaptive Federated Learning — OB016。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Federated Learning with Buffered Asynchronous Aggregation — OB017。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Decentralized Stochastic Optimization and Gossip Algorithms with Compressed Communication — OB018。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Scalable Training of Deep Learning Machines by Incremental Block Training with Intra-block Parallel Optimization and Blockwise Model-Update Filtering — OB019。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Gradient-based Hyperparameter Optimization through Reversible Learning — OP049。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Online Learning Rate Adaptation with Hypergradient Descent — OP053。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• The Road Less Scheduled — OP027。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• The Surprising Agreement Between Convex Optimization Theory and Learning-Rate Scheduling for Large Model Training — OP028。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Random Search for Hyper-Parameter Optimization — OA001。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Practical Bayesian Optimization of Machine Learning Algorithms — OA002。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Forward and Reverse Gradient-Based Hyperparameter Optimization — OP050。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Understanding Short-Horizon Bias in Stochastic Meta-Optimization — OA003。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Marthe: Scheduling the Learning Rate Via Online Hypergradients — OA004。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• No More Pesky Learning Rates — OA005。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Population Based Training of Neural Networks — OP054。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Neural Optimizer Search with Reinforcement Learning — OP056。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Learning to learn by gradient descent by gradient descent — OP055。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Learned Optimizers that Scale and Generalize — OA006。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Super-Convergence: Very Fast Training of Neural Networks Using Large Learning Rates — OA007。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Descending through a Crowded Valley - Benchmarking Deep Learning Optimizers — OA008。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Power Scheduler: A Batch Size and Token Number Agnostic Learning Rate Scheduler — OA009。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• An Automatic Learning Rate Schedule Algorithm for Achieving Faster Convergence and Steeper Descent — OA010。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly — OA011。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Optimization Hyper-parameter Laws for Large Language Models — OA012。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• A Multi-Power Law for Loss Curve Prediction Across Learning Rate Schedules — OA013。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Relationship between Batch Size and Number of Steps Needed for Nonconvex Optimization of Stochastic Gradient Descent using Armijo Line Search — OA014。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• kDecay: Just adding k-decay items on Learning-Rate Schedule to improve Neural Networks — OA015。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Optimizer Benchmarking Needs to Account for Hyperparameter Tuning — OA016。出現元：Learning Rate Schedule Survey: Shape・Scale・適応化

• Old Optimizer, New Norm: An Anthology — OP018。出現元：Muonの数理: Frobenius・スペクトル・核ノルムと最急降下、Muon Optimizer: Newton–Schulz直交化の定義・理論・実装上の注意

• Natural Gradient Works Efficiently in Learning — OP080。出現元：NGD: Natural Gradient

• New Insights and Perspectives on the Natural Gradient Method — OB001。出現元：NGD: Natural Gradient

• Limitations of the Empirical Fisher Approximation for Natural Gradient Descent — MD07。出現元：NGD: Natural Gradient

• Sharpness-Aware Minimization for Efficiently Improving Generalization — OP045。出現元：SAM: Sharpness Aware Minimization

• Why Does Sharpness-Aware Minimization Generalize Better Than SGD? — OA017。出現元：SAM: Sharpness Aware Minimization

• Normalization Layers Are All That Sharpness-Aware Minimization Needs — OP062。出現元：SAM: Sharpness Aware Minimization

• How Does Sharpness-Aware Minimization Minimize Sharpness? — OP047。出現元：SAM: Sharpness Aware Minimization

• Sharpness-Aware Minimization Leads to Low-Rank Features — OP046。出現元：SAM: Sharpness Aware Minimization

• Rethinking Sharpness-Aware Minimization as Variational Inference — OA018。出現元：SAM: Sharpness Aware Minimization

• SAM as an Optimal Relaxation of Bayes — OA019。出現元：SAM: Sharpness Aware Minimization

• When Do Flat Minima Optimizers Work? — OA020。出現元：SAM: Sharpness Aware Minimization

• On the Maximum Hessian Eigenvalue and Generalization — OA021。出現元：SAM: Sharpness Aware Minimization

• The Hessian perspective into the Nature of Convolutional Neural Networks — OA022。出現元：SAM: Sharpness Aware Minimization

• Escaping Saddle Points for Effective Generalization on Class-Imbalanced Data — OA023。出現元：SAM: Sharpness Aware Minimization

• Latent Space Oddity: on the Curvature of Deep Generative Models — OA024。出現元：SAM: Sharpness Aware Minimization

• Noise Stability Optimization for Finding Flat Minima: A Hessian-based Regularization Approach — OA025。出現元：SAM: Sharpness Aware Minimization

• Improved Deep Neural Network Generalization Using m-Sharpness-Aware Minimization — OA026。出現元：SAM: Sharpness Aware Minimization

• Make Sharpness-Aware Minimization Stronger: A Sparsified Perturbation Approach — OA027。出現元：SAM: Sharpness Aware Minimization

• TRAM: Bridging Trust Regions and Sharpness Aware Minimization — OA028。出現元：SAM: Sharpness Aware Minimization

• Random Sharpness-Aware Minimization — OA029。出現元：SAM: Sharpness Aware Minimization

• Randomized Sharpness-Aware Training for Boosting Computational Efficiency in Deep Learning — OA030。出現元：SAM: Sharpness Aware Minimization

• Towards Efficient and Scalable Sharpness-Aware Minimization — OA031。出現元：SAM: Sharpness Aware Minimization

• AdaSAM: Boosting Sharpness-Aware Minimization with Adaptive Learning Rate and Momentum for Training Deep Neural Networks — OA032。出現元：SAM: Sharpness Aware Minimization

• ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networks — OP087。出現元：SAM: Sharpness Aware Minimization

• When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations — OA033。出現元：SAM: Sharpness Aware Minimization

• Improving Sharpness-Aware Minimization with Fisher Mask for Better Generalization on Language Models — OA034。出現元：SAM: Sharpness Aware Minimization

• Improving Shape Awareness and Interpretability in Deep Networks Using Geometric Moments — OA035。出現元：SAM: Sharpness Aware Minimization

• Stochastic Collapse: How Gradient Noise Attracts SGD Dynamics Towards Simpler Subnetworks — OA036。出現元：SAM: Sharpness Aware Minimization

• Beyond Deep Ensembles: A Large-Scale Evaluation of Bayesian Deep Learning under Distribution Shift — OA037。出現元：SAM: Sharpness Aware Minimization

• The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization — GE136, OA038。出現元：SAM: Sharpness Aware Minimization

• The Implicit Regularization of Momentum Gradient Descent with Early Stopping — OA039。出現元：SAM: Sharpness Aware Minimization

• Shampoo: Preconditioned Stochastic Tensor Optimization — OP006。出現元：Shampoo: Shampoo Optimizer

• Scalable Second Order Optimization for Deep Learning — OP007。出現元：Shampoo: Shampoo Optimizer

• A New Perspective on Shampoo’s Preconditioner — OB020。出現元：Shampoo: Shampoo Optimizer

• PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel — SY004。出現元：Parallel Training

未再検証の候補一覧

• Measuring the Effects of Data Parallelism on Neural Network Training — 未再検証・旧ラベル。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録

• Accelerating Neural Network Training: An Analysis of the AlgoPerf Competition — 未再検証・旧ラベル。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録

• Priya Kasimbeg — 未再検証・旧ラベル。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録

• Sourabh Medapati — 未再検証・旧ラベル。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録

• Shankar Krishnan — 未再検証・旧ラベル。出現元：Student Researcher @Google DeepMind 2024-2025 備忘録

• ICML — 未再検証・旧ラベル。出現元：Microsoft Research Intern 2024 備忘録

• (TMLR2023 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (TMLR2025 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (IJCNN2023 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (ICLR2026 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (ICML2024 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (AISTATS2023 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (TMLR2025 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (ICLR2024 WS — 未再検証・旧ラベル。出現元：Looking for Next Position

• (Pattern Recognition Letters - PRL2025 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (TMLR2024 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (ICLR2025 — 未再検証・旧ラベル。出現元：Looking for Next Position

• (ICLR2026 — 未再検証・旧ラベル。出現元：Looking for Next Position

• Greedy Layer-Wise Training of Deep Networks — 未再検証・旧ラベル。出現元：2020 Automne Q4

• 原論文: Richard Zhang, ICML 2019 — 未再検証・旧ラベル。出現元：2021 Autome Q3

• Paper / BarsCTR: Open Benchmarking for Click-Through Rate Prediction — 未再検証・旧ラベル。出現元：使ってる実装関連

• PDF 機械学習で楽しむ JAX/NumPyro v0.3 — 未再検証・旧ラベル。出現元：JAX

• arXiv Submission Guidelines — 未再検証・旧ラベル。出現元：arXiv への upload

• NeurIPS2023 — 未再検証・旧ラベル。出現元：日報 2023年 4 月、2022年のスケジュール、ML Conference

• ICML — 未再検証・旧ラベル。出現元：2022年のスケジュール、ML Conference

• WACV — 未再検証・旧ラベル。出現元：2022年のスケジュール

• 1/26 — 未再検証・旧ラベル。出現元：2023年のスケジュール

• Workshop Website — 未再検証・旧ラベル。出現元：ACM HPC-AI 2026 School

• On the Implicit Geometry of Cross-Entropy Parameterizations for Label-Imbalanced Data — 未再検証・旧ラベル。出現元：AISTATS2023

• The Role of Codeword-to-Class Assignments in Error Correcting Codes: An Empirical Study — 未再検証・旧ラベル。出現元：AISTATS2023

• Do Bayesian Neural Networks Need To Be Fully Stochastic? — 未再検証・旧ラベル。出現元：AISTATS2023

• Toward Fairness in Text Generation via Mutual Information Minimization based on Importance Sampling — 未再検証・旧ラベル。出現元：AISTATS2023

• Domain Adaptation under Missingness Shift — 未再検証・旧ラベル。出現元：AISTATS2023

• Freeze then Train: Towards Provable Representation Learning under Spurious Correlations and Feature Noise — 未再検証・旧ラベル。出現元：AISTATS2023

• Adapting to Latent Subgroup Shifts via Concepts and Proxies — 未再検証・旧ラベル。出現元：AISTATS2023

• On the Neural Tangent Kernel Analysis of Randomly Pruned Neural Networks — 未再検証・旧ラベル。出現元：AISTATS2023

• Federated Learning under Distributed Concept Drift — 未再検証・旧ラベル。出現元：AISTATS2023

• On double-descent in uncertainty quantification in overparametrized models — 未再検証・旧ラベル。出現元：AISTATS2023

• Don’t be fooled: label leakage in explanation methods and the importance of their quantitative evaluation — 未再検証・旧ラベル。出現元：AISTATS2023

• Explicit Regularization in Overparametrized Models via Noise Injection — 未再検証・旧ラベル。出現元：AISTATS2023

• Covariate-informed Representation Learning to Prevent Posterior Collapse of iVAE — 未再検証・旧ラベル。出現元：AISTATS2023

• The ELBO of Variational Autoencoders Converges to a Sum of Entropies — 未再検証・旧ラベル。出現元：AISTATS2023

• Linear Convergence of Gradient Descent For Overparametrized Finite Width Two-Layer Linear Networks With General Initialization — 未再検証・旧ラベル。出現元：AISTATS2023

• Mind the (optimality) Gap: A Gap-Aware Learning Rate Scheduler for Adversarial Nets — 未再検証・旧ラベル。出現元：AISTATS2023

• Riemannian Accelerated Gradient Methods via Extrapolation — 未再検証・旧ラベル。出現元：AISTATS2023

• Global-Local Regularization Via Distributional Robustness — 未再検証・旧ラベル。出現元：AISTATS2023

• Unifying local and global model explanations by functional decomposition of low dimensional structures — 未再検証・旧ラベル。出現元：AISTATS2023

• arxiv — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Accelerating neural network training: An analysis of the AlgoPerf competition (ICLR2025) — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Caspr Adaptive / Combining Axes Preconditioners through Kronecker Approximation for Deep Learning — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• AdEMAMIX — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Learning Rate Grafting: Transferability of Optimizer Tuning / Naman Agarwal, Rohan Anil, Elad Hazan, Tomer Koren, Cyril Zhang (ICLR2022 Rejected) — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Can We Remove the Square-Root in Adaptive Gradient Methods? A Second-Order Perspective / Wu Lin, Felix Dangel, Runa Eschenhagen, Juhan Bae, Richard E. Turner, Alireza Makhzani (ICML2024) — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Making the Last Iterate of SGD Information Theoretically Optimal — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Optimal Linear Decay Learning Rate Schedules and Further Refinements — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Niccolò Ajroldi — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Understanding Optimization in Deep Learning with Central Flows — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025、ICLR2025

• Challenges in Training PINNs: A Loss Landscape Perspective — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• The Road Less Scheduled — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• Directional Smoothness and Gradient Methods: Convergence and Adaptivity — 未再検証・旧ラベル。出現元：AlgoPerf Workshop 2025

• list of accepted paper — 未再検証・旧ラベル。出現元：ICCV2023

• Workshop on Uncertainty Quantification for Computer Vision — 未再検証・旧ラベル。出現元：ICCV2023

• pdf — 未再検証・旧ラベル。出現元：ICCV2023

• pdf — 未再検証・旧ラベル。出現元：ICCV2023

• pdf — 未再検証・旧ラベル。出現元：ICCV2023

• pdf — 未再検証・旧ラベル。出現元：ICCV2023

• pdf — 未再検証・旧ラベル。出現元：ICCV2023

• pdf — 未再検証・旧ラベル。出現元：ICCV2023

• Sharpness-Aware Minimization: General Analysis and Improved Rates — 未再検証・旧ラベル。出現元：ICLR2025

• Test-time Adaptation for Regression by Subspace Alignment — 未再検証・旧ラベル。出現元：ICLR2025

• What Does It Mean to Be a Transformer? Insights from a Theoretical Hessian Analysis — 未再検証・旧ラベル。出現元：ICLR2025

• Multimodal Lego: Model Merging and Fine-Tuning Across Topologies and Modalities in Biomedicine — 未再検証・旧ラベル。出現元：ICLR2025

• Do Stochastic, Feel Noiseless: Stable Stochastic Optimization via a Double Momentum Mechanism — 未再検証・旧ラベル。出現元：ICLR2025

• The Hyperfitting Phenomenon: Sharpening and Stabilizing LLMs for Open-Ended Text Generation — 未再検証・旧ラベル。出現元：ICLR2025

• A Geometric Framework for Understanding Memorization in Generative Models — 未再検証・旧ラベル。出現元：ICLR2025

• Instance-dependent Early Stopping — 未再検証・旧ラベル。出現元：ICLR2025

• Inverse Scaling: When Bigger Isn’t Better — 未再検証・旧ラベル。出現元：ICLR2025

• Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass — 未再検証・旧ラベル。出現元：ICLR2025

• Accelerating Training with Neuron Interaction and Nowcasting Networks — 未再検証・旧ラベル。出現元：ICLR2025

• Expressivity of Neural Networks with Random Weights and Learned Biases — 未再検証・旧ラベル。出現元：ICLR2025

• A Second-Order Perspective on Model Compositionality and Incremental Learning — 未再検証・旧ラベル。出現元：ICLR2025

• Mitigating Parameter Interference in Model Merging via Sharpness-Aware Fine-Tuning — 未再検証・旧ラベル。出現元：ICLR2025

• Geometric Inductive Biases of Deep Networks: The Role of Data and Architecture — 未再検証・旧ラベル。出現元：ICLR2025

• arxiv — 未再検証・旧ラベル。出現元：ICML2024

• ICML portal — 未再検証・旧ラベル。出現元：ICML2024

• paper — 未再検証・旧ラベル。出現元：ICML2024

• Information Complexity of Stochastic Convex Optimization: Applications to Generalization and Memorization — 未再検証・旧ラベル。出現元：ICML2024

• LEVI: Generalizable Fine-tuning via Layer-wise Ensemble of Different Views — 未再検証・旧ラベル。出現元：ICML2024

• Interpretability Illusions in the Generalization of Simplified Models — 未再検証・旧ラベル。出現元：ICML2024

• Position: Leverage Foundational Models for Black-Box Optimization — 未再検証・旧ラベル。出現元：ICML2024

• General Analysis of LMO-based Optimizers: Beyond Bounded Variance — 未再検証・旧ラベル。出現元：ICML2026 参加記

• On the Role of Batch Size in Stochastic Conditional Gradient Methods — 未再検証・旧ラベル。出現元：ICML2026 参加記

• On the Interaction of Batch Noise, Adaptivity, and Compression, under $(L_0,L_1)$-Smoothness: An SDE Approach — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• From Muon to Gluon: Bridging Theory and Practice of LMO-based Optimizers for LLMs — 未再検証・旧ラベル。出現元：ICML2026 参加記

• Depth scaling and Muon enable balanced expert usage in MoE training — 未再検証・旧ラベル。出現元：ICML2026 参加記

• MuLoCo: Muon is a practical inner optimizer for DiLoCo — 未再検証・旧ラベル。出現元：ICML2026 参加記

• One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining — 未再検証・旧ラベル。出現元：ICML2026 参加記

• Controlled LLM Training on Spectral Sphere — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• LoRDO: Distributed Low-Rank Optimization with Infrequent Communication — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Factored Gossip DiLoCo: Reducing Blocking Communication within DiLoCo — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• OLion: Approaching the Hadamard Ideal by Intersecting Spectral and L inf Implicit Biases — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Exploiting weight-space symmetries for approximating curvature — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Rethinking 1-bit Optimization Leveraging Pre-trained Large Language Models — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Unveiling the Potential of Quantization with MXFP4: Strategies for Quantization Error Reduction — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• PRISM: Distribution-free Adaptive Computation of Matrix　Functions for Accelerating Neural Network Training — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• FPTQuant: Function-Preserving Transforms for LLM Quantization — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Gradient Smoothing: Coupling Layer-wise Updates for Improved Optimization — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• A Theory of How Pretraining Shapes Inductive Bias in Fine-Tuning — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Can Muon Fine-tune Adam-Pretrained Models? — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• ECO: Quantized Training without Full-Precision Master Weights — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• GradientStabilizer: Fix the Norm, Not the Gradient — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• AdaGC: Enhancing LLM Pretraining Stability via Adaptive Gradient Clipping — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• cs.ubc.ca/~schmidtm/Documents/2026_ICML_Tutorial.pdf — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/invited-talk/67264 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/media/icml-2026/Slides/71055.pdf — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/63313 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/60589 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/65256 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/61920 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/63590 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/62433 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/63669 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/61112 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/poster/61819 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/invited-talk/67274 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• icml.cc/virtual/2026/oral/71156 — 未再検証・旧ラベル。出現元：ICML2026 参加記、ICML2026

• Philip Zmushko — 未再検証・旧ラベル。出現元：ICML2026

• Shinji Ito — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Bai Cong — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Ryosuke Yamaki — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Kotaro Yoshida — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Mohammad Pezeshki — 未再検証・旧ラベル。出現元：NeurIPS 2024

• url — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Pseudo-Asynchronous Local SGD: Robust and Efficient Data-Parallel Training — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Addax: Utilizing Zeroth-Order Gradients to Improve Memory Efficiency and Performance of SGD for Fine-Tuning Language Models — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Scaling Collapse Reveals Universal Dynamics in Compute-Optimally Trained Neural Networks — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Second-Order Forward-Mode Automatic Differentiation for Optimization — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Don’t Be So Positive: Negative Step Sizes in Second-Order Methods — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Scaling Collapse Reveals Universal Dynamics in Compute-Optimally Trained Neural Networks — 未再検証・旧ラベル。出現元：NeurIPS 2024

• SOAP: Improving and Stabilizing Shampoo using Adam — 未再検証・旧ラベル。出現元：NeurIPS 2024

• μLO: Compute-Efficient Meta-Generalization of Learned Optimizers — 未再検証・旧ラベル。出現元：NeurIPS 2024

• VeLO: Training Versatile Learned Optimizers by Scaling Up — 未再検証・旧ラベル。出現元：NeurIPS 2024

• Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning — 未再検証・旧ラベル。出現元：NeurIPS 2024

• 深層学習の汎化の謎をめぐって — 未再検証・旧ラベル。出現元：Learning Resource

• Link — 未再検証・旧ラベル。出現元：ML Conference

• Link — 未再検証・旧ラベル。出現元：ML Conference

• Link — 未再検証・旧ラベル。出現元：ML Conference

• Link — 未再検証・旧ラベル。出現元：ML Conference

• Fishr: Invariant Gradient Variances for Out-of-distribution Generalization — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• On the Variance of the Fisher Information for Deep Learning — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• ViViT: Curvature access through the generalized Gauss-Newton’s low-rank structure — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• A Loss Curvature Perspective on Training Instability in Deep Learning — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• SGD: The Role of Implicit Regularization, Batch-size and Multiple-epochs — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• On the Explicit Role of Initialization on the Convergence and Generalization Properties of Overparametrized Linear Networks — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• Beyond BatchNorm: Towards a General Understanding of Normalization in Deep Learning — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• A Closer Look at the Optimization Landscapes of Generative Adversarial Networks — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• Hessian based analysis of SGD for Deep Nets: Dynamics and Generalization — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• Gradient Descent Happens in a Tiny Subspace — 未再検証・旧ラベル。出現元：読まなきゃと思ってる Paper List

• How to Read a Pape — 未再検証・旧ラベル。出現元：Paper Writing 関連

• ref — 未再検証・旧ラベル。出現元：GAN: Generative Adversarial Networks

• cs.cmu.edu/~morency/MMML-Tutorial-ACL2017.pdf — 未再検証・旧ラベル。出現元：MultiModal

• Calibrated ensembles - a simple way to mitigate ID-OOD accuracy tradeoffs — 未再検証・旧ラベル。出現元：Calibration

• When Training and Test Sets are Different: Characterising Learning Transfer — 未再検証・旧ラベル。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• 自分でやったまとめ PDF — 未再検証・旧ラベル。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• ref — 未再検証・旧ラベル。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• slide — 未再検証・旧ラベル。出現元：Distribution Shift入門: Covariate・Label・Concept Shift

• A Critical Analysis of Distribution Shift — 未再検証・旧ラベル。出現元：Distribution Shift入門: Covariate・Label・Concept Shift、OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Fishr: Invariant Gradient Variances for Out-of-distribution Generalization — 未再検証・旧ラベル。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Loss Function Learning for Domain Generalization by Implicit Gradient — 未再検証・旧ラベル。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• Head2Toe: Utilizing Intermediate Representations for Better OOD Generalization — 未再検証・旧ラベル。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• How does a neural network’s architecture impact its robustness to noisy labels? — 未再検証・旧ラベル。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• IRM Slides — 未再検証・旧ラベル。出現元：OOD入門: Distribution Shift・Domain Generalization・OOD Detection

• 深層学習の数理：ランダム行列と統計力学的視点 — 未再検証・旧ラベル。出現元：NTK: Neural Tangent Kernel

• Google Research Automatic differentiation — 未再検証・旧ラベル。出現元：AD: Automatic Differentiation

• UBC Automatic Differentiation (1) — 未再検証・旧ラベル。出現元：AD: Automatic Differentiation

• Ju Sun (孙举) University of Minnesota のスライド — 未再検証・旧ラベル。出現元：AD: Automatic Differentiation

• Vector Institute Grosse 先生の資料 — 未再検証・旧ラベル。出現元：AD: Automatic Differentiation

• 拡散と流れに基づく学習と推論（岡野原大輔） — 未再検証・旧ラベル。出現元：Flow Matching入門: 連続の式・条件付き目的・ODE生成

• Invariant Risk Minimization / Slide — 未再検証・旧ラベル。出現元：Bi-Level Optimization

• 情報数理工学・コンピュータサイエンス実験第二大規模連立一次方程式に対する共役勾配法 — 未再検証・旧ラベル。出現元：CGD: Conjugate Gradient

• 最急降下法と共役傾斜法について — 未再検証・旧ラベル。出現元：CGD: Conjugate Gradient

• 反復学習制御に対する共役勾配法と準ニュートン法 — 未再検証・旧ラベル。出現元：CGD: Conjugate Gradient

• Mustafa et al. 2018-11-07-Large-Batch-Training-Mustafa.pdf — 未再検証・旧ラベル。出現元：Large-Batch Training Survey: スケーリング則・限界・最適化手法

• Jiajun Shen — 未再検証・旧ラベル。出現元：Local SGD入門: Periodic Averaging・FedAvg・非同期・分散最適化

• Natural gradients and K-FAC — 未再検証・旧ラベル。出現元：NGD: Natural Gradient

• 解説 PDF — 未再検証・旧ラベル。出現元：NGD: Natural Gradient

• Provable Sharpness-Aware Minimization with Adaptive Learning Rate — 未再検証・旧ラベル。出現元：SAM: Sharpness Aware Minimization

• Sharpness-Aware Minimization in Large-Batch Training: Training Vision Transformer In Minutes — 未再検証・旧ラベル。出現元：SAM: Sharpness Aware Minimization