# 論文・研究業績 — Hiroki Naganuma

Canonical HTML: https://hiroki11x.github.io/ja/publications/

## ジャーナル論文

- [What do near-optimal learning rate schedules look like?](https://hiroki11x.github.io/ja/publications/what-do-near-optimal-learning-rate-schedules-look-like/) — TMLR 2026 (2026); 機械学習の最適化
- [Pseudo-Asynchronous Local SGD: Robust and Efficient Data-Parallel Training](https://hiroki11x.github.io/ja/publications/pseudo-asynchronous-local-sgd/) — TMLR 2025 (2025); スケーラブルな事前学習と分散学習
- [An Empirical Study of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration](https://hiroki11x.github.io/ja/publications/pre-trained-model-selection-for-ood-generalization-and-calibration/) — TMLR 2025 (2025); 汎化と基盤モデル
- [Geometric Insights into Focal Loss: Reducing Curvature for Enhanced Model Calibration](https://hiroki11x.github.io/ja/publications/geometric-insights-into-focal-loss/) — PRL (2025); 機械学習の最適化; 汎化と基盤モデル
- [Towards Understanding Variants of Invariant Risk Minimization from the Perspective of Calibration](https://hiroki11x.github.io/ja/publications/invariant-risk-minimization-variants-and-calibration/) — TMLR 2024 (2024); 汎化と基盤モデル
- [Empirical Study on Optimizer Selection for Out-of-Distribution Generalizations](https://hiroki11x.github.io/ja/publications/optimizer-selection-for-ood-generalization/) — TMLR 2023 (2023); 機械学習の最適化; 汎化と基盤モデル
- [Generalization Measures under Controlled Covariate Shift: A Regime-Aware Benchmark](https://hiroki11x.github.io/ja/publications/generalization-measures-beyond-iid/) — TMLR 2026 (2026); 汎化と基盤モデル

## 国際会議論文

- [Adaptive Batch Sizes Using Non-Euclidean Gradient Noise Scales for Stochastic Sign and Spectral Descent](https://hiroki11x.github.io/ja/publications/adaptive-batch-sizes-using-non-euclidean-gradient-noise-scales/) — ICML 2026 (2026); スケーラブルな事前学習と分散学習; 機械学習の最適化
- [On Fairness of Task Arithmetic: The Role of Task Vectors](https://hiroki11x.github.io/ja/publications/fairness-of-task-arithmetic/) — ICLR 2026 (2026); 汎化と基盤モデル
- [DiTaC: Conditioning Task Vectors via Distillation for Robust Model Merging](https://hiroki11x.github.io/ja/publications/ditac-conditioning-task-vectors-via-distillation/) — ICLR 2026 (2026); 汎化と基盤モデル
- [Mastering Task Arithmetic: τJp as a Key Indicator for Weight Disentanglement](https://hiroki11x.github.io/ja/publications/mastering-task-arithmetic-tau-jp/) — ICLR 2025 (2025); 汎化と基盤モデル
- [No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths](https://hiroki11x.github.io/ja/publications/no-wrong-turns-neural-network-optimization-paths/) — ICML 2024 (2024); 機械学習の最適化
- [How Image Corruption and Perturbation Affect Out-Of-Distribution Generalization and Calibration](https://hiroki11x.github.io/ja/publications/image-corruption-ood-generalization-and-calibration/) — IJCNN 2023 (2023); 汎化と基盤モデル
- [Conjugate Gradient Method for Generative Adversarial Networks](https://hiroki11x.github.io/ja/publications/conjugate-gradient-method-for-generative-adversarial-networks/) — AISTATS 2023 (2023); 機械学習の最適化
- [Optimal Transport Meets Noisy Label Robust Loss and MixUp Regularization for Domain Adaptation](https://hiroki11x.github.io/ja/publications/optimal-transport-noisy-label-robust-loss-and-mixup/) — CoLLAs 2022 (2022); 汎化と基盤モデル
- [Accelerating Convolutional Neural Networks Using Low Precision Arithmetic](https://hiroki11x.github.io/ja/publications/accelerating-convolutional-neural-networks-using-low-precision-arithmetic/) — HPC Asia 2018 (2018); スケーラブルな事前学習と分散学習
- [Accelerating Matrix Multiplication in Deep Learning by using Low-Rank Approximation](https://hiroki11x.github.io/ja/publications/accelerating-matrix-multiplication-using-low-rank-approximation/) — HPCS 2017 (2017); スケーラブルな事前学習と分散学習

## プレプリント・査読中

- [From Inner Randomness to Outer Stability](https://hiroki11x.github.io/ja/publications/from-inner-randomness-to-outer-stability/) — Under Review (2026); 機械学習の最適化
- [Which Geometry on Which Layer? A Principled Criterion for Mixed-Optimizer Training](https://hiroki11x.github.io/ja/publications/which-geometry-on-which-layer/) — Under Review (2026); スケーラブルな事前学習と分散学習; 機械学習の最適化
- [Orth-Dion: Eliminating Geometric Mismatch in Distributed Low-Rank Spectral Optimization](https://hiroki11x.github.io/ja/publications/orth-dion/) — Under Review (2026); スケーラブルな事前学習と分散学習; 機械学習の最適化
- [Convergence Bound and Critical Batch Size of Muon Optimizer](https://hiroki11x.github.io/ja/publications/convergence-bound-and-critical-batch-size-of-muon/) — Under Review (2025); スケーラブルな事前学習と分散学習; 機械学習の最適化
- [When Does Alignment Help? A Comparative Study of DCCA and Fusion-Based Approaches for Multi-modal Chest X-ray Analysis](https://hiroki11x.github.io/ja/publications/when-does-alignment-help-multimodal-chest-x-ray-analysis/) — SSRN (2025); 汎化と基盤モデル
- [Augmenting NER Datasets with LLMs: Towards Automated and Refined Annotation](https://hiroki11x.github.io/ja/publications/augmenting-ner-datasets-with-llms/) — arXiv (2024); 汎化と基盤モデル
- [Takeuchi's Information Criteria as Generalization Measures for DNNs Close to NTK Regime](https://hiroki11x.github.io/ja/publications/takeuchi-information-criteria-for-dnns-close-to-ntk/) — arXiv (2021); 機械学習の最適化; 汎化と基盤モデル

## ワークショップ論文

- [The Geometry of Spectral Gradient Descent: Layerwise Criteria for SignSGD vs SpecSGD](https://hiroki11x.github.io/ja/publications/geometry-of-spectral-gradient-descent/) — ICLR 2026 Workshop (2026); 機械学習の最適化
- [Smoothness-Adaptive Sharpness Aware Minimization for Finding Flatter Minima](https://hiroki11x.github.io/ja/publications/smoothness-adaptive-sharpness-aware-minimization/) — ICLR 2024 Workshop (2024); 機械学習の最適化; 汎化と基盤モデル
- [Story-to-Images Translation: Leveraging Diffusion Models and Large Language Models for Sequence Image Generation](https://hiroki11x.github.io/ja/publications/story-to-images-translation/) — NarSUM 2023 (2023); 汎化と基盤モデル
- [On the Interplay of Curvature, Calibration and Out-of-Distribution Generalization: Insights from SAM and Focal Loss Analyses](https://hiroki11x.github.io/ja/publications/curvature-calibration-and-ood-generalization/) — UNCV 2023 (2023); 機械学習の最適化; 汎化と基盤モデル
- [Necessary and Sufficient Hypothesis of Curvature: Understanding Connection Between Out-of-Distribution Generalization and Calibration](https://hiroki11x.github.io/ja/publications/necessary-and-sufficient-hypothesis-of-curvature/) — ICLR 2023 Workshop (2023); 機械学習の最適化; 汎化と基盤モデル
- [Towards Understanding the Relationship of Batch Size and Iterations in Deep Learning](https://hiroki11x.github.io/ja/publications/batch-size-and-iterations-in-deep-learning/) — MLSS 2020 (2020); スケーラブルな事前学習と分散学習; 機械学習の最適化
- [On Empirical Analysis of Layer-wise Learning Rate Schedule](https://hiroki11x.github.io/ja/publications/layer-wise-learning-rate-schedule/) — ACML 2019 Workshop (2019); 機械学習の最適化
- [A Performance Improvement Approach for Second-Order Optimization in Large Mini-batch Training](https://hiroki11x.github.io/ja/publications/second-order-optimization-in-large-mini-batch-training/) — HPML 2019 (2019); スケーラブルな事前学習と分散学習; 機械学習の最適化
- [Noise Injection Leads to Better Generalization in Large Mini-Batch Training](https://hiroki11x.github.io/ja/publications/noise-injection-for-large-mini-batch-training/) — Tokyo Tech-Stony Brook Joint Meeting (2019); スケーラブルな事前学習と分散学習; 機械学習の最適化