Hiroki Naganuma

29 papers A* 2A 1B 2Journal 21Unranked 3

Year	Rank	Type	Title / Venue / Authors
2026	J	`jnl`	Adaptive Batch Sizes Using Non-Euclidean Gradient Noise Scales for Stochastic Sign and Spectral Descent. CoRR Hiroki Naganuma, Shagun Gupta, Youssef Briki, Ioannis Mitliagkas, Irina Rish, Parameswaran Raman, Hao-Jun Michael Shi
2026	J	`jnl`	Revisiting Generalization Measures Beyond IID: An Empirical Study under Distributional Shift. CoRR Sora Nakai, Youssef Fadhloun, Kacem Mathlouthi, Kotaro Yoshida, Ganesh Talluri, Ioannis Mitliagkas, Hiroki Naganuma
2026	J	`jnl`	Takeuchi's Information Criteria as Generalization Measures for DNNs Close to NTK Regime. CoRR Hiroki Naganuma, Taiji Suzuki, Rio Yokota, Masahiro Nomura, Kohta Ishikawa, Ikuro Sato
2026	J	`jnl`	What do near-optimal learning rate schedules look like? CoRR Hiroki Naganuma, Atish Agarwala, Priya Kasimbeg, George E. Dahl
2025	J	`jnl`	An Empirical Study of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration. Trans. Mach. Learn. Res. Hiroki Naganuma, Ryuichiro Hataya, Kotaro Yoshida, Ioannis Mitliagkas
2025	J	`jnl`	Analysis of Muon's Convergence and Critical Batch Size. CoRR Naoki Sato, Hiroki Naganuma, Hideaki Iiduka
2025	J	`jnl`	DisTaC: Conditioning Task Vectors via Distillation for Robust Model Merging. CoRR Kotaro Yoshida, Yuji Naraki, Takafumi Horie, Ryotaro Shimizu, Hiroki Naganuma
2025	J	`jnl`	Geometric insights into focal loss: Reducing curvature for enhanced model calibration. Pattern Recognit. Lett. Masanari Kimura, Hiroki Naganuma
2025	A*	`conf`	Mastering Task Arithmetic: τJp as a Key Indicator for Weight Disentanglement. ICLR Kotaro Yoshida, Yuji Naraki, Takafumi Horie, Ryosuke Yamaki, Ryotaro Shimizu, Yuki Saito, Julian J. McAuley, Hiroki Naganuma
2025	J	`jnl`	On Fairness of Task Arithmetic: The Role of Task Vectors. CoRR Hiroki Naganuma, Kotaro Yoshida, Laura Gomezjurado Gonzalez, Takafumi Horie, Yuji Naraki, Ryotaro Shimizu
2025	J	`jnl`	Pseudo-Asynchronous Local SGD: Robust and Efficient Data-Parallel Training. CoRR Hiroki Naganuma, Xinzhi Zhang, Man-Chung Yue, Ioannis Mitliagkas, Philipp A. Witte, Russell J. Hewett, Yin Tat Lee
2025	J	`jnl`	Pseudo-Asynchronous Local SGD: Robust and Efficient Data-Parallel Training. Trans. Mach. Learn. Res. Hiroki Naganuma, Xinzhi Zhang, Man-Chung Yue, Ioannis Mitliagkas, Russell J. Hewett, Philipp A. Witte, Yin Tat Lee
2024	J	`jnl`	Augmenting NER Datasets with LLMs: Towards Automated and Refined Annotation. CoRR Yuji Naraki, Ryosuke Yamaki, Yoshikazu Ikeda, Takafumi Horie, Hiroki Naganuma
2024	J	`jnl`	Geometric Insights into Focal Loss: Reducing Curvature for Enhanced Model Calibration. CoRR Masanari Kimura, Hiroki Naganuma
2024	A*	`conf`	No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths. ICML Charles Guille-Escuret, Hiroki Naganuma, Kilian Fatras, Ioannis Mitliagkas
2024	J	`jnl`	Towards Understanding Variants of Invariant Risk Minimization through the Lens of Calibration. CoRR Kotaro Yoshida, Hiroki Naganuma
2024	J	`jnl`	Towards Understanding Variants of Invariant Risk Minimization through the Lens of Calibration. Trans. Mach. Learn. Res. Kotaro Yoshida, Hiroki Naganuma
2023	J	`jnl`	An Empirical Investigation of Pre-trained Model Selection for Out-of-Distribution Generalization and Calibration. CoRR Hiroki Naganuma, Ryuichiro Hataya
2023	A	`conf`	Conjugate Gradient Method for Generative Adversarial Networks. AISTATS Hiroki Naganuma, Hideaki Iiduka
2023	J	`jnl`	Empirical Study on Optimizer Selection for Out-of-Distribution Generalization. Trans. Mach. Learn. Res. Hiroki Naganuma, Kartik Ahuja, Shiro Takagi, Tetsuya Motokawa, Rio Yokota, Kohta Ishikawa, Ikuro Sato, Ioannis Mitliagkas
2023	B	`conf`	How Image Corruption and Perturbation Affect Out-of-Distribution Generalization and Calibration. IJCNN Keigo Tada, Hiroki Naganuma
2023	J	`jnl`	No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths. CoRR Charles Guille-Escuret, Hiroki Naganuma, Kilian Fatras, Ioannis Mitliagkas
2023	—	`conf`	Story-to-Images Translation: Leveraging Diffusion Models and Large Language Models for Sequence Image Generation. NarSUM@MM Haruka Kumagai, Ryosuke Yamaki, Hiroki Naganuma
2022	J	`jnl`	Conjugate Gradient Method for Generative Adversarial Networks. CoRR Hiroki Naganuma, Hideaki Iiduka
2022	J	`jnl`	Empirical Study on Optimizer Selection for Out-of-Distribution Generalization. CoRR Hiroki Naganuma, Kartik Ahuja, Shiro Takagi, Tetsuya Motokawa, Rio Yokota, Kohta Ishikawa, Ikuro Sato, Ioannis Mitliagkas
2022	—	`conf`	Optimal Transport meets Noisy Label Robust Loss and MixUp Regularization for Domain Adaptation. CoLLAs Kilian Fatras, Hiroki Naganuma, Ioannis Mitliagkas
2022	J	`jnl`	Optimal transport meets noisy label robust loss and MixUp regularization for domain adaptation. CoRR Kilian Fatras, Hiroki Naganuma, Ioannis Mitliagkas
2019	B	`conf`	A Performance Improvement Approach for Second-Order Optimization in Large Mini-batch Training. CCGRID Hiroki Naganuma, Rio Yokota
2017	—	`conf`	Accelerating Matrix Multiplication in Deep Learning by Using Low-Rank Approximation. HPCS Kazuki Osawa, Akira Sekiya, Hiroki Naganuma, Rio Yokota