Shunta Akiyama

15 papers A* 6Journal 9

Year	Rank	Type	Title / Venue / Authors
2026	J	`jnl`	Don't Eliminate Cut: Exponential Separations in LLM-Based Theorem Proving. CoRR Sho Sonoda, Shunta Akiyama, Yuya Uezato
2026	J	`jnl`	Why Agentic Theorem Prover Works: A Statistical Provability Theory of Mathematical Reasoning Models. CoRR Sho Sonoda, Shunta Akiyama, Yuya Uezato
2025	J	`jnl`	Block Coordinate Descent for Neural Networks Provably Finds Global Minima. CoRR Shunta Akiyama
2025	A*	`conf`	Survival Analysis via Density Estimation. ICML Hiroki Yanagisawa, Shunta Akiyama
2024	A*	`conf`	SILVER: Single-loop variance reduction and application to federated learning. ICML Kazusato Oko, Shunta Akiyama, Denny Wu, Tomoya Murata, Taiji Suzuki
2023	A*	`conf`	Diffusion Models are Minimax Optimal Distribution Estimators. ICML Kazusato Oko, Shunta Akiyama, Taiji Suzuki
2023	J	`jnl`	Diffusion Models are Minimax Optimal Distribution Estimators. CoRR Kazusato Oko, Shunta Akiyama, Taiji Suzuki
2023	A*	`conf`	Excess Risk of Two-Layer ReLU Neural Networks in Teacher-Student Settings and its Superiority to Kernel Methods. ICLR Shunta Akiyama, Taiji Suzuki
2022	J	`jnl`	Excess Risk of Two-Layer ReLU Neural Networks in Teacher-Student Settings and its Superiority to Kernel Methods. CoRR Shunta Akiyama, Taiji Suzuki
2022	J	`jnl`	Optimal design of lottery with cumulative prospect theory. CoRR Shunta Akiyama, Mitsuaki Obara, Yasushi Kawase
2022	J	`jnl`	Versatile Single-Loop Method for Gradient Estimator: First and Second Order Optimality, and its Application to Federated Learning. CoRR Kazusato Oko, Shunta Akiyama, Tomoya Murata, Taiji Suzuki
2021	A*	`conf`	Benefit of deep learning with non-convex noisy gradient descent: Provable excess risk bound and superiority to kernel methods. ICLR Taiji Suzuki, Shunta Akiyama
2021	A*	`conf`	On Learnability via Gradient Method for Two-Layer ReLU Neural Networks in Teacher-Student Setting. ICML Shunta Akiyama, Taiji Suzuki
2021	J	`jnl`	On Learnability via Gradient Method for Two-Layer ReLU Neural Networks in Teacher-Student Setting. CoRR Shunta Akiyama, Taiji Suzuki
2020	J	`jnl`	Benefit of deep learning with non-convex noisy gradient descent: Provable excess risk bound and superiority to kernel methods. CoRR Taiji Suzuki, Shunta Akiyama