研究

Preprints

• Tomoya Matsumoto*, Shokichi Takakura*, Shun Takagi, and Satoshi Hasegawa: DPSQL+: A Differentially Private SQL Library with a Minimum Frequency Rule. (*Equal contribution)
• Akifumi Wachi, Hirota Kinoshita, Shokichi Takakura, Rei Higuchi, Taiji Suzuki: A Relative-Budget Theory for Reinforcement Learning with Verifiable Rewards in Large Language Model Reasoning.
• Shokichi Takakura, Akifumi Wachi, Rei Higuchi, Kohei Miyaguchi, Taiji Suzuki: Inference-Aware Meta-Alignment of LLMs via Non-Linear GRPO
• Shokichi Takakura, Seng Pei Liew, and Satoshi Hasegawa: Accelerating Differentially Private Federated Learning via Adaptive Extrapolation.

Journal papers (Refereed)

• Shokichi Takakura, and Kazuhiro Sato: Structured Output Feedback Control for Linear Quadratic Regulator Using Policy Gradient Method. IEEE Transactions on Automatic Control, vol. 69, no. 1, pp. 363-370, 2022.

International Conference papers (Refereed)

• Shokichi Takakura, Seng Pei Liew, Satoshi Hasegawa: Optimal Variance and Covariance Estimation under Differential Privacy in the Add-Remove Model and Beyond. 29th Annual Conference on Artificial Intelligence and Statistics (AISTATS).
• Shokichi Takakura, Seng Pei Liew, Satoshi Hasegawa: FedDuA: Doubly Adaptive Federated Learning. 29th Annual Conference on Artificial Intelligence and Statistics (AISTATS).
• Shokichi Takakura, and Taiji Suzuki: Mean-field Analysis on Two-layer Neural Networks from a Kernel Perspective. Proceedings of the 41th International Conference on Machine Learning (ICML2024), PMLR 235:47475-47509, 2024.
• Shokichi Takakura, and Taiji Suzuki: Approximation and Estimation Ability of Transformers for Sequence-to-Sequence Functions with Infinite Dimensional Input. Proceedings of the 40th International Conference on Machine Learning (ICML2023), PMLR 202:33416–33447, 2023.

Presentations

• Shokichi Takakura, Seng Pei Liew, and Satoshi Hasegawa: Optimal Variance and Covariance Estimation under Differential Privacy in the Add-Remove Model. 第18回データ工学と情報マネジメントに関するフォーラム (DEIM2026), Hyogo, Japan, March. 2026.
• Shokichi Takakura, Seng Pei Liew, and Satoshi Hasegawa: FedDuA: Doubly Adaptive Federated Learning. The 28th Information-Based Induction Science Workshop, Okinawa, Japan, Nov. 2025.
• Shokichi Takakura, Seng Pei Liew, and Satoshi Hasegawa: Accelerating Differentially Private Federated Learning via Adaptive Extrapolation. Will Synthetic Data Finally Solve the Data Access Problem? (ICLR 2025 workshop), Singapore, April. 2025.
• Shokichi Takakura, and Taiji Suzuki: カーネルの観点による二層ニューラルネットワークの平均場解析. Japanese Joint Statistical Meeting, Tokyo, Japan, Sep. 2024.
• Shokichi Takakura, and Taiji Suzuki: Mean-field Analysis on Two-layer Neural Networks from a Kernel Perspective. The 41th International Conference on Machine Learning (ICML2024), Vienna, Austria, Jul. 2024.
• Shokichi Takakura, and Taiji Suzuki: Mean-field Analysis on Two-layer Neural Networks from a Kernel Perspective. Workshop on Functional Inference and Machine Intelligence, Bristol, UK, Mar. 2024.
• Shokichi Takakura, and Taiji Suzuki: Grokking in Linear Diagonal Neural Networks. The 26th Information-Based Induction Science Workshop, Hukuoka, Japan, Oct. 2023.
• Shokichi Takakura, and Taiji Suzuki: Approximation and Estimation Ability of Transformers for Sequence-to-Sequence Functions with Infinite Dimensional Input. The 26th Information-Based Induction Science Workshop, Hukuoka, Japan, Oct. 2023.
• Shokichi Takakura, and Taiji Suzuki: 無限次元入力sequence-to-sequence関数に対するトランスフォーマーの近似及び推定能力. Japanese Joint Statistical Meeting, Kyoto, Japan, Sep. 2023.
• Shokichi Takakura, and Taiji Suzuki: Approximation and Estimation Ability of Transformers for Sequence-to-Sequence Functions with Infinite Dimensional Input. The 40th International Conference on Machine Learning (ICML2023), Hawaii, USA, Jul. 2023.
• Shokichi Takakura, and Kazuhiro Sato: 方策勾配法を用いた線形2次レギュレータの制約付き出力フィードバック制御. 第66回システム制御情報学会研究発表講演会 (SCI’22), Kyoto, Japan, May 2022.