論文表記の著者名
Taiji Suzuki
東京大学Machine Learning2022–2026 LIVE
対象論文 69 本
著者名は論文に記載された表記をそのまま使用しており、同一人物の名寄せは行っていません。
印刷日： 全論文
論文上の著者表記・公開時所属による集計です。人物の名寄せ・現在所属・全業績を表すものではありません。
対象論文
論文一覧
- Transformers as Measure-Theoretic Associative Memory: A Statistical Perspective and Minimax Optimality
- In-Context Learning Is Provably Bayesian Inference: A Generalization Theory for Meta-Learning
- Inference-Aware Meta-Alignment of LLMs via Non-Linear GRPO
- Inference-time Alignment with Rewards in Besov Spaces: Provable Advantages of Feature Learning and Multi-Step Policy Updates
- Provable Sample Efficiency of Curriculum Post-Training for Transformer Reasoning
- Zero-Flow Encoders
- Direct Distributional Optimization for Provable Alignment of Diffusion Models
- Flow matching achieves almost minimax optimal convergence
- On the Optimization and Generalization of Two-layer Transformers with Sign Gradient Descent
- Optimality and Adaptivity of Deep Neural Features for Instrumental Variable Regression
- State Space Models are Provably Comparable to Transformers in Dynamic Token Selection
- Transformers Provably Solve Parity Efficiently with Chain of Thought
- Weighted Point Set Embedding for Multimodal Contrastive Learning Toward Optimal Similarity Metric
- Direct Density Ratio Optimization: A Statistically Consistent Approach to Aligning Large Language Models
- Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillation
- Mixture of Experts Provably Detect and Learn the Latent Cluster Structure in Gradient-Based Learning
- Nonlinear transformers can perform inference-time feature learning
- On the Role of Label Noise in the Feature Learning Process
- Propagation of Chaos for Mean-Field Langevin Dynamics and its Application to Model Ensemble
- Provable In-Context Vector Arithmetic via Retrieving Task Concepts
- Quantifying Memory Utilization with Effective State-Size
- Degrees of Freedom for Linear Attention: Distilling Softmax Attention with Optimal Feature Efficiency
- From Shortcut to Induction Head: How Data Diversity Shapes Algorithm Selection in Transformers
- Generalization Bound of Gradient Flow through Training Trajectory and Data-dependent Kernel
- Hessian-guided Perturbed Wasserstein Gradient Flows for Escaping Saddle Points
- How Does Label Noise Gradient Descent Improve Generalization in the Low SNR Regime?
- State Size Independent Statistical Error Bound for Discrete Diffusion Models
- Trained Mamba Emulates Online Gradient Descent in In-Context Linear Regression
- Learning sum of diverse features: computational hardness and efficient gradient-based training for ridge combinations
- Improved statistical and computational complexity of the mean-field Langevin dynamics under structured data
- Koopman-based generalization bound: New aspect for full-rank weights
- Minimax optimality of convolutional neural networks for infinite dimensional input-output problems and separation from kernel methods
- Optimal criterion for feature learning of two-layer linear neural network in high dimensional interpolation regime
- Symmetric Mean-field Langevin Dynamics for Distributional Minimax Problems
- Understanding Convergence and Generalization in Federated Learning through Feature Learning Theory
- High-Dimensional Kernel Methods under Covariate Shift: Data-Dependent Implicit Regularization
- How do Transformers Perform In-Context Autoregressive Learning ?
- Mean Field Langevin Actor-Critic: Faster Convergence and Global Optimality beyond Lazy Learning
- Mean-field Analysis on Two-layer Neural Networks from a Kernel Perspective
- Mechanistic Design and Scaling of Hybrid Architectures
- Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples
- SILVER: Single-loop variance reduction and application to federated learning
- State-Free Inference of State-Space Models: The *Transfer Function* Approach
- Transformers Learn Nonlinear Features In Context: Nonconvex Mean-field Dynamics on the Attention Landscape
- Neural network learns low-dimensional polynomials with SGD near the information-theoretic limit
- On the Comparison between Multi-modal and Single-modal Contrastive Learning
- Pretrained Transformer Efficiently Learns Low-Dimensional Target Functions In-Context
- Provably Transformers Harness Multi-Concept Word Semantics for Efficient In-Context Learning
- Transformers are Minimax Optimal Nonparametric In-Context Learners
- Unveil Benign Overfitting for Transformer in Vision: Training Dynamics, Convergence, and Generalization
- Excess Risk of Two-Layer ReLU Neural Networks in Teacher-Student Settings and its Superiority to Kernel Methods
- Uniform-in-time propagation of chaos for the mean-field gradient Langevin dynamics
- Approximation and Estimation Ability of Transformers for Sequence-to-Sequence Functions with Infinite Dimensional Input
- DIFF2: Differential Private Optimization via Gradient Differences for Nonconvex Distributed Learning
- Diffusion Models are Minimax Optimal Distribution Estimators
- Primal and Dual Analysis of Entropic Fictitious Play for Finite-sum Problems
- Tight and fast generalization error bound of graph embedding in metric space
- Convergence of mean-field Langevin dynamics: time-space discretization, stochastic gradient, and variance reduction
- Feature learning via mean-field Langevin dynamics: classifying sparse parities and beyond
- Gradient-Based Feature Learning under Structured Data
- Learning in the Presence of Low-dimensional Structure: A Spiked Random Matrix Perspective
- Dimension-free convergence rates for gradient Langevin dynamics in RKHS
- Learnability of convolutional neural networks for infinite dimensional input via mixed and anisotropic smoothness
- Particle Stochastic Dual Coordinate Ascent: Exponential convergent algorithm for mean field neural network optimization
- Understanding the Variance Collapse of SVGD in High Dimensions
- Escaping Saddle Points with Bias-Variance Reduced Local Perturbed SGD for Communication Efficient Nonconvex Distributed Learning
- High-dimensional Asymptotics of Feature Learning: How One Gradient Step Improves the Representation
- Improved Convergence Rate of Stochastic Gradient Langevin Dynamics with Variance Reduction and its Application to Optimization
- Two-layer neural network on infinite dimensional data: global optimization guarantee in the mean-field regime