論文表記の著者名
Masashi Sugiyama
東京大学Machine Learning2021–2025 確定版
対象論文 73 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
- Learning View-invariant World Models for Visual Robotic Manipulation
- Realistic Evaluation of Deep Partial-Label Learning Algorithms
- Sharpness-Aware Black-Box Optimization
- Towards Effective Evaluations and Comparisons for LLM Unlearning Methods
- Towards Out-of-Modal Generalization without Instance-level Modal Correspondence
- Adaptive Localization of Knowledge Negation for Continual LLM Unlearning
- Non-stationary Online Learning for Curved Losses: Improved Dynamic Regret via Mixability
- Parallel Simulation for Log-concave Sampling and Score-based Diffusion Models
- Generalized Linear Bandits: Almost Optimal Regret with One-Pass Update
- The Adaptive Complexity of Minimizing Relative Fisher Information
- Accurate Forgetting for Heterogeneous Federated Continual Learning
- Robust Similarity Learning with Difference Alignment Regularization
- Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks
- A General Framework for Learning from Weak Supervision
- Balancing Similarity and Complementarity for Federated Learning
- Counterfactual Reasoning for Multi-Label Image Classification via Patching-Based Training
- Efficient Non-stationary Online Learning by Wavelets with Applications to Online Distribution Shift Adaptation
- Generating Chain-of-Thoughts with a Pairwise-Comparison Approach to Searching for the Most Promising Intermediate Thought
- Learning with Complementary Labels Revisited: The Selected-Completely-at-Random Setting Is More Practical
- Locally Estimated Global Perturbations are Better than Local Perturbations for Federated Sharpness-aware Minimization
- Enriching Disentanglement: From Logical Definitions to Quantitative Metrics
- Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations
- Slight Corruption in Pre-training Data Makes Better Diffusion Models
- Test-time Adaptation in Non-stationary Environments via Adaptive Representation Alignment
- What Makes Partial-Label Learning Algorithms Effective?
- Is the Performance of My Deep Network Too Good to Be True? A Direct Approach to Estimating the Bayes Error in Binary Classification
- Seeing Differently, Acting Similarly: Heterogeneously Observable Imitation Learning
- A Category-theoretical Meta-analysis of Definitions of Disentanglement
- Diversity-enhancing Generative Network for Few-shot Hypothesis Adaptation
- GAT: Guided Adversarial Training with Pareto-optimal Auxiliary Tasks
- Optimality of Thompson Sampling with Noninformative Priors for Pareto Bandits
- Adapting to Continuous Covariate Shift via Online Density Ratio Estimation
- Binary Classification with Confidence Difference
- Class-Distribution-Aware Pseudo-Labeling for Semi-Supervised Multi-Label Learning
- Distributional Pareto-Optimal Multi-Objective Reinforcement Learning
- Diversified Outlier Exposure for Out-of-Distribution Detection via Informative Extrapolation
- Efficient Adversarial Contrastive Learning via Robustness-Aware Coreset Selection
- Enhancing Adversarial Contrastive Learning via Adversarial Invariant Regularization
- Generalizing Importance Weighting to A Universal Solver for Distribution Shift Problems
- Imitation Learning from Vague Feedback
- On the Overlooked Pitfalls of Weight Decay and How to Mitigate Them: A Gradient-Norm Perspective
- Online (Multinomial) Logistic Bandit: Improved Regret and Constant Computation Cost
- Exploiting Class Activation Value for Partial-Label Learning
- Federated Learning from Only Unlabeled Data with Class-conditional-sharing Clients
- Meta Discovery: Learning to Discover Novel Classes given Very Limited Data
- Rethinking Class-Prior Estimation for Positive-Unlabeled Learning
- Sample Selection with Uncertainty of Losses for Learning with Noisy Labels
- Adaptive Inertia: Disentangling the Effects of Adaptive Learning Rate and Momentum
- Adversarial Attack and Defense for Non-Parametric Two-Sample Tests
- To Smooth or Not? When Label Smoothing Meets Noisy Labels
- Adapting to Online Label Shift with Provable Guarantees
- Adversarial Training with Complementary Labels: On the Benefit of Gradually Informative Attacks
- Generalizing Consistent Multi-Class Classification with Rejection to be Compatible with Arbitrary Losses
- Learning Contrastive Embedding in Low-Dimensional Space
- Synergy-of-Experts: Collaborate to Improve Adversarial Robustness
- A Diffusion Theory For Deep Learning Dynamics: Stochastic Gradient Descent Exponentially Favors Flat Minima
- Geometry-aware Instance-reweighted Adversarial Training
- Binary Classification from Multiple Unlabeled Datasets via Surrogate Set Classification
- CIFS: Improving Adversarial Robustness of CNNs via Channel-wise Importance-based Feature Selection
- Classification with Rejection Based on Cost-sensitive Classification
- Confidence Scores Make Instance-dependent Label-noise Learning Possible
- Large-Margin Contrastive Learning with Distance Polarization Regularizer
- Learning Diverse-Structured Networks for Adversarial Robustness
- Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization
- Learning from Similarity-Confidence Data
- Lower-Bounded Proper Losses for Weakly Supervised Classification
- Maximum Mean Discrepancy Test is Aware of Adversarial Attacks
- Mediated Uncoupled Learning: Learning Functions without Direct Input-output Correspondences
- Pointwise Binary Classification with Pairwise Confidence Comparisons
- Positive-Negative Momentum: Manipulating Stochastic Gradient Noise to Improve Generalization
- Provably End-to-end Label-noise Learning without Anchor Points
- Loss function based second-order Jensen inequality and its application to particle variational inference
- Probabilistic Margins for Instance Reweighting in Adversarial Training