Masashi Sugiyama 研究室

主宰者：Masashi Sugiyama

理化学研究所・RIKEN Center for Advanced Intelligence Project

AI 要約（直近 5 年の研究成果）

本研究室の中心的な関心は、実データに付随する不完全性や曖昧性を扱う機械学習の理論と方法論にあります。具体的には、含まれる誤ったラベル、複数の候補ラベル、部分的な注釈といった現実的な課題に直面するとき、機械学習モデルがどのように堅牢に学習できるかを追求しています。これらの課題は特に大規模データセットを活用する際に重要です。不正確な教師信号から本質的な知識を引き出すための統計的手法やロバスト性の強化に関する理論的知見と実践的な手法を開発しています。また、多様な学習シナリオに対する効率的なアプローチを探索しています。限定的なラベル情報と大量の未ラベル手法データが存在する半教師学習、タスク関連の特性を活用した少量データでの学習、分布のズレに対応した知識転移など、現実的な制約下での学習戦略を扱っています。分子性質予測といった応用領域では、グラフ構造を持つデータの複雑な関係性を効果的に捉えるためのモデル設計にも取り組んでいます。さらに、オンライン学習や意思決定問題への応用も進めており、バンディット問題における探索と利用のバランスや、強化学習における方針学習といった基本的な問題に対して、ノイズや不確実性の影響を考慮した理論的枠組みを構築しています。

※ AI（Claude）が、公開されている論文要旨から研究の問い・手法・主要な発見を事実情報として抽出・再構成して自動生成しています。誤りを含む可能性があるため、正確性は研究室公式情報でご確認ください。

外部リンク

研究成果（66 件）

[2026] Early Stopping Without Validation Data in Weakly Supervised Learning
DOI: https://doi.org/10.1109/tpami.2026.3712551
[2026] Instance-dependent Early Stopping for Adaptive Data Pruning
DOI: https://doi.org/10.1109/tpami.2026.3693108
[2026] Atom-Motif Contrastive Transformer for Molecular Property Prediction
DOI: https://doi.org/10.1145/3787204
[2025] Safety at Scale: A Comprehensive Survey of Large Model Safety
DOI: https://doi.org/10.70777/si.v2i2.14609
[2025] Action-Agnostic Point-Level Supervision for Temporal Action Detection
DOI: https://doi.org/10.1609/aaai.v39i9.33037
[2025] EXPERIMENTAL STUDY ON CARBONATION REDUCTION EFFECT AND LENGTH CHANGE REDUCTION EFFECT OF HOUSING BASE CONCRETE USING CURING SHEETS
DOI: https://doi.org/10.3130/aijt.31.86
[2025] Impact of Noisy Supervision in Foundation Model Learning
DOI: https://doi.org/10.1109/tpami.2025.3552309
[2025] Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety
DOI: https://doi.org/10.1561/3300000051
[2025] Label Distribution Learning with Biased Annotations Assisted by Multi-Label Learning
DOI: https://doi.org/10.24963/ijcai.2025/617
[2024] The Choice of Noninformative Priors for Thompson Sampling in Multiparameter Bandit Models
DOI: https://doi.org/10.1609/aaai.v38i12.29240

続きを表示（残り 56 件）

[2024] Overcoming Continuous Distribution Shifts: Challenges in Online Machine Learning
DOI: https://doi.org/10.1109/bci60775.2024.10480518
[2024] BadLabel: A Robust Perspective on Evaluating and Enhancing Label-Noise Learning
DOI: https://doi.org/10.1109/tpami.2024.3355425
[2024] Appearance-Based Curriculum for Semi-Supervised Learning with Multi-Angle Unlabeled Data
DOI: https://doi.org/10.1109/wacv57701.2024.00276
[2024] Slight Corruption in Pre-training Data Makes Better Diffusion Models
DOI: https://doi.org/10.52202/079017-4008
[2024] Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations
DOI: https://doi.org/10.52202/079017-1905
[2024] Test-time Adaptation in Non-stationary Environments via Adaptive Representation Alignment
DOI: https://doi.org/10.52202/079017-2999
[2024] Enriching Disentanglement: From Logical Definitions to Quantitative Metrics
DOI: https://doi.org/10.52202/079017-2297
[2024] What Makes Partial-Label Learning Algorithms Effective?
DOI: https://doi.org/10.52202/079017-2841
[2024] A Fast Algorithm for the Real-Valued Combinatorial Pure Exploration of the Multi-Armed Bandit
DOI: https://doi.org/10.1162/neco_a_01728
[2024] Acetylxylan esterase is the key to the host specialization of wood-decay fungi predicted by random forest machine-learning algorithm
DOI: https://doi.org/10.1186/s10086-024-02159-9
[2024] Estimating Per-Class Statistics for Label Noise Learning
DOI: https://doi.org/10.1109/tpami.2024.3466182
[2024] Learning explainable task-relevant state representation for model-free deep reinforcement learning
DOI: https://doi.org/10.1016/j.neunet.2024.106741
[2024] EXPERIMENTAL STUDY ON CARBONATION REDUCTION EFFECT OF HOUSING BASE CONCRETE BY ELASTIC PAINT
DOI: https://doi.org/10.3130/aijt.30.599
[2024] Random forest machine-learning algorithm classifies white- and brown-rot fungi according to the number of the genes encoding Carbohydrate-Active enZyme families
DOI: https://doi.org/10.1128/aem.00482-24
[2024] An offline learning of behavior correction policy for vision-based robotic manipulation
DOI: https://doi.org/10.1109/icra57147.2024.10610177
[2024] Thompson Sampling for Real-Valued Combinatorial Pure Exploration of Multi-Armed Bandit
DOI: https://doi.org/10.1609/aaai.v38i13.29355
[2023] Audio Signal Enhancement with Learning from Positive and Unlabeled Data
DOI: https://doi.org/10.1109/icassp49357.2023.10095988
[2023] Distribution Shift Matters for Knowledge Distillation with Webly Collected Images
DOI: https://doi.org/10.1109/iccv51070.2023.01602
[2023] Multi-Label Knowledge Distillation
DOI: https://doi.org/10.1109/iccv51070.2023.01584
[2023] Boundary-restricted metric learning
DOI: https://doi.org/10.1007/s10994-023-06380-3
[2023] Learning Intention-Aware Policies in Deep Reinforcement Learning
DOI: https://doi.org/10.1162/neco_a_01607
[2023] On the Effectiveness of Adversarial Training Against Backdoor Attacks
DOI: https://doi.org/10.1109/tnnls.2023.3281872
[2023] On the Robustness of Average Losses for Partial-Label Learning
DOI: https://doi.org/10.1109/tpami.2023.3275249
[2022] Learning and Mining with Noisy Labels
DOI: https://doi.org/10.1145/3511808.3557504
[2022] Representation learning for continuous action spaces is beneficial for efficient policy learning
DOI: https://doi.org/10.1016/j.neunet.2022.12.009
[2022] Learning With Proper Partial Labels
DOI: https://doi.org/10.1162/neco_a_01554
[2022] Towards Adversarially Robust Deep Image Denoising
DOI: https://doi.org/10.24963/ijcai.2022/211
[2022] Positive-unlabeled classification under class-prior shift: a prior-invariant approach based on density ratio estimation
DOI: https://doi.org/10.1007/s10994-022-06190-z
[2022] Instance-Dependent Label-Noise Learning with Manifold-Regularized Transition Matrix Estimation
DOI: https://doi.org/10.1109/cvpr52688.2022.01613
[2022] Spatiotemporal dynamics of odor representations in the human brain revealed by EEG decoding
DOI: https://doi.org/10.1073/pnas.2114966119
[2022] Discovering diverse solutions in deep reinforcement learning by maximizing state–action-based mutual information
DOI: https://doi.org/10.1016/j.neunet.2022.04.009
[2022] Improving imbalanced classification using near-miss instances
DOI: https://doi.org/10.1016/j.eswa.2022.117130
[2022] Root cause estimation of faults in production processes: a novel approach inspired by approximate Bayesian computation
DOI: https://doi.org/10.1080/00207543.2022.2042611
[2022] Learning from Noisy Complementary Labels with Robust Loss Functions
DOI: https://doi.org/10.1587/transinf.2021edp7035
[2022] Class-Wise Denoising for Robust Learning under Label Noise
DOI: https://doi.org/10.1109/tpami.2022.3178690
[2021] Probabilistic Margins for Instance Reweighting in Adversarial Training
[2021] Loss function based second-order Jensen inequality and its application to particle variational inference
[2021] Classification with Rejection Based on Cost-sensitive Classification
[2021] Large-Margin Contrastive Learning with Distance Polarization Regularizer
[2021] Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization
[2021] Confidence Scores Make Instance-dependent Label-noise Learning Possible
[2021] Positive-Negative Momentum: Manipulating Stochastic Gradient Noise to Improve Generalization
[2021] CIFS: Improving Adversarial Robustness of CNNs via Channel-wise Importance-based Feature Selection
[2021] Binary Classification from Multiple Unlabeled Datasets via Surrogate Set Classification
[2021] Lower-Bounded Proper Losses for Weakly Supervised Classification
[2021] Artificial Neural Variability for Deep Learning: On Overfitting, Noise Memorization, and Catastrophic Forgetting
DOI: https://doi.org/10.1162/neco_a_01403
[2021] Direction Matters: On Influence-Preserving Graph Summarization and Max-Cut Principle for Directed Graphs
DOI: https://doi.org/10.1162/neco_a_01402
[2021] A One-Step Approach to Covariate Shift Adaptation
DOI: https://doi.org/10.1007/s42979-021-00678-6
[2021] On Focal Loss for Class-Posterior Probability Estimation: A Theoretical Perspective
DOI: https://doi.org/10.1109/cvpr46437.2021.00516
[2021] On the Role of Pre-training for Meta Few-Shot Learning
[2021] A Diffusion Theory For Deep Learning Dynamics: Stochastic Gradient Descent Exponentially Favors Flat Minima
[2021] Fenchel-Young Losses with Skewed Entropies for Class-posterior Probability Estimation
[2021] Incorporating Causal Graphical Prior Knowledge into Predictive Modeling via Simple Data Augmentation
[2021] Classification From Pairwise Similarities/Dissimilarities and Unlabeled Data via Empirical Risk Minimization
DOI: https://doi.org/10.1162/neco_a_01373
[2021] Constraint learning for control tasks with limited duration barrier functions
DOI: https://doi.org/10.1016/j.automatica.2021.109504
[2021] LocalDrop: A Hybrid Regularization for Deep Neural Networks
DOI: https://doi.org/10.1109/tpami.2021.3061463

科研費（0 件）

まだデータがありません（KAKEN 取り込み後に表示）。

所属学会・役職（0 件）

まだデータがありません（学会データ連携後に表示）。

AI 要約（直近 5 年の研究成果）

外部リンク

関連研究室(8 件)

研究成果（66 件）

科研費（0 件）

所属学会・役職（0 件）