革新知能統合研究センター

機械学習の主要なカンファレンスであるICLR 2024において、AIPセンターから19本の論文が採択されました。

[Accepted Papers] https://openreview.net/group?id=ICLR.cc/2024/Conference

[Spotlight Presentation]

• Analyzing Feed-Forward Blocks in Transformers through the Lens of Attention Map
  Goro Kobayashi (Tohoku University / RIKEN AIP)
  Tatsuki Kuribayashi (MBZUAI / RIKEN AIP)
  Sho Yokoi (Tohoku University / RIKEN AIP)
  Kentaro Inui (MBZUAI / Tohoku University / RIKEN AIP)
• Deep Geodesic Canonical Correlation Analysis for Covariance-Based Neuroimaging Data
  *Ce JU (NTU)
  *Reinmar KOBLER (ATR/RIKEN AIP)
  Liyao TANG (University of Sydney)
  Cuntai GUAN (NTU)
  Motoaki KAWANABE (ATR/RIKEN AIP)
  * shared first co-authorship
• Symmetric Mean-field Langevin Dynamics for Distributional Minimax Problems
  Juno Kim (The University of Tokyo / RIKEN AIP)
  Kakei Yamamoto (Massachusetts Institute of Technology)
  Kazusato Oko (The University of Tokyo / RIKEN AIP)
  Zhuoran Yang (Yale University)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks
  Hao Chen (Carnegie Mellon University)
  Jindong Wang (Microsoft Research)
  Ankit Shah (Carnegie Mellon University)
  Ran Tao (Carnegie Mellon University)
  Hongxin Wei (Southern University of Science and Technology)
  Xing Xie (Microsoft Research)
  Masashi Sugiyama (RIKEN AIP/The University of Tokyo)
  Bhiksha Raj (Carnegie Mellon University)

[Poster Presentation]

• Accurate Forgetting for Heterogeneous Federated Continual Learning
  Abudukelimu Wuerkaixi (Tsinghua University)
  Sen Cui (Tsinghua University)
  Jingfeng Zhang (University of Auckland)
  Kunda Yan (Tsinghua University)
  Bo Han (Hong Kong Baptist University)
  Gang Niu (RIKEN AIP)
  Lei Fang (Tsinghua University)
  Changshui Zhang (Tsinghua University)
  Masashi Sugiyama (RIKEN AIP/The University of Tokyo)
• Adversarial Training on Purification (AToP): Advancing Both Robustness and Generalization
  Guang Lin (Tokyo University of Agriculture and Technology / RIKEN AIP)
  Chao Li (RIKEN AIP)
  Jianhai Zhang (Hangzhou Dianzi University)
  Toshihisa Tanaka (Tokyo University of Agriculture and Technology / RIKEN AIP)
  Qibin Zhao (RIKEN AIP / Tokyo University of Agriculture and Technology)
• Anisotropy helps: improved statistical and computational complexity of the mean-field Langevin dynamics under structured data
  Atsushi Nitanda (A*STAR)
  Kazusato Oko (The University of Tokyo / RIKEN AIP)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
  Denny Wu (New York University / Flatiron Institute)
• An LLM can Fool Itself: A Prompt-Based Adversarial Attack
  Xilie Xu* (National University of Singapore)
  Keyi Kong* (Shandong University)
  Ning Liu (Shandong University )
  Lizhen Cui (Shandong University )
  Di Wang (King Abdullah University of Science and Technology)
  Jingfeng Zhang (University of Auckland / RIKEN AIP)
  Mohan Kankanhalli (National University of Singapore)
• AutoLoRa: A Parameter-Free Automated Robust Fine-Tuning Framework
  Xilie Xu (National University of Singapore)
  Jingfeng Zhang (University of Auckland / RIKEN AIP)
  Mohan Kankanhalli (National University of Singapore)
• Conformal Prediction via Regression-as-Classification,
  Etash Kumar Guha (RIKEN AIP)
  Shlok Natarajan (Salesforce)
  Thomas Möllenhoff (RIKEN AIP)
  Mohammad Emtiyaz Khan (RIKEN AIP)
  Eugene Ndiaye (Apple)
• GPAvatar: Generalizable and Precise Head Avatar from Image(s)
  Xuangeng Chu (The University of Tokyo/International Digital Economy Academy)
  Yu Li (International Digital Economy Academy)
  Ailing Zeng (International Digital Economy Academy)
  Tianyu Yang (International Digital Economy Academy)
  Lijian Lin (International Digital Economy Academy)
  Yunfei Liu (International Digital Economy Academy)
  Tatsuya Harada (The University of Tokyo/RIKEN AIP)
• Koopman-based generalization bound: New aspect for full-rank weights
  Yuka Hashimoto (NTT Corporation / RIKEN AIP)
  Sho Sonoda (RIKEN AIP)
  Isao Ishikawa (Ehime University / RIKEN AIP)
  Atsushi Nitanda (A*STAR)
  Taiji Suzuki (The University of Tokyo /RIKEN AIP)
• Minimax optimality of convolutional neural networks for infinite dimensional input-output problems and separation from kernel methods
  Yuto Nishimura (The University of Tokyo)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• Model Merging by Uncertainty-Based Gradient Matching,
  Nico Daheim (Technical University Darmstadt)
  Thomas Möllenhoff (RIKEN AIP)
  Edoardo Ponti (University of Edinburgh)
  Iryna Gurevych (Technical University Darmstadt)
  Mohammad Emtiyaz Khan (RIKEN AIP)
• Optimal criterion for feature learning of two-layer linear neural network in high dimensional interpolation regime
  Keita Suzuki (Preferred Networks, Inc.)
  Taiji Suzuki (The University of Tokyo / AIP RIKEN)
• Robust Similarity Learning with Difference Alignment Regularization
  Shuo Chen (RIKEN AIP)
  Gang Niu (RIKEN AIP)
  Chen Gong (Nanjing University of Science and Technology)
  Okan Koc (RIKEN AIP)
  Jian Yang (Nanjing University of Science and Technology)
  Masashi Sugiyama (RIKEN AIP/The University of Tokyo)
• SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer
  Yuhta Takida (Sony AI)
  Masaaki Imaizumi (The University of Tokyo/RIKEN AIP)
  Takashi Shibuya (Sony AI)
  Chieh-Hsin Lai (Sony AI)
  Toshimitsu Uesaka (Sony AI)
  Naoki Murata (Sony AI)
  Yuki Mitsufuji (Sony AI/Sony Group Corporation)
• Understanding Convergence and Generalization in Federated Learning through Feature Learning Theory
  Wei Huang (AIP-RIKEN)
  Ye Shi (ShanghaiTech University)
  Zhongyi Cai (ShanghaiTech University)
  Taiji Suzuki (The University of Tokyo / AIP RIKEN)
• When Semantic Segmentation Meets Frequency Aliasing
  Linwei Chen (Beijing Institute of Technology)
  Lin Gu (RIKEN AIP)
  Ying Fu (Beijing Institute of Technology)

関連研究室