Research and development of nonlinear Selective Inference for high-dimensional and small number of samples data

高维小样本数据非线性选择性推理研究进展

• 批准号: 20H04243
• 负责人: 山田 誠
• 金额: $ 11.32万
• 依托单位: Okinawa Institute of Science and Technology Graduate University (2023)Kyoto University (2020-2022)
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B)
• 财政年份: 2020
• 资助国家: 日本
• 起止时间: 2020-04-01 至 2024-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-20H04243/
• 关键词: 選択的推論; カーネル法; 木構造最適輸送距離; 特徴選択; 統計的推論; 機械学習

本年度は、カーネル法に基づいた選択的推論手法を提案し, 提案手法をバイオデータに適用した. そして, 機械学習の難関国際会議であるAISTATS2022にて報告した [1]. 上記の結果に加え, 高次元データ解析手法を複数提案した. 具体的には, 木構造最適輸送に基づいたBarycenterの推定手法の提案　[2], 木構造最適輸送距離の学習方法を提案した [3]. これらの研究成果に関しても, AISTATS2022とTransactions on Machine Learning (TMLR)にそれぞれ報告した.[1] Benjamin Poignard, Peter J. Naylor, Hector Climente-Gonzlez, Makoto Yamada: Feature screening with kernel knockoffs. AISTATS 2022: 1935-1974[2]Yuki Takezawa, Ryoma Sato, Zornitsa Kozareva, Sujith Ravi, Makoto Yamada: Fixed Support Tree-Sliced Wasserstein Barycenter. AISTATS 2022: 1120-1137[3]Makoto Yamada, Yuki Takezawa, Ryoma Sato, Han Bao, Zornitsa Kozareva, Sujith Ravi:Approximating 1-Wasserstein Distance with Trees. TMLR 2022

This year, the basic method of selection is proposed, and the proposed method is applicable. Report of the International Conference on Mechanical Learning and Difficulties 2022 [1]. The results of the above note are added, and the high dimensional analysis method is proposed. Specific proposals for estimating the optimum transport distance of wood structures [2] and for learning the optimum transport distance of wood structures [3]. AISTATS2022 Transactions on Machine Learning (TMLR) is a report on the results of this research. [1] Benjamin Poignard, Peter J. Naylor, Hector Climente-Gonzlez, Makoto Yamada: Feature screening with kernel knockoffs. AISTATS 2022: 1935-1974[2]Yuki Takezawa, Ryoma Sato, Zornitsa Kozareva, Sujith Ravi, Makoto Yamada: Fixed Support Tree-Sliced Wasserstein Barycenter. AISTATS 2022: 1120-1137[3]Makoto Yamada, Yuki Takezawa, Ryoma Sato, Han Bao, Zornitsa Kozareva, Sujith Ravi:Approximating 1-Wasserstein Distance with Trees. TMLR 2022

项目成果

Dynamic Sasvi: Strong Safe Screening for Norm-Regularized Least Squares

• 发表时间: 2021-02
• 作者: Hiroaki Yamada;M. Yamada

Feature screening with kernel knockoffs

使用内核仿制品进行功能筛选

• 发表时间: 2022
• 作者: Benjamin Poignard Peter J Naylor;Hector Climente Gonzarez;Makoto Yamada.
• 通讯作者: Makoto Yamada.

Random Features Strengthen Graph Neural Networks

• DOI: 10.1137/1.9781611976700.38
• 发表时间: 2020-02
• 作者: R. Sato;M. Yamada;H. Kashima

LSMI-Sinkhorn: Semi-supervised Mutual Information Estimation with Optimal Transport

• DOI: 10.1007/978-3-030-86486-6_40
• 发表时间: 2019-09
• 作者: Yanbin Liu;M. Yamada;Yao-Hung Hubert Tsai;Tam Le;R. Salakhutdinov;Yi Yang

Supervised Tree-Wasserstein Distance

• 发表时间: 2021-01
• 期刊: ArXiv
• 作者: Yuki Takezawa;R. Sato;M. Yamada