第63回統計的機械学習セミナー / The 63rd Statistical Machine Learning Seminar
- 【日時】
- 2024年12月9日(月) 15:00 - 16:30
参加無料 / Admission Free
- 【場所】
- 統計数理研究所・D棟3階セミナー室5 (ハイブリッド)
オンライン参加を希望される場合は、以下の google form に登録し、Zoom情報をお受け取りください．
https://forms.gle/KEcBov7qHNGx5GAU6
(現地参加の場合は登録不要です)
【Speaker】
Prof. Ichiro Takeuchi (Nagoya University / RIKEN AIP )
【Title】
Statistical Testing for Hypotheses Generated by Complex Machine Learning Models Using Selective Inference
【Abstract】
In data-driven science and technology, hypotheses about research and development targets are often generated from data using complex machine learning models, enabling the discovery of promising insights that might elude human experts. However, these data-driven hypotheses are prone to be biased toward the data, and their reliability cannot be adequately evaluated using traditional statistical inference methods. Selective inference, an emerging approach for statistical inference in data-driven contexts, addresses this issue by conditioning on the hypothesis selection process, allowing for accurate quantification of statistical significance. In this talk, we present our recent findings on applying selective inference to statistical testing in deep learning and data analysis pipelines.
【Organizers】
Kenji Fukumizu, Yoshiyuki Ninomiya @ ISM
Research Center for Statistical Machine Learning/Statistical Machine Learning Collaboratory
