[日時] 2024年7月26日(金) 13:00--14:00
[場所] 東京大学 本郷キャンパス 工14号館 534室
[講演者] 坂上晋作 (東京大学)
[題目(Title)] Online Structured Prediction with Fenchel–Young Losses and Improved Surrogate Regret for Online Multiclass Classification with Logistic Loss
[概要(Abstract)]
This talk is about online structured prediction with full-information feedback. For online multiclass classification, van der Hoeven (NeurIPS 2020) established surrogate regret bounds independent of the time horizon, or "finite," by introducing an elegant "exploit-the-surrogate-gap" framework. However, this framework has been limited to multiclass classification primarily because it relies on a classification-specific procedure for converting estimated scores to outputs. We extend the exploit-the-surrogate-gap framework to online structured prediction with "Fenchel–Young losses," a large family of surrogate losses that includes the logistic loss for multiclass classification as a special case, obtaining finite surrogate regret bounds in various structured prediction problems. To this end, we propose and analyze "randomized decoding," which converts estimated scores to general structured outputs. Moreover, by applying our decoding to online multiclass classification with the logistic loss, we obtain a surrogate regret bound of O(||U||_F^2), where U is the best offline linear estimator and || ||_F denotes the Frobenius norm. This bound is tight up to logarithmic factors and improves the previous bound of O(d ||U||_F^2) due to van der Hoeven (NeurIPS 2020) by a factor of d, the number of classes. This is a joint work with Han Bao, Taira Tsuchiya, and Taihei Oki, which will appear in COLT 2024.