日時：2016年8月23日(火) 14:00-16:00
場所：九州大学　伊都キャンパス　ウエスト2号館　3階313号室
主催：新学術領域研究「多面的アプローチの統合による計算限界の解明」


講師：Elad Hazan 氏 (Princeton University)


Title：How to classify from sparse data

Abstract:
For many data sets that are sparse and structured, data reconstruction is  computationally hard.
A notable example is  the case for media-recommendation data, which is sparse and low-rank.
We consider general model for classification and regression tasks where we have missing data 
and assume that the (clean) data resides in a low dimensional subspace.
We give a non-reconstructive learning approach for this setting with provable guarantees.

based on joint work with Tomer Koren, Roi Livni and Yishay Mansour (COLT 2016)

ホスト：瀧本英二，畑埜晃平