[2Aa01-14] データ駆動表面科学研究部会「データ駆動アプローチ研究の最近の動向」

[2Aa10] Data-driven diagnosis for compressed sensing

*Yoshinori Nakanishi-Ohno¹ (1. Faculty of Culture and Information Science, Doshisha University)

We discuss the reliability of the results of compressed sensing. Compressed sensing enables us to decrease the required amount of data with a method from informatics such as sparse modeling. However, there is a limit for compressed sensing and how much data should be acquired is not known in advance. We argue a statistical method based on cross validation to explore the limit and to diagnose the success or failure of compressed sensing.