Academy of Mathematics and Systems Science, CAS
Colloquia & Seminars

Speaker:	Prof. Chunming Zhang，University of Wisconsin-Madison
Inviter:	陈敏研究员
Title:	MaxICA with augmented genetic algorithm and application to EEG data
Time & Venue:	2019.1.9 14:00 N613
Abstract:	In many scientific disciplines, finding hidden influential factors behind observational data is essential but challenging. The majority of existing approaches rely on linear transformation, i.e., hidden components are linear combinations of original sources. Motivated from analyzing non-linear temporal signals in finance, genetics and neuroscience, this paper proposes the "maximum independent component analysis" (MaxICA), based on max-linear combinations of original sources. In contrast to existing methods, MaxICA benefits from focusing on major information while filtering out minor information. A major tool for parameter learning of MaxICA is the proposed ECD_GA algorithm, consisting of three schemes for the elite weighted sum selection, combined crossover and dynamic mutation, and its convergence properties are discussed. Extensive empirical evaluations demonstrate the effectiveness of MaxICA in either extracting max-linearly combined essential sources in many applications or supplying a better approximation for nonlinearly combined source signals, such as EEG recordings analyzed in this paper.