Event-related Potential analysis using elastic net logistic regression
엘라스틱 넷 로지스틱 회귀분석을 통한 사건 관련 전위 분석

DC Field Value Language
dc.description학위논문 (석사)-- 서울대학교 대학원 : 협동과정 인지과학전공, 2016. 2. 김청택.-
dc.description.abstractThe objective of the thesis is to explore whether regularization techniques can be
applied to ERP analysis, and which type of regularization is adequate. This thesis
proposes elastic net regularization logistic regression as a good candidate of
data analytic method for Event-Related Potential analysis (ERP). Specifically,
regularization techniques are used to identify latency in ERP. Study 1 tested
whether regularization logistic regression can classify latency using simulated
ERP data. It showed that ridge and lasso could identify latency information. In
study 2, the same analyses were applied to actual ERP data. Ridge regression
can identify latency information wheras lasso cannot.
dc.description.tableofcontentsChapter 1 Introduction 1
1.1 Regression 3
1.1.1 Linear regression 3
1.1.2 Logistic regression 4
1.1.3 Regularization methods 4

Chapter 2 Study 1 7
2.1 Simulation 7
2.2 Results and discussion 8

Chapter 3 Study 2 11
3.1 EEG acquisition 11
3.2 Experiment design 11
3.2.1 Stimuli 13
3.2.2 Participants 14
3.2.3 Procedure 14
3.3 Results 14
3.3.1 Preprocessing 14
3.3.2 ERP analysis 15
3.3.3 Logistic regression 20
3.3.4 Lasso logistic regression 20
3.3.5 Ridge logistic regression 22
3.3.6 Discussion 31

Chapter 4 Conculsion 33
4.1 Conclusion and Limitations 33
4.2 Further research 34

References 35

초록 37
dc.format.extent6569005 bytes-
dc.publisher서울대학교 대학원-
dc.subjectElastic net-
dc.subjectEvent-related potential-
dc.subjectLogistic regression-
dc.titleEvent-related Potential analysis using elastic net logistic regression-
dc.title.alternative엘라스틱 넷 로지스틱 회귀분석을 통한 사건 관련 전위 분석-
dc.contributor.affiliation인문대학 협동과정 인지과학전공-
Appears in Collections:
College of Humanities (인문대학)Program in Cognitive Science (협동과정-인지과학전공)Theses (Master's Degree_협동과정-인지과학전공)
Files in This Item:
  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.