Transformer-based Explainable Artificial Intelligence (XAI) for the Classification of Brain fNIRS Signals
- Title
- Transformer-based Explainable Artificial Intelligence (XAI) for the Classification of Brain fNIRS Signals
- Alternative Title
- 뇌 기능적 근적외선 분광법(fNIRS) 신호 분류를 위한 변압기 기반 설명 가능 인공지능(XAI)
- Author(s)
- Jim Knebel
- DGIST Authors
- Jim Knebel ; Jae Youn Hwang ; Kyung Hwan Jin
- Advisor
- 황재윤
- Co-Advisor(s)
- Kyung Hwan Jin
- Issued Date
- 2023
- Awarded Date
- 2023-02-01
- Type
- Thesis
- Description
- Vision Transformer, Time series, Explainable Artificial Intelligence (XAI), Layerwise Relevance Propagation, Classification, fNIRS, Brain Computer Interface,비전 트랜스포머(ViT), 시계열, 설명가능 인공지능(XAI), 계층별 관련성 전파(LRP), 분류, 기능적 근적외선 분광법(fNIRS), 뇌-컴퓨터 인터페이스(BCI)
- Table Of Contents
-
Ⅰ. Introduction 1
1.1 Motivation 1
1.2 Background 5
1.2.1 Functional Near-Infrared Spectroscopy in BCI 5
1.2.2 XAI for inspecting model output 8
1.2.3 Transformer and Vision Transformer Model Architectures 10
II. Material and Methods 12
2.1 Datasets 12
2.2 Data Preprocessing & Model Configurations 16
2.3 XAI: Layerwise Relevance Propagation 20
2.4 Locality Based Patch Embedding 24
2.5 Reconstructing & Combining Attention Maps 26
III. Results 28
3.1 Patch embedding method : Performance Comparison 28
3.2 Patch embedding method : Qualitative Evaluation 31
3.3 Patch Embedding Perturbation Experiments 33
3.4 Comparing XAI methods 39
IV. Discussion 42
V. Conclusion 45
VI. References 46
- URI
-
http://hdl.handle.net/20.500.11750/45744
http://dgist.dcollection.net/common/orgView/200000658414
- Degree
- Master
- Publisher
- DGIST
- Related Researcher
-
-
Hwang, Jae Youn
- Research Interests Multimodal Imaging; High-Frequency Ultrasound Microbeam; Ultrasound Imaging and Analysis; 스마트 헬스케어; Biomedical optical system
-
- Files in This Item:
-
There are no files associated with this item.
- Appears in Collections:
- Department of Electrical Engineering and Computer Science Theses Master
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.