jin's blog

  • 홈
  • 태그
  • 미디어로그
  • 위치로그
  • 방명록

Concept vector 1

[XAI] Interpretability Beyond Feature Attribution:Quantitative Testing with Concept Activation Vectors (TCAV)

성별, 인종, 특정 패턴(줄무늬)같은 내가 생각한 컨셉이 실제 모델 예측에 있어서 얼마나 sensitivity를 주었는지 확인해볼수있는 방법으로, 검증하고자 하는 컨셉 C를 고르고 random data를 골라서 classifier를 학습시킨 후 -> 해당 classifier의 orthogonal vector를 CAV (weight vector) 라고 정의함. CAV를 이용해서 원본 image에서 컨셉벡터단위로 움직여보고 해당 특정 클래스에 대한 sensitivity값이 특정 값 이상으로 나오는지를 통해 모델 해석 설명하는 논문임 아래 내용은 https://beenkim.github.io/slides/TCAV_ICML_pdf.pdf 및 논문에서 가져와 재구성하여 만든 발표 자료로 컨텐츠 모두 beenkim..

IT/Paper 2022.01.05
이전
1
다음
더보기
프로필사진

jin's blog

Endure

  • 분류 전체보기 (50)
    • IT (44)
      • Paper (23)
      • Reinforcement Learni.. (0)
      • Probability (0)
      • Deep learning (6)
      • Spark (5)
      • Python (4)
      • Computer vision (4)
      • Data Structure (1)
    • 관심사 (2)
      • 낚시 (0)
      • 피아노 (2)
      • 일상 (0)

Tag

vision transformer, CAV, R-CNN, Fast R-CNN, RL논문, XAI, TCAV, Learning Directed Exploration Strategies, Regularizing Trajectory Optimization with Denoising Autoencoders, Quantifying Attention Flow in Transformers, Never Give Up, intergrated gradient, Adversarial Examples Are Not Bugs, Interpretability Beyond Feature Attribution:Quantitative Testing with Concept Activation Vectors, smoothGrad, Concept vector, Deconvolution Network, Axiomatic Attribution for Deep Networks, Paper리뷰, They Are Features,

최근글과 인기글

  • 최근글
  • 인기글

최근댓글

공지사항

페이스북 트위터 플러그인

  • Facebook
  • Twitter

Archives

Calendar

«   2025/11   »
일 월 화 수 목 금 토
1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30

방문자수Total

  • Today :
  • Yesterday :

Copyright © Kakao Corp. All rights reserved.

티스토리툴바