Statistically Significant Concept-based Explanation of Image Classifiers via Model Knockoffs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Kaiwen, Fukuchi, Kazuto, Akimoto, Youhei, Sakuma, Jun
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917571135012864
author Xu, Kaiwen
Fukuchi, Kazuto
Akimoto, Youhei
Sakuma, Jun
author_facet Xu, Kaiwen
Fukuchi, Kazuto
Akimoto, Youhei
Sakuma, Jun
contents A concept-based classifier can explain the decision process of a deep learning model by human-understandable concepts in image classification problems. However, sometimes concept-based explanations may cause false positives, which misregards unrelated concepts as important for the prediction task. Our goal is to find the statistically significant concept for classification to prevent misinterpretation. In this study, we propose a method using a deep learning model to learn the image concept and then using the Knockoff samples to select the important concepts for prediction by controlling the False Discovery Rate (FDR) under a certain value. We evaluate the proposed method in our synthetic and real data experiments. Also, it shows that our method can control the FDR properly while selecting highly interpretable concepts to improve the trustworthiness of the model.
format Preprint
id arxiv_https___arxiv_org_abs_2305_18362
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Statistically Significant Concept-based Explanation of Image Classifiers via Model Knockoffs
Xu, Kaiwen
Fukuchi, Kazuto
Akimoto, Youhei
Sakuma, Jun
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
A concept-based classifier can explain the decision process of a deep learning model by human-understandable concepts in image classification problems. However, sometimes concept-based explanations may cause false positives, which misregards unrelated concepts as important for the prediction task. Our goal is to find the statistically significant concept for classification to prevent misinterpretation. In this study, we propose a method using a deep learning model to learn the image concept and then using the Knockoff samples to select the important concepts for prediction by controlling the False Discovery Rate (FDR) under a certain value. We evaluate the proposed method in our synthetic and real data experiments. Also, it shows that our method can control the FDR properly while selecting highly interpretable concepts to improve the trustworthiness of the model.
title Statistically Significant Concept-based Explanation of Image Classifiers via Model Knockoffs
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2305.18362