Concept Bottleneck Models Without Predefined Concepts

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Schrodi, Simon, Schur, Julian, Argus, Max, Brox, Thomas
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914859386404864
author Schrodi, Simon
Schur, Julian
Argus, Max
Brox, Thomas
author_facet Schrodi, Simon
Schur, Julian
Argus, Max
Brox, Thomas
contents There has been considerable recent interest in interpretable concept-based models such as Concept Bottleneck Models (CBMs), which first predict human-interpretable concepts and then map them to output classes. To reduce reliance on human-annotated concepts, recent works have converted pretrained black-box models into interpretable CBMs post-hoc. However, these approaches predefine a set of concepts, assuming which concepts a black-box model encodes in its representations. In this work, we eliminate this assumption by leveraging unsupervised concept discovery to automatically extract concepts without human annotations or a predefined set of concepts. We further introduce an input-dependent concept selection mechanism that ensures only a small subset of concepts is used across all classes. We show that our approach improves downstream performance and narrows the performance gap to black-box models, while using significantly fewer concepts in the classification. Finally, we demonstrate how large vision-language models can intervene on the final model weights to correct model errors.
format Preprint
id arxiv_https___arxiv_org_abs_2407_03921
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Concept Bottleneck Models Without Predefined Concepts
Schrodi, Simon
Schur, Julian
Argus, Max
Brox, Thomas
Machine Learning
Computer Vision and Pattern Recognition
There has been considerable recent interest in interpretable concept-based models such as Concept Bottleneck Models (CBMs), which first predict human-interpretable concepts and then map them to output classes. To reduce reliance on human-annotated concepts, recent works have converted pretrained black-box models into interpretable CBMs post-hoc. However, these approaches predefine a set of concepts, assuming which concepts a black-box model encodes in its representations. In this work, we eliminate this assumption by leveraging unsupervised concept discovery to automatically extract concepts without human annotations or a predefined set of concepts. We further introduce an input-dependent concept selection mechanism that ensures only a small subset of concepts is used across all classes. We show that our approach improves downstream performance and narrows the performance gap to black-box models, while using significantly fewer concepts in the classification. Finally, we demonstrate how large vision-language models can intervene on the final model weights to correct model errors.
title Concept Bottleneck Models Without Predefined Concepts
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2407.03921