Saved in:
Bibliographic Details
Main Authors: Nicolson, Angus, Gal, Yarin, Noble, J. Alison
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.08652
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914914039234560
author Nicolson, Angus
Gal, Yarin
Noble, J. Alison
author_facet Nicolson, Angus
Gal, Yarin
Noble, J. Alison
contents Concept-based interpretability methods are a popular form of explanation for deep learning models which provide explanations in the form of high-level human interpretable concepts. These methods typically find concept activation vectors (CAVs) using a probe dataset of concept examples. This requires labelled data for these concepts -- an expensive task in the medical domain. We introduce TextCAVs: a novel method which creates CAVs using vision-language models such as CLIP, allowing for explanations to be created solely using text descriptions of the concept, as opposed to image exemplars. This reduced cost in testing concepts allows for many concepts to be tested and for users to interact with the model, testing new ideas as they are thought of, rather than a delay caused by image collection and annotation. In early experimental results, we demonstrate that TextCAVs produces reasonable explanations for a chest x-ray dataset (MIMIC-CXR) and natural images (ImageNet), and that these explanations can be used to debug deep learning-based models.
format Preprint
id arxiv_https___arxiv_org_abs_2408_08652
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle TextCAVs: Debugging vision models using text
Nicolson, Angus
Gal, Yarin
Noble, J. Alison
Machine Learning
Artificial Intelligence
Human-Computer Interaction
I.2.1; I.2.6
Concept-based interpretability methods are a popular form of explanation for deep learning models which provide explanations in the form of high-level human interpretable concepts. These methods typically find concept activation vectors (CAVs) using a probe dataset of concept examples. This requires labelled data for these concepts -- an expensive task in the medical domain. We introduce TextCAVs: a novel method which creates CAVs using vision-language models such as CLIP, allowing for explanations to be created solely using text descriptions of the concept, as opposed to image exemplars. This reduced cost in testing concepts allows for many concepts to be tested and for users to interact with the model, testing new ideas as they are thought of, rather than a delay caused by image collection and annotation. In early experimental results, we demonstrate that TextCAVs produces reasonable explanations for a chest x-ray dataset (MIMIC-CXR) and natural images (ImageNet), and that these explanations can be used to debug deep learning-based models.
title TextCAVs: Debugging vision models using text
topic Machine Learning
Artificial Intelligence
Human-Computer Interaction
I.2.1; I.2.6
url https://arxiv.org/abs/2408.08652