TU Darmstadt / ULB / TUbiblio

Rule Extraction From Binary Neural Networks With Convolutional Rules for Model Validation

Burkhardt, Sophie ; Brugger, Jannis ; Wagner, Nicolas ; Ahmadi, Zahra ; Kersting, Kristian ; Kramer, Stefan (2022):
Rule Extraction From Binary Neural Networks With Convolutional Rules for Model Validation. (Publisher's Version)
In: Frontiers in Artificial Intelligence, 4, Frontiers Media S.A., e-ISSN 2624-8212,
DOI: 10.26083/tuprints-00020098,
[Article]

Abstract

Classification approaches that allow to extract logical rules such as decision trees are often considered to be more interpretable than neural networks. Also, logical rules are comparatively easy to verify with any possible input. This is an important part in systems that aim to ensure correct operation of a given model. However, for high-dimensional input data such as images, the individual symbols, i.e. pixels, are not easily interpretable. Therefore, rule-based approaches are not typically used for this kind of high-dimensional data. We introduce the concept of first-order convolutional rules, which are logical rules that can be extracted using a convolutional neural network (CNN), and whose complexity depends on the size of the convolutional filter and not on the dimensionality of the input. Our approach is based on rule extraction from binary neural networks with stochastic local search. We show how to extract rules that are not necessarily short, but characteristic of the input, and easy to visualize. Our experiments show that the proposed approach is able to model the functionality of the neural network while at the same time producing interpretable logical rules. Thus, we demonstrate the potential of rule-based approaches for images which allows to combine advantages of neural networks and rule learning.

Item Type: Article
Erschienen: 2022
Creators: Burkhardt, Sophie ; Brugger, Jannis ; Wagner, Nicolas ; Ahmadi, Zahra ; Kersting, Kristian ; Kramer, Stefan
Origin: Secondary publication DeepGreen
Status: Publisher's Version
Title: Rule Extraction From Binary Neural Networks With Convolutional Rules for Model Validation
Language: English
Abstract:

Classification approaches that allow to extract logical rules such as decision trees are often considered to be more interpretable than neural networks. Also, logical rules are comparatively easy to verify with any possible input. This is an important part in systems that aim to ensure correct operation of a given model. However, for high-dimensional input data such as images, the individual symbols, i.e. pixels, are not easily interpretable. Therefore, rule-based approaches are not typically used for this kind of high-dimensional data. We introduce the concept of first-order convolutional rules, which are logical rules that can be extracted using a convolutional neural network (CNN), and whose complexity depends on the size of the convolutional filter and not on the dimensionality of the input. Our approach is based on rule extraction from binary neural networks with stochastic local search. We show how to extract rules that are not necessarily short, but characteristic of the input, and easy to visualize. Our experiments show that the proposed approach is able to model the functionality of the neural network while at the same time producing interpretable logical rules. Thus, we demonstrate the potential of rule-based approaches for images which allows to combine advantages of neural networks and rule learning.

Journal or Publication Title: Frontiers in Artificial Intelligence
Volume of the journal: 4
Publisher: Frontiers Media S.A.
Collation: 14 Seiten
Uncontrolled Keywords: k-term DNF, stochastic local search, convolutional neural networks, logical rules, rule extraction, interpretability
Divisions: 20 Department of Computer Science
20 Department of Computer Science > Artificial Intelligence and Machine Learning
Forschungsfelder
Forschungsfelder > Information and Intelligence
Forschungsfelder > Information and Intelligence > Cognitive Science
Zentrale Einrichtungen
Zentrale Einrichtungen > hessian.AI - The Hessian Center for Artificial Intelligence
Date Deposited: 13 May 2022 13:48
DOI: 10.26083/tuprints-00020098
URL / URN: https://tuprints.ulb.tu-darmstadt.de/20098
URN: urn:nbn:de:tuda-tuprints-200985
PPN:
Corresponding Links:
Export:
Suche nach Titel in: TUfind oder in Google
Send an inquiry Send an inquiry

Options (only for editors)
Show editorial Details Show editorial Details