Knowledge Extraction from Auto-Encoders on Anomaly Detection Tasks Using Co-activation Graphs

Daniyal Selani, Ilaria Tiddi

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

Deep neural networks have exploded in popularity and different types of networks are used to solve a multitude of complex tasks. One such task is anomaly detection, that a type of deep neural network called auto-encoder has become extremely proficient at solving. The low level neural activity, produced by such a network, generates extremely rich representations of the data, which can be used to extract task specific knowledge. In this paper, we built upon previous work and used co-activation graph analysis to extract knowledge from auto-encoders, that were trained for the specific task of anomaly detection. First, we outlined a method for extracting co-activation graphs from auto-encoders. Then, we performed graph analysis to discover that task specific knowledge from the auto-encoder was being encoded into the co-activation graph, and that the extracted knowledge could be used to reveal the role of individual neurons in the network.

Original languageEnglish
Title of host publicationK-CAP 2021
Subtitle of host publicationProceedings of the 11th Knowledge Capture Conference
PublisherAssociation for Computing Machinery, Inc
Pages65-71
Number of pages7
ISBN (Electronic)9781450384575
DOIs
Publication statusPublished - Dec 2021
Event11th ACM International Conference on Knowledge Capture, K-CAP 2021 - Virtual, Online, United States
Duration: 2 Dec 20213 Dec 2021

Conference

Conference11th ACM International Conference on Knowledge Capture, K-CAP 2021
Country/TerritoryUnited States
CityVirtual, Online
Period2/12/213/12/21

Bibliographical note

Publisher Copyright:
© 2021 ACM.

Keywords

  • anomaly detection
  • auto-encoders
  • co-activation graphs

Fingerprint

Dive into the research topics of 'Knowledge Extraction from Auto-Encoders on Anomaly Detection Tasks Using Co-activation Graphs'. Together they form a unique fingerprint.

Cite this