Node co-activations as a means of error detection—Towards fault-tolerant neural networks

Show full item record



Permalink

http://hdl.handle.net/10138/345966

Citation

Myllyaho , L , Nurminen , J K & Mikkonen , T 2022 , ' Node co-activations as a means of error detection—Towards fault-tolerant neural networks ' , Array , vol. 15 . https://doi.org/10.1016/j.array.2022.100201

Title: Node co-activations as a means of error detection—Towards fault-tolerant neural networks
Author: Myllyaho, Lalli; Nurminen, Jukka K; Mikkonen, Tommi
Contributor organization: Department of Computer Science
Empirical Software Engineering research group
Date: 2022-09
Language: eng
Belongs to series: Array
ISSN: 2590-0056
DOI: https://doi.org/10.1016/j.array.2022.100201
URI: http://hdl.handle.net/10138/345966
Abstract: Context: Machine learning has proved an efficient tool, but the systems need tools to mitigate risks during runtime. One approach is fault tolerance: detecting and handling errors before they cause harm. Objective: This paper investigates whether rare co-activations – pairs of usually segregated nodes activating together – are indicative of problems in neural networks (NN). These could be used to detect concept drift and flagging untrustworthy predictions. Methods: We trained four NNs. For each, we studied how often each pair of nodes activates together. In a separate test set, we counted how many rare co-activations occurred with each input, and grouped the inputs based on whether its classification was correct, incorrect, or whether its class was absent during training. Results: Rare co-activations are much more common in inputs from a class that was absent during training. Incorrectly classified inputs averaged a larger number of rare co-activations than correctly classified inputs, but the difference was smaller. Conclusions: As rare co-activations are more common in unprecedented inputs, they show potential for detecting concept drift. There is also some potential in detecting single inputs from untrained classes. The small difference between correctly and incorrectly predicted inputs is less promising and needs further research.
Subject: 113 Computer and information sciences
Peer reviewed: Yes
Rights: cc_by
Usage restriction: openAccess
Self-archived version: publishedVersion


Files in this item

Total number of downloads: Loading...

Files Size Format View
1_s2.0_S2590005622000509_main.pdf 962.0Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record