Visualising basins of attraction for the cross-entropy and the squared error neural network loss functions

Loading...
Thumbnail Image
File version
Accepted Manuscript (AM)
Author(s)
Bosman, A
Engelbrecht, AP
Helbig, M
Griffith University Author(s)
Primary Supervisor
Other Supervisors
Editor(s)
Date
2020
Size
File type(s)
Location
License
https://creativecommons.org/licenses/by-nc-nd/4.0/
Abstract

Quantification of the stationary points and the associated basins of attraction of neural network loss surfaces is an important step towards a better understanding of neural network loss surfaces at large. This work proposes a novel method to visualise basins of attraction together with the associated stationary points via gradient-based stochastic sampling. The proposed technique is used to perform an empirical study of the loss surfaces generated by two different error metrics: quadratic loss and entropic loss. The empirical observations confirm the theoretical hypothesis regarding the nature of neural network attraction basins. Entropic loss is shown to exhibit stronger gradients and fewer stationary points than quadratic loss, indicating that entropic loss has a more searchable landscape. Quadratic loss is shown to be more resilient to overfitting than entropic loss. Both losses are shown to exhibit local minima, but the number of local minima is shown to decrease with an increase in dimensionality. Thus, the proposed visualisation technique successfully captures the local minima properties exhibited by the neural network loss surfaces, and can be used for the purpose of fitness landscape analysis of neural networks.

Journal Title
Neurocomputing
Conference Title
Book Title
Edition
Volume
400
Issue
Thesis Type
Degree Program
School
Publisher link
Patent number
Funder(s)
Grant identifier(s)
Rights Statement
Rights Statement
© 2020 Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International Licence (http://creativecommons.org/licenses/by-nc-nd/4.0/) which permits unrestricted, non-commercial use, distribution and reproduction in any medium, providing that the work is properly cited.
Item Access Status
Note
Access the data
Related item(s)
Subject
Neural networks
Artificial intelligence
Science & Technology
Computer Science, Artificial Intelligence
Fitness landscape analysis
Persistent link to this record
Citation
Bosman, A; Engelbrecht, AP; Helbig, M, Visualising basins of attraction for the cross-entropy and the squared error neural network loss functions, Neurocomputing, 2020, 400, pp. 113-136
Collections