Visualising the Training Process of Convolutional Neural Networks for Non-Experts

Michelle Peters, Lindsay Kempen, Meike Nauta, Christin Seifert

Research output: Contribution to conferencePaperAcademicpeer-review

31 Downloads (Pure)

Abstract

Convolutional neural networks are very complex and not easily interpretable by humans. Several tools give more insight into the training process and decision making of neural networks but are not un- derstandable for people with no or limited knowledge about artificial neural networks. Since these non-experts sometimes do need to rely on the decisions of a neural network, we developed an open-source tool that intuitively visualises the training process of a neural network. We visualize neuron activity using the dimensionality reduction method UMAP. By plotting neuron activity after every epoch, we create a video that shows how the neural network improves itself throughout the training phase. We evaluated our method by analysing the visualization on a CNN training on a sketch data set. We show how a video of the training over time gives more insight than a static visualisation at the end of training, as well as which features are useful to visualise for non-experts. We conclude that most of the useful deductions made from the videos are suitable for non-experts, which indicates that the visualization tool might be helpful in practice.
Original languageEnglish
Publication statusPublished - 2019
Event31th Benelux Conference on Artificial Intelligence, BNAIC 2019 - Ateliers Des Tanneurs, Brussels, Belgium
Duration: 6 Nov 20198 Nov 2019
Conference number: 31
https://bnaic19.brussels

Conference

Conference31th Benelux Conference on Artificial Intelligence, BNAIC 2019
Abbreviated titleBNAIC
CountryBelgium
CityBrussels
Period6/11/198/11/19
Internet address

Fingerprint

Neural networks
Visualization
Neurons
Decision making

Cite this

Peters, M., Kempen, L., Nauta, M., & Seifert, C. (2019). Visualising the Training Process of Convolutional Neural Networks for Non-Experts. Paper presented at 31th Benelux Conference on Artificial Intelligence, BNAIC 2019, Brussels, Belgium.
Peters, Michelle ; Kempen, Lindsay ; Nauta, Meike ; Seifert, Christin . / Visualising the Training Process of Convolutional Neural Networks for Non-Experts. Paper presented at 31th Benelux Conference on Artificial Intelligence, BNAIC 2019, Brussels, Belgium.
@conference{5f8c848661a54307bd8014c5fd0bf506,
title = "Visualising the Training Process of Convolutional Neural Networks for Non-Experts",
abstract = "Convolutional neural networks are very complex and not easily interpretable by humans. Several tools give more insight into the training process and decision making of neural networks but are not un- derstandable for people with no or limited knowledge about artificial neural networks. Since these non-experts sometimes do need to rely on the decisions of a neural network, we developed an open-source tool that intuitively visualises the training process of a neural network. We visualize neuron activity using the dimensionality reduction method UMAP. By plotting neuron activity after every epoch, we create a video that shows how the neural network improves itself throughout the training phase. We evaluated our method by analysing the visualization on a CNN training on a sketch data set. We show how a video of the training over time gives more insight than a static visualisation at the end of training, as well as which features are useful to visualise for non-experts. We conclude that most of the useful deductions made from the videos are suitable for non-experts, which indicates that the visualization tool might be helpful in practice.",
author = "Michelle Peters and Lindsay Kempen and Meike Nauta and Christin Seifert",
year = "2019",
language = "English",
note = "31th Benelux Conference on Artificial Intelligence, BNAIC 2019, BNAIC ; Conference date: 06-11-2019 Through 08-11-2019",
url = "https://bnaic19.brussels",

}

Peters, M, Kempen, L, Nauta, M & Seifert, C 2019, 'Visualising the Training Process of Convolutional Neural Networks for Non-Experts' Paper presented at 31th Benelux Conference on Artificial Intelligence, BNAIC 2019, Brussels, Belgium, 6/11/19 - 8/11/19, .

Visualising the Training Process of Convolutional Neural Networks for Non-Experts. / Peters, Michelle; Kempen, Lindsay; Nauta, Meike ; Seifert, Christin .

2019. Paper presented at 31th Benelux Conference on Artificial Intelligence, BNAIC 2019, Brussels, Belgium.

Research output: Contribution to conferencePaperAcademicpeer-review

TY - CONF

T1 - Visualising the Training Process of Convolutional Neural Networks for Non-Experts

AU - Peters, Michelle

AU - Kempen, Lindsay

AU - Nauta, Meike

AU - Seifert, Christin

PY - 2019

Y1 - 2019

N2 - Convolutional neural networks are very complex and not easily interpretable by humans. Several tools give more insight into the training process and decision making of neural networks but are not un- derstandable for people with no or limited knowledge about artificial neural networks. Since these non-experts sometimes do need to rely on the decisions of a neural network, we developed an open-source tool that intuitively visualises the training process of a neural network. We visualize neuron activity using the dimensionality reduction method UMAP. By plotting neuron activity after every epoch, we create a video that shows how the neural network improves itself throughout the training phase. We evaluated our method by analysing the visualization on a CNN training on a sketch data set. We show how a video of the training over time gives more insight than a static visualisation at the end of training, as well as which features are useful to visualise for non-experts. We conclude that most of the useful deductions made from the videos are suitable for non-experts, which indicates that the visualization tool might be helpful in practice.

AB - Convolutional neural networks are very complex and not easily interpretable by humans. Several tools give more insight into the training process and decision making of neural networks but are not un- derstandable for people with no or limited knowledge about artificial neural networks. Since these non-experts sometimes do need to rely on the decisions of a neural network, we developed an open-source tool that intuitively visualises the training process of a neural network. We visualize neuron activity using the dimensionality reduction method UMAP. By plotting neuron activity after every epoch, we create a video that shows how the neural network improves itself throughout the training phase. We evaluated our method by analysing the visualization on a CNN training on a sketch data set. We show how a video of the training over time gives more insight than a static visualisation at the end of training, as well as which features are useful to visualise for non-experts. We conclude that most of the useful deductions made from the videos are suitable for non-experts, which indicates that the visualization tool might be helpful in practice.

UR - http://ceur-ws.org/Vol-2491/paper108.pdf

M3 - Paper

ER -

Peters M, Kempen L, Nauta M, Seifert C. Visualising the Training Process of Convolutional Neural Networks for Non-Experts. 2019. Paper presented at 31th Benelux Conference on Artificial Intelligence, BNAIC 2019, Brussels, Belgium.