Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks
Understanding how deep hierarchical models build their knowledge is a key issue in the usage of artificial intelligence to interpret the reality behind data. Depending on the discipline and models used, such knowledge may be represented in ways that are more or less intelligible for humans, limiting...
Published in: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
---|---|
Main Authors: | , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
IEEE
2022
|
Subjects: | |
Online Access: | https://doi.org/10.1109/JSTARS.2022.3155967 https://doaj.org/article/9a0b5112bc5c4c49998d976d76a30003 |
id |
ftdoajarticles:oai:doaj.org/article:9a0b5112bc5c4c49998d976d76a30003 |
---|---|
record_format |
openpolar |
spelling |
ftdoajarticles:oai:doaj.org/article:9a0b5112bc5c4c49998d976d76a30003 2023-05-15T13:58:57+02:00 Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez 2022-01-01T00:00:00Z https://doi.org/10.1109/JSTARS.2022.3155967 https://doaj.org/article/9a0b5112bc5c4c49998d976d76a30003 EN eng IEEE https://ieeexplore.ieee.org/document/9726918/ https://doaj.org/toc/2151-1535 2151-1535 doi:10.1109/JSTARS.2022.3155967 https://doaj.org/article/9a0b5112bc5c4c49998d976d76a30003 IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol 15, Pp 2311-2325 (2022) Knowledge based systems learning (artificial intelligence) supervised learning machine learning deep learning representation learning Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809 article 2022 ftdoajarticles https://doi.org/10.1109/JSTARS.2022.3155967 2022-12-31T03:51:04Z Understanding how deep hierarchical models build their knowledge is a key issue in the usage of artificial intelligence to interpret the reality behind data. Depending on the discipline and models used, such knowledge may be represented in ways that are more or less intelligible for humans, limiting further improvements on the performance of the existing models. In order to delve into the characterization and modeling of volcano-seismic signals, this article emphasizes the idea of deciphering what and how recurrent neural networks (RNNs) model, and how this knowledge can be used to improve data interpretation. The key to accomplishing these objectives is both analyzing the hidden state dynamics associated with their hidden units as well as pruning/trimming based on the specialization of neurons. In this article, we process, analyze, and visualize the hidden states activation maps of two RNN architectures when managing different types of volcano-seismic events. As a result, the class-dependent discriminative behavior of most active neurons is analyzed, thereby increasing the comprehension of the detection and classification tasks. A representative dataset from the deception island volcano (Antarctica), containing volcano-tectonic earthquakes, long period events, volcanic tremors, and hybrid events, is used to train the models. Experimental analysis shows how neural activity and its associated specialization skills change depending on the architecture chosen and the type of event analyzed. Article in Journal/Newspaper Antarc* Antarctica Deception Island Directory of Open Access Journals: DOAJ Articles Deception Island ENVELOPE(-60.633,-60.633,-62.950,-62.950) IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 15 2311 2325 |
institution |
Open Polar |
collection |
Directory of Open Access Journals: DOAJ Articles |
op_collection_id |
ftdoajarticles |
language |
English |
topic |
Knowledge based systems learning (artificial intelligence) supervised learning machine learning deep learning representation learning Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809 |
spellingShingle |
Knowledge based systems learning (artificial intelligence) supervised learning machine learning deep learning representation learning Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809 Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks |
topic_facet |
Knowledge based systems learning (artificial intelligence) supervised learning machine learning deep learning representation learning Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809 |
description |
Understanding how deep hierarchical models build their knowledge is a key issue in the usage of artificial intelligence to interpret the reality behind data. Depending on the discipline and models used, such knowledge may be represented in ways that are more or less intelligible for humans, limiting further improvements on the performance of the existing models. In order to delve into the characterization and modeling of volcano-seismic signals, this article emphasizes the idea of deciphering what and how recurrent neural networks (RNNs) model, and how this knowledge can be used to improve data interpretation. The key to accomplishing these objectives is both analyzing the hidden state dynamics associated with their hidden units as well as pruning/trimming based on the specialization of neurons. In this article, we process, analyze, and visualize the hidden states activation maps of two RNN architectures when managing different types of volcano-seismic events. As a result, the class-dependent discriminative behavior of most active neurons is analyzed, thereby increasing the comprehension of the detection and classification tasks. A representative dataset from the deception island volcano (Antarctica), containing volcano-tectonic earthquakes, long period events, volcanic tremors, and hybrid events, is used to train the models. Experimental analysis shows how neural activity and its associated specialization skills change depending on the architecture chosen and the type of event analyzed. |
format |
Article in Journal/Newspaper |
author |
Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez |
author_facet |
Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez |
author_sort |
Manuel Titos |
title |
Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks |
title_short |
Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks |
title_full |
Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks |
title_fullStr |
Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks |
title_full_unstemmed |
Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks |
title_sort |
toward knowledge extraction in classification of volcano-seismic events: visualizing hidden states in recurrent neural networks |
publisher |
IEEE |
publishDate |
2022 |
url |
https://doi.org/10.1109/JSTARS.2022.3155967 https://doaj.org/article/9a0b5112bc5c4c49998d976d76a30003 |
long_lat |
ENVELOPE(-60.633,-60.633,-62.950,-62.950) |
geographic |
Deception Island |
geographic_facet |
Deception Island |
genre |
Antarc* Antarctica Deception Island |
genre_facet |
Antarc* Antarctica Deception Island |
op_source |
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol 15, Pp 2311-2325 (2022) |
op_relation |
https://ieeexplore.ieee.org/document/9726918/ https://doaj.org/toc/2151-1535 2151-1535 doi:10.1109/JSTARS.2022.3155967 https://doaj.org/article/9a0b5112bc5c4c49998d976d76a30003 |
op_doi |
https://doi.org/10.1109/JSTARS.2022.3155967 |
container_title |
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
container_volume |
15 |
container_start_page |
2311 |
op_container_end_page |
2325 |
_version_ |
1766267314299207680 |