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...
| Главные авторы: | , , , |
|---|---|
| Формат: | Статья |
| Язык: | English |
| Опубликовано: |
IEEE
2022-01-01
|
| Серии: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Предметы: | |
| Online-ссылка: | https://ieeexplore.ieee.org/document/9726918/ |
| _version_ | 1828872427844665344 |
|---|---|
| author | Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez |
| author_facet | Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez |
| author_sort | Manuel Titos |
| collection | DOAJ |
| 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 <italic>what</italic> and <italic>how</italic> 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 <italic>deception island volcano</italic> (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. |
| first_indexed | 2024-12-13T06:53:40Z |
| format | Article |
| id | doaj.art-9a0b5112bc5c4c49998d976d76a30003 |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-12-13T06:53:40Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-9a0b5112bc5c4c49998d976d76a300032022-12-21T23:56:05ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352022-01-01152311232510.1109/JSTARS.2022.31559679726918Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural NetworksManuel Titos0https://orcid.org/0000-0002-8279-2341Luz Garcia1https://orcid.org/0000-0001-5904-5412Milad Kowsari2Carmen Benitez3https://orcid.org/0000-0002-5407-8335Department of Signal Theory, Telematics and Communications, University of Granada, Granada, SpainDepartment of Signal Theory, Telematics and Communications, University of Granada, Granada, SpainFaculty of Civil and Environmental Engineering, School of Engineering and Natural Sciences, University of Iceland, Reykjavík, IcelandDepartment of Signal Theory, Telematics and Communications, University of Granada, Granada, SpainUnderstanding 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 <italic>what</italic> and <italic>how</italic> 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 <italic>deception island volcano</italic> (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.https://ieeexplore.ieee.org/document/9726918/Knowledge based systemslearning (artificial intelligence)supervised learningmachine learningdeep learningrepresentation learning |
| spellingShingle | Manuel Titos Luz Garcia Milad Kowsari Carmen Benitez Toward Knowledge Extraction in Classification of Volcano-Seismic Events: Visualizing Hidden States in Recurrent Neural Networks IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Knowledge based systems learning (artificial intelligence) supervised learning machine learning deep learning representation learning |
| title | 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_short | 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 |
| topic | Knowledge based systems learning (artificial intelligence) supervised learning machine learning deep learning representation learning |
| url | https://ieeexplore.ieee.org/document/9726918/ |
| work_keys_str_mv | AT manueltitos towardknowledgeextractioninclassificationofvolcanoseismiceventsvisualizinghiddenstatesinrecurrentneuralnetworks AT luzgarcia towardknowledgeextractioninclassificationofvolcanoseismiceventsvisualizinghiddenstatesinrecurrentneuralnetworks AT miladkowsari towardknowledgeextractioninclassificationofvolcanoseismiceventsvisualizinghiddenstatesinrecurrentneuralnetworks AT carmenbenitez towardknowledgeextractioninclassificationofvolcanoseismiceventsvisualizinghiddenstatesinrecurrentneuralnetworks |