Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network
International audience In a previous work, we have developed a framework for the multimodal and hierarchical classification of images from soil remediation reports. We extended this work using Deep Metric Learning (DML) as an additional training step to improve embeddings quality and obtained 84.24%...
Main Authors: | , , |
---|---|
Other Authors: | , |
Format: | Conference Object |
Language: | English |
Published: |
HAL CCSD
2022
|
Subjects: | |
Online Access: | https://hal.science/hal-03981883 https://hal.science/hal-03981883/document https://hal.science/hal-03981883/file/W18P05.pdf |
id |
ftunivnantes:oai:HAL:hal-03981883v1 |
---|---|
record_format |
openpolar |
spelling |
ftunivnantes:oai:HAL:hal-03981883v1 2023-05-15T16:01:54+02:00 Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network Rysbayeva, Korlan Giot, Romain Journet, Nicholas Laboratoire Bordelais de Recherche en Informatique (LaBRI) Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS) Montréal, Canada 2022-08-21 https://hal.science/hal-03981883 https://hal.science/hal-03981883/document https://hal.science/hal-03981883/file/W18P05.pdf en eng HAL CCSD hal-03981883 https://hal.science/hal-03981883 https://hal.science/hal-03981883/document https://hal.science/hal-03981883/file/W18P05.pdf info:eu-repo/semantics/OpenAccess ICPR 2022 workshops 2-nd Workshop on Explainable and Ethical AI – ICPR 2022 https://hal.science/hal-03981883 2-nd Workshop on Explainable and Ethical AI – ICPR 2022, Aug 2022, Montréal, Canada https://xaie-icpr.labri.fr/ Graph analysis Hierarchical embeddings eXplainable Artificial Intelligence [INFO]Computer Science [cs] [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI] info:eu-repo/semantics/conferenceObject Conference papers 2022 ftunivnantes 2023-03-01T00:59:36Z International audience In a previous work, we have developed a framework for the multimodal and hierarchical classification of images from soil remediation reports. We extended this work using Deep Metric Learning (DML) as an additional training step to improve embeddings quality and obtained 84.24% of weighted F1 score for the level 5th hierarchical level. However, the standard classifier performance metrics are insufficient to explain the decision process reasoning. So far of our knowledge, there are no methods to analyze hierarchical classification algorithms. In this work, we propose a method of graph analysis to describe the embeddings that represent the extended classifier, which we believe properly interprets the obtained results than classification metrics. We illustrate the method of analyzing hierarchical classification algorithms on private dataset, but the method remains generic enough to be used in other contexts. Conference Object DML Université de Nantes: HAL-UNIV-NANTES Canada |
institution |
Open Polar |
collection |
Université de Nantes: HAL-UNIV-NANTES |
op_collection_id |
ftunivnantes |
language |
English |
topic |
Graph analysis Hierarchical embeddings eXplainable Artificial Intelligence [INFO]Computer Science [cs] [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI] |
spellingShingle |
Graph analysis Hierarchical embeddings eXplainable Artificial Intelligence [INFO]Computer Science [cs] [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI] Rysbayeva, Korlan Giot, Romain Journet, Nicholas Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network |
topic_facet |
Graph analysis Hierarchical embeddings eXplainable Artificial Intelligence [INFO]Computer Science [cs] [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI] |
description |
International audience In a previous work, we have developed a framework for the multimodal and hierarchical classification of images from soil remediation reports. We extended this work using Deep Metric Learning (DML) as an additional training step to improve embeddings quality and obtained 84.24% of weighted F1 score for the level 5th hierarchical level. However, the standard classifier performance metrics are insufficient to explain the decision process reasoning. So far of our knowledge, there are no methods to analyze hierarchical classification algorithms. In this work, we propose a method of graph analysis to describe the embeddings that represent the extended classifier, which we believe properly interprets the obtained results than classification metrics. We illustrate the method of analyzing hierarchical classification algorithms on private dataset, but the method remains generic enough to be used in other contexts. |
author2 |
Laboratoire Bordelais de Recherche en Informatique (LaBRI) Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS) |
format |
Conference Object |
author |
Rysbayeva, Korlan Giot, Romain Journet, Nicholas |
author_facet |
Rysbayeva, Korlan Giot, Romain Journet, Nicholas |
author_sort |
Rysbayeva, Korlan |
title |
Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network |
title_short |
Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network |
title_full |
Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network |
title_fullStr |
Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network |
title_full_unstemmed |
Graph-based Analysis of Hierarchical Embedding Generated by Deep Neural Network |
title_sort |
graph-based analysis of hierarchical embedding generated by deep neural network |
publisher |
HAL CCSD |
publishDate |
2022 |
url |
https://hal.science/hal-03981883 https://hal.science/hal-03981883/document https://hal.science/hal-03981883/file/W18P05.pdf |
op_coverage |
Montréal, Canada |
geographic |
Canada |
geographic_facet |
Canada |
genre |
DML |
genre_facet |
DML |
op_source |
ICPR 2022 workshops 2-nd Workshop on Explainable and Ethical AI – ICPR 2022 https://hal.science/hal-03981883 2-nd Workshop on Explainable and Ethical AI – ICPR 2022, Aug 2022, Montréal, Canada https://xaie-icpr.labri.fr/ |
op_relation |
hal-03981883 https://hal.science/hal-03981883 https://hal.science/hal-03981883/document https://hal.science/hal-03981883/file/W18P05.pdf |
op_rights |
info:eu-repo/semantics/OpenAccess |
_version_ |
1766397591269933056 |