File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Evaluating Capability of Deep Neural Networks for Image Classification via Information Plane

TitleEvaluating Capability of Deep Neural Networks for Image Classification via Information Plane
Authors
KeywordsImage classification
Information bottleneck
Mutual information
Neural networks
Issue Date2018
Citation
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2018, v. 11215 LNCS, p. 181-195 How to Cite?
AbstractInspired by the pioneering work of information bottleneck principle for Deep Neural Networks (DNNs) analysis, we design an information plane based framework to evaluate the capability of DNNs for image classification tasks, which not only helps understand the capability of DNNs, but also helps us choose a neural network which leads to higher classification accuracy more efficiently. Further, with experiments, the relationship among the model accuracy, I(X; T) and I(T; Y) are analyzed, where I(X; T) and I(T; Y) are the mutual information of DNN’s output T with input X and label Y. We also show the information plane is more informative than loss curve and apply mutual information to infer the model’s capability for recognizing objects of each class. Our studies would facilitate a better understanding of DNNs.
Persistent Identifierhttp://hdl.handle.net/10722/345237
ISSN
2023 SCImago Journal Rankings: 0.606

 

DC FieldValueLanguage
dc.contributor.authorCheng, Hao-
dc.contributor.authorLian, Dongze-
dc.contributor.authorGao, Shenghua-
dc.contributor.authorGeng, Yanlin-
dc.date.accessioned2024-08-15T09:26:05Z-
dc.date.available2024-08-15T09:26:05Z-
dc.date.issued2018-
dc.identifier.citationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2018, v. 11215 LNCS, p. 181-195-
dc.identifier.issn0302-9743-
dc.identifier.urihttp://hdl.handle.net/10722/345237-
dc.description.abstractInspired by the pioneering work of information bottleneck principle for Deep Neural Networks (DNNs) analysis, we design an information plane based framework to evaluate the capability of DNNs for image classification tasks, which not only helps understand the capability of DNNs, but also helps us choose a neural network which leads to higher classification accuracy more efficiently. Further, with experiments, the relationship among the model accuracy, I(X; T) and I(T; Y) are analyzed, where I(X; T) and I(T; Y) are the mutual information of DNN’s output T with input X and label Y. We also show the information plane is more informative than loss curve and apply mutual information to infer the model’s capability for recognizing objects of each class. Our studies would facilitate a better understanding of DNNs.-
dc.languageeng-
dc.relation.ispartofLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)-
dc.subjectImage classification-
dc.subjectInformation bottleneck-
dc.subjectMutual information-
dc.subjectNeural networks-
dc.titleEvaluating Capability of Deep Neural Networks for Image Classification via Information Plane-
dc.typeConference_Paper-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1007/978-3-030-01252-6_11-
dc.identifier.scopuseid_2-s2.0-85055132133-
dc.identifier.volume11215 LNCS-
dc.identifier.spage181-
dc.identifier.epage195-
dc.identifier.eissn1611-3349-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats