Evaluation of CNN Performance in Semantically Relevant Latent Spaces

Jeroen van Doorenmalen, Vlado Menkovski

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

1 Citaat (Scopus)


We examine deep neural network (DNN) performance and behavior using contrasting explanations generated from a semantically relevant latent space. We develop a semantically relevant latent space by training a variational autoencoder (VAE) augmented by a metric learning loss on the latent space. The properties of the VAE provide for a smooth latent space supported by a simple density and the metric learning term organizes the space in a semantically relevant way with respect to the target classes. In this space we can both linearly separate the classes and generate meaningful interpolation of contrasting data points across decision boundaries. This allows us to examine the DNN model beyond its performance on a test set for potential biases and its sensitivity to perturbations of individual factors disentangled in the latent space.

Originele taal-2Engels
TitelAdvances in Intelligent Data Analysis XVIII - 18th International Symposium on Intelligent Data Analysis, IDA 2020, Proceedings
RedacteurenMichael R. Berthold, Ad Feelders, Georg Krempl
Aantal pagina's13
ISBN van geprinte versie9783030445836
StatusGepubliceerd - 2020
Evenement18th International Conference on Intelligent Data Analysis, IDA 2020 - Konstanz, Duitsland
Duur: 27 apr. 202029 apr. 2020

Publicatie series

NaamLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12080 LNCS
ISSN van geprinte versie0302-9743
ISSN van elektronische versie1611-3349


Congres18th International Conference on Intelligent Data Analysis, IDA 2020


Duik in de onderzoeksthema's van 'Evaluation of CNN Performance in Semantically Relevant Latent Spaces'. Samen vormen ze een unieke vingerafdruk.

Citeer dit