Saved in:
Bibliographic Details
Main Authors: Denoodt, Fabian, de Boer, Bart, Oramas, José
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.12936
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908415325896704
author Denoodt, Fabian
de Boer, Bart
Oramas, José
author_facet Denoodt, Fabian
de Boer, Bart
Oramas, José
contents We introduce Smooth InfoMax (SIM), a self-supervised representation learning method that incorporates interpretability constraints into the latent representations at different depths of the network. Based on $β$-VAEs, SIM's architecture consists of probabilistic modules optimized locally with the InfoNCE loss to produce Gaussian-distributed representations regularized toward the standard normal distribution. This creates smooth, well-defined, and better-disentangled latent spaces, enabling easier post-hoc analysis. Evaluated on speech data, SIM preserves the large-scale training benefits of Greedy InfoMax while improving the effectiveness of post-hoc interpretability methods across layers.
format Preprint
id arxiv_https___arxiv_org_abs_2408_12936
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Smooth InfoMax -- Towards Easier Post-Hoc Interpretability
Denoodt, Fabian
de Boer, Bart
Oramas, José
Machine Learning
Artificial Intelligence
We introduce Smooth InfoMax (SIM), a self-supervised representation learning method that incorporates interpretability constraints into the latent representations at different depths of the network. Based on $β$-VAEs, SIM's architecture consists of probabilistic modules optimized locally with the InfoNCE loss to produce Gaussian-distributed representations regularized toward the standard normal distribution. This creates smooth, well-defined, and better-disentangled latent spaces, enabling easier post-hoc analysis. Evaluated on speech data, SIM preserves the large-scale training benefits of Greedy InfoMax while improving the effectiveness of post-hoc interpretability methods across layers.
title Smooth InfoMax -- Towards Easier Post-Hoc Interpretability
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2408.12936