<div class="csl-bib-body">
<div class="csl-entry">Pichler, G., Colombo, P. J. A., Boudiaf, M., Koliander, G., & Piantanida, P. (2022). A Differential Entropy Estimator for Training Neural Networks. In K. Chaudhuri, S. Jegelka, L. Song, C. Szepesvari, G. Niu, & S. Sabato (Eds.), <i>Proceedings of the 39th International Conference on Machine Learning</i> (pp. 17691–17715). PMLR. http://hdl.handle.net/20.500.12708/135918</div>
</div>
-
dc.identifier.uri
http://hdl.handle.net/20.500.12708/135918
-
dc.description.abstract
Mutual Information (MI) has been widely used as a loss regularizer for training neural networks. This has been particularly effective when learn disentangled or compressed representations of high dimensional data. However, differential entropy (DE), another fundamental measure of information, has not found widespread use in neural network training. Although DE offers a potentially wider range of applications than MI, off-the-shelf DE estimators are either non differentiable, computationally intractable or fail to adapt to changes in the underlying distribution. These drawbacks prevent them from being used as regularizers in neural networks training. To address shortcomings in previously proposed estimators for DE, here we introduce KNIFE, a fully parameterized, differentiable kernel-based estimator of DE. The flexibility of our approach also allows us to construct KNIFE-based estimators for conditional (on either discrete or continuous variables) DE, as well as MI. We empirically validate our method on high-dimensional synthetic data and further apply it to guide the training of neural networks for real-world tasks. Our experiments on a large variety of tasks, including visual domain adaptation, textual fair classification, and textual fine-tuning demonstrate the effectiveness of KNIFE-based estimation. Code can be found at https://github.com/g-pichler/knife.
en
dc.language.iso
en
-
dc.subject
Mutual Information
en
dc.subject
information measures
en
dc.subject
Kernel Density Estimation
en
dc.subject
differential entropy
en
dc.title
A Differential Entropy Estimator for Training Neural Networks
en
dc.type
Inproceedings
en
dc.type
Konferenzbeitrag
de
dc.contributor.affiliation
Université Paris Saclay (MICS)
-
dc.contributor.affiliation
ETS Montreal
-
dc.contributor.affiliation
Austrian Academy of Sciences
-
dc.contributor.affiliation
International Laboratory on Learning Systems (ILLS)
-
dc.contributor.editoraffiliation
MIT EECS
-
dc.contributor.editoraffiliation
Mohamed bin Zayed University of Artificial Intelligence
-
dc.contributor.editoraffiliation
University of Alberta
-
dc.contributor.editoraffiliation
RIKEN Center for Advanced Intelligence Project
-
dc.contributor.editoraffiliation
Ben-Gurion University of the Negev
-
dc.description.startpage
17691
-
dc.description.endpage
17715
-
dcterms.dateSubmitted
2022-01-27
-
dc.type.category
Poster Contribution
-
tuw.booktitle
Proceedings of the 39th International Conference on Machine Learning
-
tuw.container.volume
162
-
tuw.relation.publisher
PMLR
-
tuw.researchTopic.id
I2
-
tuw.researchTopic.id
I4a
-
tuw.researchTopic.id
C5
-
tuw.researchTopic.name
Computer Engineering and Software-Intensive Systems