Layerwise Bregman Representation Learning with Applications to Knowledge Distillation

In this work, we propose a novel approach for layerwise representation learning of a trained neural network. In particular, we form a Bregman divergence based on the layer's transfer function and construct an extension of the original Bregman PCA formulation by incorporating a mean vector and n...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2022-09
Main Authors: Amid, Ehsan, Rohan, Anil, Fifty, Christopher, Warmuth, Manfred K
Format: Article
Language:English
Subjects:
Distillation
Divergence
Machine learning
Neural networks
Representations
Teachers
Transfer functions
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this work, we propose a novel approach for layerwise representation learning of a trained neural network. In particular, we form a Bregman divergence based on the layer's transfer function and construct an extension of the original Bregman PCA formulation by incorporating a mean vector and normalizing the principal directions with respect to the geometry of the local convex function around the mean. This generalization allows exporting the learned representation as a fixed layer with a non-linearity. As an application to knowledge distillation, we cast the learning problem for the student network as predicting the compression coefficients of the teacher's representations, which are passed as the input to the imported layer. Our empirical findings indicate that our approach is substantially more effective for transferring information between networks than typical teacher-student training using the teacher's penultimate layer representations and soft labels.
ISSN:2331-8422