Learning Neural Networks with Sparse Activations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Awasthi, Pranjal, Dikkala, Nishanth, Kamath, Pritish, Meka, Raghu
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909231659089920
author Awasthi, Pranjal
Dikkala, Nishanth
Kamath, Pritish
Meka, Raghu
author_facet Awasthi, Pranjal
Dikkala, Nishanth
Kamath, Pritish
Meka, Raghu
contents A core component present in many successful neural network architectures, is an MLP block of two fully connected layers with a non-linear activation in between. An intriguing phenomenon observed empirically, including in transformer architectures, is that, after training, the activations in the hidden layer of this MLP block tend to be extremely sparse on any given input. Unlike traditional forms of sparsity, where there are neurons/weights which can be deleted from the network, this form of {\em dynamic} activation sparsity appears to be harder to exploit to get more efficient networks. Motivated by this we initiate a formal study of PAC learnability of MLP layers that exhibit activation sparsity. We present a variety of results showing that such classes of functions do lead to provable computational and statistical advantages over their non-sparse counterparts. Our hope is that a better theoretical understanding of {\em sparsely activated} networks would lead to methods that can exploit activation sparsity in practice.
format Preprint
id arxiv_https___arxiv_org_abs_2406_17989
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning Neural Networks with Sparse Activations
Awasthi, Pranjal
Dikkala, Nishanth
Kamath, Pritish
Meka, Raghu
Machine Learning
A core component present in many successful neural network architectures, is an MLP block of two fully connected layers with a non-linear activation in between. An intriguing phenomenon observed empirically, including in transformer architectures, is that, after training, the activations in the hidden layer of this MLP block tend to be extremely sparse on any given input. Unlike traditional forms of sparsity, where there are neurons/weights which can be deleted from the network, this form of {\em dynamic} activation sparsity appears to be harder to exploit to get more efficient networks. Motivated by this we initiate a formal study of PAC learnability of MLP layers that exhibit activation sparsity. We present a variety of results showing that such classes of functions do lead to provable computational and statistical advantages over their non-sparse counterparts. Our hope is that a better theoretical understanding of {\em sparsely activated} networks would lead to methods that can exploit activation sparsity in practice.
title Learning Neural Networks with Sparse Activations
topic Machine Learning
url https://arxiv.org/abs/2406.17989