Finding Strong Lottery Ticket Networks with Genetic Algorithms

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Altmann, Philipp, Schönberger, Julian, Zorn, Maximilian, Gabor, Thomas
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910687773589504
author Altmann, Philipp
Schönberger, Julian
Zorn, Maximilian
Gabor, Thomas
author_facet Altmann, Philipp
Schönberger, Julian
Zorn, Maximilian
Gabor, Thomas
contents According to the Strong Lottery Ticket Hypothesis, every sufficiently large neural network with randomly initialized weights contains a sub-network which - still with its random weights - already performs as well for a given task as the trained super-network. We present the first approach based on a genetic algorithm to find such strong lottery ticket sub-networks without training or otherwise computing any gradient. We show that, for smaller instances of binary classification tasks, our evolutionary approach even produces smaller and better-performing lottery ticket networks than the state-of-the-art approach using gradient information.
format Preprint
id arxiv_https___arxiv_org_abs_2411_04658
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Finding Strong Lottery Ticket Networks with Genetic Algorithms
Altmann, Philipp
Schönberger, Julian
Zorn, Maximilian
Gabor, Thomas
Neural and Evolutionary Computing
According to the Strong Lottery Ticket Hypothesis, every sufficiently large neural network with randomly initialized weights contains a sub-network which - still with its random weights - already performs as well for a given task as the trained super-network. We present the first approach based on a genetic algorithm to find such strong lottery ticket sub-networks without training or otherwise computing any gradient. We show that, for smaller instances of binary classification tasks, our evolutionary approach even produces smaller and better-performing lottery ticket networks than the state-of-the-art approach using gradient information.
title Finding Strong Lottery Ticket Networks with Genetic Algorithms
topic Neural and Evolutionary Computing
url https://arxiv.org/abs/2411.04658