Saved in:
Bibliographic Details
Main Authors: Malviya, Pranshu, Huang, Jerry, Baratin, Aristide, Fournier, Quentin, Chandar, Sarath
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.15895
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909649456857088
author Malviya, Pranshu
Huang, Jerry
Baratin, Aristide
Fournier, Quentin
Chandar, Sarath
author_facet Malviya, Pranshu
Huang, Jerry
Baratin, Aristide
Fournier, Quentin
Chandar, Sarath
contents Determining the optimal model for a given task often requires training multiple models from scratch, which becomes impractical as dataset and model sizes grow. A more efficient alternative is to expand smaller pre-trained models, but this approach is underutilized due to a limited understanding of its impact on the training dynamics. Existing methods for quantifying this impact have notable limitations, including computation cost. To address this, we introduce a new perspective based on the loss landscape, which has been shown to contain a manifold of linearly connected minima. Specifically, we propose a metric that estimates the size of this manifold to study the impact of model expansion. Our experiments reveal a strong correlation between performance gains and our manifold metric, enabling more informed model comparison and offering a first step toward a geometry-driven approach for reliable model expansion. Notably, our metric outperforms other baselines, even when different types of expansion with equivalent number of parameters are applied to a model.
format Preprint
id arxiv_https___arxiv_org_abs_2405_15895
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Manifold Metric: A Loss Landscape Approach for Predicting Model Performance
Malviya, Pranshu
Huang, Jerry
Baratin, Aristide
Fournier, Quentin
Chandar, Sarath
Machine Learning
Determining the optimal model for a given task often requires training multiple models from scratch, which becomes impractical as dataset and model sizes grow. A more efficient alternative is to expand smaller pre-trained models, but this approach is underutilized due to a limited understanding of its impact on the training dynamics. Existing methods for quantifying this impact have notable limitations, including computation cost. To address this, we introduce a new perspective based on the loss landscape, which has been shown to contain a manifold of linearly connected minima. Specifically, we propose a metric that estimates the size of this manifold to study the impact of model expansion. Our experiments reveal a strong correlation between performance gains and our manifold metric, enabling more informed model comparison and offering a first step toward a geometry-driven approach for reliable model expansion. Notably, our metric outperforms other baselines, even when different types of expansion with equivalent number of parameters are applied to a model.
title Manifold Metric: A Loss Landscape Approach for Predicting Model Performance
topic Machine Learning
url https://arxiv.org/abs/2405.15895