A Second-Order Perspective on Pruning at Initialization and Knowledge Transfer

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Iurada, Leonardo, Occhiena, Beatrice, Tommasi, Tatiana
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911182347042816
author Iurada, Leonardo
Occhiena, Beatrice
Tommasi, Tatiana
author_facet Iurada, Leonardo
Occhiena, Beatrice
Tommasi, Tatiana
contents The widespread availability of pre-trained vision models has enabled numerous deep learning applications through their transferable representations. However, their computational and storage costs often limit practical deployment. Pruning-at-Initialization has emerged as a promising approach to compress models before training, enabling efficient task-specific adaptation. While conventional wisdom suggests that effective pruning requires task-specific data, this creates a challenge when downstream tasks are unknown in advance. In this paper, we investigate how data influences the pruning of pre-trained vision models. Surprisingly, pruning on one task retains the model's zero-shot performance also on unseen tasks. Furthermore, fine-tuning these pruned models not only improves performance on original seen tasks but can recover held-out tasks' performance. We attribute this phenomenon to the favorable loss landscapes induced by extensive pre-training on large-scale datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2509_24066
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Second-Order Perspective on Pruning at Initialization and Knowledge Transfer
Iurada, Leonardo
Occhiena, Beatrice
Tommasi, Tatiana
Computer Vision and Pattern Recognition
Artificial Intelligence
The widespread availability of pre-trained vision models has enabled numerous deep learning applications through their transferable representations. However, their computational and storage costs often limit practical deployment. Pruning-at-Initialization has emerged as a promising approach to compress models before training, enabling efficient task-specific adaptation. While conventional wisdom suggests that effective pruning requires task-specific data, this creates a challenge when downstream tasks are unknown in advance. In this paper, we investigate how data influences the pruning of pre-trained vision models. Surprisingly, pruning on one task retains the model's zero-shot performance also on unseen tasks. Furthermore, fine-tuning these pruned models not only improves performance on original seen tasks but can recover held-out tasks' performance. We attribute this phenomenon to the favorable loss landscapes induced by extensive pre-training on large-scale datasets.
title A Second-Order Perspective on Pruning at Initialization and Knowledge Transfer
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2509.24066