Model Diffusion for Certifiable Few-shot Transfer Learning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Rezk, Fady, Lee, Royson, Gouk, Henry, Hospedales, Timothy, Kim, Minyoung
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913862805094400
author Rezk, Fady
Lee, Royson
Gouk, Henry
Hospedales, Timothy
Kim, Minyoung
author_facet Rezk, Fady
Lee, Royson
Gouk, Henry
Hospedales, Timothy
Kim, Minyoung
contents In contemporary deep learning, a prevalent and effective workflow for solving low-data problems is adapting powerful pre-trained foundation models (FMs) to new tasks via parameter-efficient fine-tuning (PEFT). However, while empirically effective, the resulting solutions lack generalisation guarantees to certify their accuracy - which may be required for ethical or legal reasons prior to deployment in high-importance applications. In this paper we develop a novel transfer learning approach that is designed to facilitate non-vacuous learning theoretic generalisation guarantees for downstream tasks, even in the low-shot regime. Specifically, we first use upstream tasks to train a distribution over PEFT parameters. We then learn the downstream task by a sample-and-evaluate procedure -- sampling plausible PEFTs from the trained diffusion model and selecting the one with the highest likelihood on the downstream data. Crucially, this confines our model hypothesis to a finite set of PEFT samples. In contrast to the typical continuous hypothesis spaces of neural network weights, this facilitates tighter risk certificates. We instantiate our bound and show non-trivial generalization guarantees compared to existing learning approaches which lead to vacuous bounds in the low-shot regime.
format Preprint
id arxiv_https___arxiv_org_abs_2502_06970
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Model Diffusion for Certifiable Few-shot Transfer Learning
Rezk, Fady
Lee, Royson
Gouk, Henry
Hospedales, Timothy
Kim, Minyoung
Machine Learning
In contemporary deep learning, a prevalent and effective workflow for solving low-data problems is adapting powerful pre-trained foundation models (FMs) to new tasks via parameter-efficient fine-tuning (PEFT). However, while empirically effective, the resulting solutions lack generalisation guarantees to certify their accuracy - which may be required for ethical or legal reasons prior to deployment in high-importance applications. In this paper we develop a novel transfer learning approach that is designed to facilitate non-vacuous learning theoretic generalisation guarantees for downstream tasks, even in the low-shot regime. Specifically, we first use upstream tasks to train a distribution over PEFT parameters. We then learn the downstream task by a sample-and-evaluate procedure -- sampling plausible PEFTs from the trained diffusion model and selecting the one with the highest likelihood on the downstream data. Crucially, this confines our model hypothesis to a finite set of PEFT samples. In contrast to the typical continuous hypothesis spaces of neural network weights, this facilitates tighter risk certificates. We instantiate our bound and show non-trivial generalization guarantees compared to existing learning approaches which lead to vacuous bounds in the low-shot regime.
title Model Diffusion for Certifiable Few-shot Transfer Learning
topic Machine Learning
url https://arxiv.org/abs/2502.06970