ModelGiF: Gradient Fields for Model Functional Distance

TL;DR

This paper introduces ModelGiF, a novel method using gradient fields to measure the functional distance between deep learning models, addressing challenges posed by model opacity and heterogeneity.

Contribution

The paper proposes ModelGiF, a new approach that captures model representations through gradient fields to quantify model similarity across diverse architectures and tasks.

Findings

ModelGiF effectively estimates task relatedness.

It aids in intellectual property protection.

It verifies model unlearning with high accuracy.

Abstract

The last decade has witnessed the success of deep learning and the surge of publicly released trained models, which necessitates the quantification of the model functional distance for various purposes. However, quantifying the model functional distance is always challenging due to the opacity in inner workings and the heterogeneity in architectures or tasks. Inspired by the concept of "field" in physics, in this work we introduce Model Gradient Field (abbr. ModelGiF) to extract homogeneous representations from the heterogeneous pre-trained models. Our main assumption underlying ModelGiF is that each pre-trained deep model uniquely determines a ModelGiF over the input space. The distance between models can thus be measured by the similarity between their ModelGiFs. We validate the effectiveness of the proposed ModelGiF with a suite of testbeds, including task relatedness estimation, intellectual property protection, and model unlearning verification. Experimental results demonstrate the versatility of the proposed ModelGiF on these tasks, with significantly superiority performance to state-of-the-art competitors. Codes are available at https://github.com/zju-vipa/modelgif.