Equi-Tuning: Group Equivariant Fine-Tuning of Pretrained Models

TL;DR

Equi-tuning is a new fine-tuning approach that transforms pretrained models into group equivariant models, enhancing their performance on various tasks by combining inductive biases with pretrained semantic knowledge.

Contribution

The paper introduces equi-tuning, a method for converting pretrained models into group equivariant models with minimal loss, applicable across diverse tasks and model architectures.

Findings

Equi-tuning improves model performance on image classification, language generalization, and fairness tasks.

The method maintains semantic priors while enforcing group equivariance.

Experimental results demonstrate broad applicability and effectiveness across multiple models and datasets.

Abstract

We introduce equi-tuning, a novel fine-tuning method that transforms (potentially non-equivariant) pretrained models into group equivariant models while incurring minimum $L_2$ loss between the feature representations of the pretrained and the equivariant models. Large pretrained models can be equi-tuned for different groups to satisfy the needs of various downstream tasks. Equi-tuned models benefit from both group equivariance as an inductive bias and semantic priors from pretrained models. We provide applications of equi-tuning on three different tasks: image classification, compositional generalization in language, and fairness in natural language generation (NLG). We also provide a novel group-theoretic definition for fairness in NLG. The effectiveness of this definition is shown by testing it against a standard empirical method of fairness in NLG. We provide experimental results for equi-tuning using a variety of pretrained models: Alexnet, Resnet, VGG, and Densenet for image classification; RNNs, GRUs, and LSTMs for compositional generalization; and GPT2 for fairness in NLG. We test these models on benchmark datasets across all considered tasks to show the generality and effectiveness of the proposed method.