APE: Aligning Pretrained Encoders to Quickly Learn Aligned Multimodal Representations

Elan Rosenfeld, Preetum Nakkiran, Hadi Pouransari, Oncel Tuzel, Fartash Faghri

Keywords: multimodal representations, zero-shot classification, pretraining, contrastive language-image pretraining, CLIP

TL;DR: We study a method for aligning pretrained unimodal encoders via small auxiliary functions. We show that in settings with limited training time and/or data this approach trains faster, reaches higher accuracy, and is more robust to distribution shift.

Abstract: Recent advances in learning aligned multimodal representations have been primarily driven by training large neural networks on massive, noisy paired-modality datasets. In this work, we ask whether it is possible to achieve similar results with substantially less training time and data. We achieve this by taking advantage of existing pretrained unimodal encoders and careful curation of alignment data relevant to the downstream task of interest. We study a natural approach to aligning existing encoders via small auxiliary functions, and we find that this method is competitive with (or outperforms) state of the art in many settings while being less prone to overfitting, less costly to train, and more robust to distribution shift. With a carefully chosen alignment distribution, our method surpasses prior state of the art for ImageNet zero-shot classification on public data while using two orders of magnitude less time and data and training 77% fewer parameters.