TAP: The Attention Patch for Cross-Modal Knowledge Transfer from Unlabeled Modality

TL;DR

This paper introduces TAP, a neural network add-on that leverages an unlabeled secondary modality via kernelized cross-attention to improve supervised learning in the primary modality, demonstrating significant generalization gains.

Contribution

The paper proposes The Attention Patch (TAP), a novel method for cross-modal knowledge transfer using an attention mechanism derived from kernel regression, applicable to various neural architectures.

Findings

TAP significantly improves generalization across domains.

It effectively utilizes unlabeled cross-modal data.

The method is compatible with different neural network architectures.

Abstract

This paper addresses a cross-modal learning framework, where the objective is to enhance the performance of supervised learning in the primary modality using an unlabeled, unpaired secondary modality. Taking a probabilistic approach for missing information estimation, we show that the extra information contained in the secondary modality can be estimated via Nadaraya-Watson (NW) kernel regression, which can further be expressed as a kernelized cross-attention module (under linear transformation). This expression lays the foundation for introducing The Attention Patch (TAP), a simple neural network add-on that can be trained to allow data-level knowledge transfer from the unlabeled modality. We provide extensive numerical simulations using real-world datasets to show that TAP can provide statistically significant improvement in generalization across different domains and different neural network architectures, making use of seemingly unusable unlabeled cross-modal data.