Decoupled Training: Return of Frustratingly Easy Multi-Domain Learning

TL;DR

This paper introduces Decoupled Training (D-Train), a simple, hyperparameter-free multi-domain learning method that improves domain independence and performance across diverse datasets with minimal complexity.

Contribution

The paper presents a novel, straightforward tri-phase training strategy for multi-domain learning that is easy to implement, hyperparameter-free, and effective across various applications.

Findings

D-Train achieves competitive results on standard benchmarks.

It effectively handles dataset bias and domain domination issues.

The method is computationally efficient and simple to deploy.

Abstract

Multi-domain learning (MDL) aims to train a model with minimal average risk across multiple overlapping but non-identical domains. To tackle the challenges of dataset bias and domain domination, numerous MDL approaches have been proposed from the perspectives of seeking commonalities by aligning distributions to reduce domain gap or reserving differences by implementing domain-specific towers, gates, and even experts. MDL models are becoming more and more complex with sophisticated network architectures or loss functions, introducing extra parameters and enlarging computation costs. In this paper, we propose a frustratingly easy and hyperparameter-free multi-domain learning method named Decoupled Training (D-Train). D-Train is a tri-phase general-to-specific training strategy that first pre-trains on all domains to warm up a root model, then post-trains on each domain by splitting into multi-heads, and finally fine-tunes the heads by fixing the backbone, enabling decouple training to achieve domain independence. Despite its extraordinary simplicity and efficiency, D-Train performs remarkably well in extensive evaluations of various datasets from standard benchmarks to applications of satellite imagery and recommender systems.