Task Addition in Multi-Task Learning by Geometrical Alignment

TL;DR

This paper introduces a task addition method for GATE, enhancing molecular property prediction by improving performance on limited data tasks with minimal extra computation.

Contribution

It proposes a scalable task addition approach for GATE that combines multi-task pre-training with task-specific modules, outperforming traditional methods.

Findings

Superior performance of task addition strategy over conventional methods

Maintains comparable computational costs

Effective in limited data scenarios

Abstract

Training deep learning models on limited data while maintaining generalization is one of the fundamental challenges in molecular property prediction. One effective solution is transferring knowledge extracted from abundant datasets to those with scarce data. Recently, a novel algorithm called Geometrically Aligned Transfer Encoder (GATE) has been introduced, which uses soft parameter sharing by aligning the geometrical shapes of task-specific latent spaces. However, GATE faces limitations in scaling to multiple tasks due to computational costs. In this study, we propose a task addition approach for GATE to improve performance on target tasks with limited data while minimizing computational complexity. It is achieved through supervised multi-task pre-training on a large dataset, followed by the addition and training of task-specific modules for each target task. Our experiments demonstrate the superior performance of the task addition strategy for GATE over conventional multi-task methods, with comparable computational costs.