Learning to Order: Task Sequencing as In-Context Optimization

TL;DR

This paper introduces a meta-learning approach using transformer architectures to improve task sequencing in deep learning, enabling quick generalization to new problems with minimal demonstrations.

Contribution

It demonstrates that deep neural networks can meta-learn over an infinite prior of synthetic task sequencing problems, achieving few-shot generalization.

Findings

Meta-learned models discover optimal sequences faster than baselines.

Transformer-based architecture effectively generalizes to new sequencing tasks.

Empirical evidence supports the efficiency of the proposed meta-learning approach.

Abstract

Task sequencing (TS) is one of the core open problems in Deep Learning, arising in a plethora of real-world domains, from robotic assembly lines to autonomous driving. Unfortunately, prior work has not convincingly demonstrated the generalization ability of meta-learned TS methods to solve new TS problems, given few initial demonstrations. In this paper, we demonstrate that deep neural networks can meta-learn over an infinite prior of synthetically generated TS problems and achieve a few-shot generalization. We meta-learn a transformer-based architecture over datasets of sequencing trajectories generated from a prior distribution that samples sequencing problems as paths in directed graphs. In a large-scale experiment, we provide ample empirical evidence that our meta-learned models discover optimal task sequences significantly quicker than non-meta-learned baselines.