Task Weighting in Meta-learning with Trajectory Optimisation

TL;DR

This paper introduces a fully-automated, trajectory-optimisation-based task weighting algorithm for meta-learning, which outperforms hand-designed methods in few-shot learning benchmarks.

Contribution

A novel meta-learning task-weighting algorithm using trajectory optimisation and iterative LQR, with proven convergence and improved empirical performance.

Findings

Outperforms hand-engineered weighting methods in benchmarks

Converges to an $0$-stationary point

Theoretically proven convergence properties

Abstract

Developing meta-learning algorithms that are un-biased toward a subset of training tasks often requires hand-designed criteria to weight tasks, potentially resulting in sub-optimal solutions. In this paper, we introduce a new principled and fully-automated task-weighting algorithm for meta-learning methods. By considering the weights of tasks within the same mini-batch as an action, and the meta-parameter of interest as the system state, we cast the task-weighting meta-learning problem to a trajectory optimisation and employ the iterative linear quadratic regulator to determine the optimal action or weights of tasks. We theoretically show that the proposed algorithm converges to an $\epsilon_{0}$-stationary point, and empirically demonstrate that the proposed approach out-performs common hand-engineering weighting methods in two few-shot learning benchmarks.