Minimax and Neyman-Pearson Meta-Learning for Outlier Languages

TL;DR

This paper introduces Minimax and Neyman-Pearson meta-learning variants of MAML to improve robustness for outlier languages in cross-lingual NLP, demonstrating enhanced performance on low-resource language tasks.

Contribution

It proposes two novel meta-learning algorithms based on alternative risk criteria, addressing MAML's limitations with outlier languages in cross-lingual NLP.

Findings

Improved average and minimum performance on low-resource languages.

Effective in zero- and few-shot learning scenarios.

Outperforms joint transfer and vanilla MAML.

Abstract

Model-agnostic meta-learning (MAML) has been recently put forth as a strategy to learn resource-poor languages in a sample-efficient fashion. Nevertheless, the properties of these languages are often not well represented by those available during training. Hence, we argue that the i.i.d. assumption ingrained in MAML makes it ill-suited for cross-lingual NLP. In fact, under a decision-theoretic framework, MAML can be interpreted as minimising the expected risk across training languages (with a uniform prior), which is known as Bayes criterion. To increase its robustness to outlier languages, we create two variants of MAML based on alternative criteria: Minimax MAML reduces the maximum risk across languages, while Neyman-Pearson MAML constrains the risk in each language to a maximum threshold. Both criteria constitute fully differentiable two-player games. In light of this, we propose a new adaptive optimiser solving for a local approximation to their Nash equilibrium. We evaluate both model variants on two popular NLP tasks, part-of-speech tagging and question answering. We report gains for their average and minimum performance across low-resource languages in zero- and few-shot settings, compared to joint multi-source transfer and vanilla MAML.