Timezone: »

Sign-MAML: Efficient Model-Agnostic Meta-Learning by SignSGD
Chen Fan · Parikshit Ram · Sijia Liu
Event URL: https://openreview.net/forum?id=bd0UOwKS_6j »

We propose a new computationally-efficient first-order algorithm for Model-Agnostic Meta-Learning (MAML). The key enabling technique is to interpret MAML as a bilevel optimization (BLO) problem and leverage the sign-based SGD (signSGD) as a lower-level optimizer of BLO. We show that MAML, through the lens of signSGD-oriented BLO, naturally yields an alternating optimization scheme that just requires first-order gradients of a learned meta-model. We term the resulting MAML algorithm Sign-MAML. Compared to the conventional first-order MAML (FO-MAML) algorithm, Sign-MAML is theoretically-grounded as it does not impose any assumption on the absence of second-order derivatives during meta training. In practice, we show that Sign-MAML outperforms FO-MAML in various few-shot image classification tasks, and compared to MAML, it achieves a much more graceful tradeoff between classification accuracy and computation efficiency.

Author Information

Chen Fan (University of Massachusetts, Amherst)
Parikshit Ram (IBM Research AI)
Sijia Liu (Michigan State University)

More from the Same Authors