English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Model-Agnostic Meta-Learning (MAML) and its variants have achieved success inmeta-learning tasks on many datasets and settings. On the other hand, we havejust started to understand and analyze how they are able to adapt fast to newtasks. For example, one popular hypothesis is that the algorithms learn goodrepresentations for transfer, as in multi-task learning. In this work, wecontribute by providing a series of empirical and theoretical studies, anddiscover several interesting yet previously unknown properties of thealgorithm. We find MAML adapts better with a deep architecture even if thetasks need only a shallow one (and thus, no representation learning is needed).While echoing previous findings by others that the bottom layers in deeparchitectures enable representation learning, we also find that upper layersenable fast adaptation by being meta-learned to perform adaptive gradientupdate when generalizing to new tasks. Motivated by these findings, we studyseveral meta-optimization approaches and propose a new one for learning tooptimize adaptively. Those approaches attain stronger performance inmeta-learning both shallower and deeper architectures than MAML.

When MAML Can Adapt Fast and How to Assist When It Cannot