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

We propose to meta-learn causal structures based on how fast a learner adaptsto new distributions arising from sparse distributional changes, e.g. due tointerventions, actions of agents and other sources of non-stationarities. Weshow that under this assumption, the correct causal structural choices lead tofaster adaptation to modified distributions because the changes areconcentrated in one or just a few mechanisms when the learned knowledge ismodularized appropriately. This leads to sparse expected gradients and a lowereffective number of degrees of freedom needing to be relearned while adaptingto the change. It motivates using the speed of adaptation to a modifieddistribution as a meta-learning objective. We demonstrate how this can be usedto determine the cause-effect relationship between two observed variables. Thedistributional changes do not need to correspond to standard interventions(clamping a variable), and the learner has no direct knowledge of theseinterventions. We show that causal structures can be parameterized viacontinuous variables and learned end-to-end. We then explore how these ideascould be used to also learn an encoder that would map low-level observedvariables to unobserved causal variables leading to faster adaptationout-of-distribution, learning a representation space where one can satisfy theassumptions of independent mechanisms and of small and sparse changes in thesemechanisms due to actions and non-stationarities.

A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms