MetaLearnerOptimizerUse

Enums · AiDotNet.MetaLearning

Which of the supplied optimizers a meta-learning algorithm’s updates go through.

For Beginners

Most meta-learners have an outer loop (learning the starting point shared by all tasks) and some also have an inner loop (adapting to one task). An optimizer can only be plugged into a loop the algorithm actually runs with gradient steps.

How It Works

An algorithm declares this so that an optimizer it has no step for is rejected when the meta-learner is built, instead of being accepted and silently never used.

Fields

Field Summary
MetaLearnerOptimizerUse Inner The inner (task adaptation) step goes through the inner optimizer.
MetaLearnerOptimizerUse InnerAndOuter Both loops go through their optimizers.
MetaLearnerOptimizerUse None Neither loop takes plain gradient steps; the algorithm’s own update rule is part of the method.
MetaLearnerOptimizerUse Outer The outer (meta-parameter) step goes through the meta-optimizer.
int value__