
Online gradientbased mixtures for transfer modulation in metalearning
Learningtolearn or metalearning leverages datadriven inductive bias ...
read it

GradientEM Bayesian Metalearning
Bayesian metalearning enables robust and fast adaptation to new tasks w...
read it

Meta Particle Flow for Sequential Bayesian Inference
We present a particle flow realization of Bayes' rule, where an ODEbase...
read it

Learning not to learn: Nature versus nurture in silico
Animals are equipped with a rich innate repertoire of sensory, behaviora...
read it

Deep Interactive Bayesian Reinforcement Learning via MetaLearning
Agents that interact with other agents often do not know a priori what t...
read it

Hierarchical Bayes Modeling for LargeScale Inference
Bayesian modeling is now ubiquitous in problems of largescale inference...
read it

Metatrained agents implement Bayesoptimal agents
Memorybased metalearning is a powerful technique to build agents that ...
read it
Recasting GradientBased MetaLearning as Hierarchical Bayes
Metalearning allows an intelligent agent to leverage prior learning episodes as a basis for quickly improving performance on a novel task. Bayesian hierarchical modeling provides a theoretical framework for formalizing metalearning as inference for a set of parameters that are shared across tasks. Here, we reformulate the modelagnostic metalearning algorithm (MAML) of Finn et al. (2017) as a method for probabilistic inference in a hierarchical Bayesian model. In contrast to prior methods for metalearning via hierarchical Bayes, MAML is naturally applicable to complex function approximators through its use of a scalable gradient descent procedure for posterior inference. Furthermore, the identification of MAML as hierarchical Bayes provides a way to understand the algorithm's operation as a metalearning procedure, as well as an opportunity to make use of computational strategies for efficient inference. We use this opportunity to propose an improvement to the MAML algorithm that makes use of techniques from approximate inference and curvature estimation.
READ FULL TEXT
Comments
There are no comments yet.