Meta-Learning Conjugate Priors for Few-Shot Bayesian Optimization
This work is significant for researchers and practitioners using Bayesian Optimization, particularly in few-shot scenarios where obtaining unbiased and informative priors is a bottleneck.
The paper addresses the challenge of finding informative priors for Bayesian Optimization, especially when data is scarce. It proposes a meta-learning approach to automate the estimation of conjugate prior distributions, enabling accurate estimation of distribution shape parameters with limited data.
Bayesian Optimization is methodology used in statistical modelling that utilizes a Gaussian process prior distribution to iteratively update a posterior distribution towards the true distribution of the data. Finding unbiased informative priors to sample from is challenging and can greatly influence the outcome on the posterior distribution if only few data are available. In this paper we propose a novel approach to utilize meta-learning to automate the estimation of informative conjugate prior distributions given a distribution class. From this process we generate priors that require only few data to estimate the shape parameters of the original distribution of the data.