Efficient Graph-Based Active Learning with Probit Likelihood via Gaussian Approximations
This work addresses active learning for graph-based SSL, offering incremental improvements in efficiency and function diversity for practitioners in semi-supervised and Bayesian learning domains.
The paper tackles the problem of active learning in graph-based semi-supervised learning under non-Gaussian Bayesian models by developing Gaussian approximations to adapt acquisition functions, resulting in efficient rank-one updates and a novel 'model change' acquisition function.
We present a novel adaptation of active learning to graph-based semi-supervised learning (SSL) under non-Gaussian Bayesian models. We present an approximation of non-Gaussian distributions to adapt previously Gaussian-based acquisition functions to these more general cases. We develop an efficient rank-one update for applying "look-ahead" based methods as well as model retraining. We also introduce a novel "model change" acquisition function based on these approximations that further expands the available collection of active learning acquisition functions for such methods.