AutoBayes: Automated Bayesian Graph Exploration for Nuisance-Robust Inference
This work addresses the problem of nuisance-robust inference for machine learning practitioners, offering an incremental advancement through automated exploration of graphical models.
The authors tackled the challenge of learning data representations that are invariant to nuisance variations by introducing AutoBayes, an automated Bayesian inference framework that explores graphical models to optimize machine learning pipelines, resulting in significant performance improvements through ensemble learning.
Learning data representations that capture task-related features, but are invariant to nuisance variations remains a key challenge in machine learning. We introduce an automated Bayesian inference framework, called AutoBayes, that explores different graphical models linking classifier, encoder, decoder, estimator and adversarial network blocks to optimize nuisance-invariant machine learning pipelines. AutoBayes also enables learning disentangled representations, where the latent variable is split into multiple pieces to impose various relationships with the nuisance variation and task labels. We benchmark the framework on several public datasets, and provide analysis of its capability for subject-transfer learning with/without variational modeling and adversarial training. We demonstrate a significant performance improvement with ensemble learning across explored graphical models.