AutoBayes: Automated Bayesian Graph Exploration for Nuisance- Robust Inference

Abstract

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. Auto-Bayes 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.

EEG-GNN: Graph Neural Networks for Classification of Electroencephalogram (EEG) Signals

“…Compared to a standard CNN classifier of similar model size, GNN models significantly improve the accuracy by 2.0% for ErrP, whereas the improvement for RSVP is 0.4%. Compared to AutoBayes classifiers [15], which detect the conditional relationship between data features, task labels, nuisance variation labels (subject IDs), and potential latent variables in DNN architectures to identify the best inference strategy, GNN models still have higher classification performance, while reducing model size more than 20x. Although GNN models don't take advantage from adversarial learning using variations in subject IDs, they perform 0.8% better for ErrP, and nearly same for RSVP.…”