Uncertainty quantification for appliance recognition in non-intrusive load monitoring using Bayesian deep learning

Non-Intrusive Load Monitoring (NILM) can be used to detect, recognize, and classify switching events of individual electrical appliances from an aggregate power signal that is measured at the main line grid connection. A limitation existing solutions deep learning models tend overfit data do not express their uncertainty when making predictions. This paper shows information obtained in a natural way by use Bayesian Neural Networks. Having this very valuable, because it supplies relevant about potential misclassifications model end-user. The source these attributed ambiguous data, or requiring more examples learn from. In work, increase generalization performance observed Stochastic Gradient Hamiltonian Monte Carlo over descent, usefulness NILM context discussed.