Uncertainty-Guided Continual Learning in Bayesian Neural Networks - Extended Abstract

Sayna Ebrahimi, Mohamed Elhoseiny, Trevor Darrell, Marcus Rohrbach; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 75-78

Continual learning aims to learn new tasks without forgetting previously learned ones. This is especially challenging when one cannot access data from previous tasks and when the model has a fixed capacity. Current regularization-based continual learning algorithms need an external representation and extra computation to measure the parameters' importance. In contrast, we propose Bayesian Continual Learning (BCL), where the learning rate adapts according to the uncertainty defined in the probability distribution of the weights in networks. We evaluate our BCL approach extensively on diverse object classification datasets with short and long sequences of tasks and report superior or on-par performance compared to existing approaches. Additionally we show that our model can be task-independent at test time, i.e. it does not presume knowledge of which task a sample belongs to.