Online Learning of Moving Gaussians

Manipulate moving gaussians in real-time and see how an online learning algorithm reacts.

In non-stationary learning problems, the target distribution changes over time, and our learned model must adapt. In this example, toy data is generated from two gaussian distributions. The learning algorithm receives data sequentially for training. In real-time, the user can manipulate the parameters of the target distribution, and see the learning algorithm react. An implementation of Online Gaussian Naive Bayes in included, with a forward-weighted option to facilitate adaptation. Other algorithms can be plugged in by conforming to a simple interface.