We intend to use Gaussian process regression to predict a set of heart rate time series data from the MIT-BIH dataset (1000 observations as the training set and 500 observations as the test set). We are going to use three sparse Gaussian process methods: SKI/KISS GP, Scalable Kernel Learning by Stochastic Lanczos Expansions, and Constant time predictive distributions on the dataset. We will evaluate above methods with their accuracy and computing efficiency.
