Scalable Gaussian Process Classification via Expectation Propagation
This work addresses scalability issues in Gaussian process classification for practitioners dealing with large datasets, though it is incremental as it offers an alternative to existing variational methods.
The paper tackles the problem of scaling Gaussian process classification to large datasets by proposing an expectation propagation method that handles millions of data instances, with experiments showing it is competitive with variational approaches.
Variational methods have been recently considered for scaling the training process of Gaussian process classifiers to large datasets. As an alternative, we describe here how to train these classifiers efficiently using expectation propagation. The proposed method allows for handling datasets with millions of data instances. More precisely, it can be used for (i) training in a distributed fashion where the data instances are sent to different nodes in which the required computations are carried out, and for (ii) maximizing an estimate of the marginal likelihood using a stochastic approximation of the gradient. Several experiments indicate that the method described is competitive with the variational approach.