On the Combination of Information-Theoretic Kernels with Generative Embeddings

lassical methods to obtain classifiers for structured objects (e.g., sequences, images) are based on generative models and adopt a classical generative Bayesian framework. To embrace discriminative approaches (namely, support vector machines), the objects have to be mapped/embedded onto a Hilbert space; one way that has been proposed to carry out such an embedding is via generative models (maybe learned from data). This type of hybrid discriminative/generative approach has been recently shown to outperform classifiers obtained directly from the generative model upon which the embedding is built.Discriminative approaches based on generative embeddings involve two key components: a generative model used to define the embedding; a discriminative learning algorithms to obtain a (maybe kernel) classifier. The literature on generative embedding is essentially focused on defining the embedding, and some standard off-the-shelf kernel and learning algorithm are usually adopted. Recently, we have proposed a different approach that exploits the probabilistic nature of generative embeddings, by using information-theoretic kernels defined on probability distributions. In this chapter, we review this approach and its building blocks. We illustrate the performance of this approach on two medical applications