## Abstract

In this paper, we consider an infinite dimensional exponential family P of probability densities, which are parametrized by functions in a reproducing kernel Hilbert space H, and show it to be quite rich in the sense that a broad class of densities on R^{d} can be approximated arbitrarily well in Kullback-Leibler (KL) divergence by elements in P. Motivated by this approximation property, the paper addresses the question of estimating an unknown density po through an element in P. Standard techniques like maximum likelihood estimation (MLE) or pseudo MLE (based on the method of sieves), which are based on minimizing the KL divergence between po and P, do not yield practically useful estimators because of their inability to efficiently handle the log-partition function. We propose an estimator p_{n} based on minimizing the Fisher divergence, J(po||p) between po and p ϵ P, which involves solving a simple finite-dimensional linear system. When po ϵ P, we show that the pro-posed estimator is consistent, and provide a convergence rate of n^{-min} {2/3, 2β+1/2β+2} in Fisher divergence under the smoothness assumption that log po ϵ 72. (C^{β}) for some β ≥ 0, where C is a certain Hilbert-Schmidt operator on H and R(C^{β}) denotes the image of C^{β}. We also investigate the misspecified case of po ϵ P and show that J(po||p_{n}) → inf_{pϵP} J(po||p) as n → ∞, and provide a rate for this convergence under a similar smoothness condition as above. Through numerical simulations we demonstrate that the proposed estimator outperforms the non-parametric kernel density estimator, and that the advantage of the proposed estimator grows as d increases.

Original language | English (US) |
---|---|

Journal | Journal of Machine Learning Research |

Volume | 18 |

State | Published - Jul 1 2017 |

## All Science Journal Classification (ASJC) codes

- Software
- Control and Systems Engineering
- Statistics and Probability
- Artificial Intelligence