Abstract | ||
---|---|---|
The kernel exponential family is a rich class of distributions,which can be fit efficiently and with statistical guarantees by score matching. Being required to choose a priori a simple kernel such as the Gaussian, however, limits its practical applicability. We provide a scheme for learning a kernel parameterized by a deep network, which can find complex location-dependent local features of the data geometry. This gives a very rich class of density models, capable of fitting complex structures on moderate-dimensional problems. Compared to deep density models fit via maximum likelihood, our approach provides a complementary set of strengths and tradeoffs: in empirical studies, the former can yield higher likelihoods, whereas the latter gives better estimates of the gradient of the log density, the score, which describes the distributionu0027s shape. |
Year | Venue | Field |
---|---|---|
2018 | international conference on machine learning | Kernel (linear algebra),Applied mathematics,Parameterized complexity,Mathematical optimization,A priori and a posteriori,Exponential family,Maximum likelihood,Gaussian,Mathematics,Empirical research |
DocType | Volume | Citations |
Journal | abs/1811.08357 | 1 |
PageRank | References | Authors |
0.35 | 0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Wenliang Li | 1 | 1 | 1.36 |
Dougal J. Sutherland | 2 | 53 | 6.76 |
Heiko Strathmann | 3 | 82 | 5.84 |
Arthur Gretton | 4 | 3638 | 226.18 |