Learning Stochastic Differential Equations With Gaussian Processes Without Gradient Matching
Machine Learning
2018-08-01 v2 Machine Learning
Abstract
We introduce a novel paradigm for learning non-parametric drift and diffusion functions for stochastic differential equation (SDE). The proposed model learns to simulate path distributions that match observations with non-uniform time increments and arbitrary sparseness, which is in contrast with gradient matching that does not optimize simulated responses. We formulate sensitivity equations for learning and demonstrate that our general stochastic distribution optimisation leads to robust and efficient learning of SDE systems.
Cite
@article{arxiv.1807.05748,
title = {Learning Stochastic Differential Equations With Gaussian Processes Without Gradient Matching},
author = {Cagatay Yildiz and Markus Heinonen and Jukka Intosalmi and Henrik Mannerström and Harri Lähdesmäki},
journal= {arXiv preprint arXiv:1807.05748},
year = {2018}
}
Comments
The accepted version of the paper to be presented in 2018 IEEE International Workshop on Machine Learning for Signal Processing