Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Modeling Continuous Stochastic Processes with Dynamic Normalizing Flows
(Submitted on 24 Feb 2020 (v1), last revised 13 Jul 2021 (this version, v4))
Abstract: Normalizing flows transform a simple base distribution into a complex target distribution and have proved to be powerful models for data generation and density estimation. In this work, we propose a novel type of normalizing flow driven by a differential deformation of the Wiener process. As a result, we obtain a rich time series model whose observable process inherits many of the appealing properties of its base process, such as efficient computation of likelihoods and marginals. Furthermore, our continuous treatment provides a natural framework for irregular time series with an independent arrival process, including straightforward interpolation. We illustrate the desirable properties of the proposed model on popular stochastic processes and demonstrate its superior flexibility to variational RNN and latent ODE baselines in a series of experiments on synthetic and real-world data.
Submission history
From: Ruizhi Deng [view email][v1] Mon, 24 Feb 2020 20:13:43 GMT (6755kb,D)
[v2] Thu, 16 Jul 2020 00:38:32 GMT (7962kb,D)
[v3] Mon, 26 Oct 2020 21:02:05 GMT (12130kb,D)
[v4] Tue, 13 Jul 2021 04:10:23 GMT (12131kb,D)
Link back to: arXiv, form interface, contact.