Neural Jump Stochastic Differential Equations

05/24/2019
by   Junteng Jia, et al.
0

Many time series can be effectively modeled with a combination of continuous flows along with random jumps sparked by discrete events. However, we usually do not have the equation of motion describing the flows, or how they are affected by jumps. To this end, we introduce Neural Jump Stochastic Differential Equations that provide a data-driven approach to learn continuous and discrete dynamic behavior, i.e., hybrid systems that both flow and jump. Our approach extends the framework of Neural Ordinary Differential Equations with a stochastic process term that models discrete events. We then model temporal point processes with a piecewise-continuous latent trajectory, where stochastic events cause an abrupt change in the latent variables. We demonstrate the predictive capabilities of our model on a range of synthetic and real-world marked point process datasets, including classical point processes such as Hawkes processes, medical records, awards on Stack Overflow, and earthquake monitoring.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
08/22/2022

Latent Neural Stochastic Differential Equations for Change Point Detection

The purpose of change point detection algorithms is to locate an abrupt ...
research
10/25/2020

ImitationFlow: Learning Deep Stable Stochastic Dynamic Systems by Normalizing Flows

We introduce ImitationFlow, a novel Deep generative model that allows le...
research
10/04/2019

Discrete Processes and their Continuous Limits

The possibility that a discrete process can be fruitfully approximated b...
research
02/24/2020

Modeling Continuous Stochastic Processes with Dynamic Normalizing Flows

Normalizing flows transform a simple base distribution into a complex ta...
research
06/08/2020

Theoretical Guarantees for Learning Conditional Expectation using Controlled ODE-RNN

Continuous stochastic processes are widely used to model time series tha...
research
02/26/2023

Invariant Neural Ordinary Differential Equations

Latent neural ordinary differential equations have been proven useful fo...
research
08/18/2022

Challenges and opportunities in applying Neural Temporal Point Processes to large scale industry data

In this work, we identify open research opportunities in applying Neural...

Please sign up or login with your details

Forgot password? Click here to reset