Estimation of inhomogeneous point processes: Theory and applications

Wednesday, April 25, 2018 - 9:30am - 10:00am
Keller 3-180
Mark Davenport (Georgia Institute of Technology)
We will present new results concerning the accuracy with which the basis coefficients that define an inhomogeneous Poisson process can be recovered using maximum-likelihood estimation. These results generalize existing guarantees that only hold when observing a Poisson counting process, i.e., a discretized version of the more-general Poisson arrival process. In the process, we remove several restrictive assumptions. We conclude with a brief discussion of applications of these results to more complex models such as Hawkes processes.