Bayesian Nonparametric Models for Time Evolving Feature Allocations and Sparse Dynamic Networks

=========================================================================
                    Joint  JSS/CSS/CSE Seminar
=========================================================================
Speaker:        Dr. Konstantina Palla
                Department of Statistics
                Oxford University

Title:          "Bayesian Nonparametric Models for Time Evolving Feature
                 Allocations and Sparse Dynamic Networks"

Date:           Monday 23 November 2015

Time:           4:30pm - 5:30pm

Venue:          Lecture Theater F (near lift no. 25/26), HKUST

Abstract:

In this talk, I will present our recent and ongoing work on dynamic models
over discrete time. More specifically I will describe two Bayesian
nonparametric priors; one over feature allocations for sequential data
called the birth-death feature allocation process (BDFP) and one for
time-varying networks.

The BDFP models the evolution of the feature allocation of a set of
objects N across a covariate (e.g. time) by creating and deleting
features. A BDFP is exchangeable, projective, stationary and reversible,
and its equilibrium distribution is given by the Indian buffet process
(IBP). I will also present the de Finetti mixing distribution underlying
the BDFP that plays the role for the BDFP that the Beta process plays for
the Indian buffet process. The utility of this prior is demonstrated on
synthetic and real world data.

The prior over dynamic networks is an ongoing work. To each node of the
network is associated a positive parameter, modelling the sociability of
that nodes. Sociabilities are assumed to evolve over time, and are
modelled via a dynamic point process model. The model is able to (a)
capture smooth evolution of the interactions between nodes, allowing edges
to appear/disappear over time (b) capture long term evolution of the
sociabilities of the nodes (c) and yields sparse graphs, where the number
of edges grows subquadratically with the number of nodes. The evolution of
the sociabilities is described by a tractable time-varying gamma process.

We provide some theoretical insights into the model.