WP 3 - Exact and approximate inference


OBJECTIVES

  • The methodological development of scalable algorithms for approximate inference in hybrid Bayesian networks and dynamic models
  • The implementation of the developed models in the AMIDST toolbox
  • To provide the AMIDST toolbox with the functionality necessary to be able to react to the continuous and massive arrival of data
 

PLANNED WORK

In what concerns inference in Hybrid models the work will focus on static domains involving both discrete and continuous variables.

Different approximation techniques will be explored in order to develop an approximate inference algorithm. Scalability will be approached from the point of view of organizing the required calculations so that they can be distributed among different computing devices when available. Scalable methods for doing MAP-based inference in hybrid models will be developed.

WP3 also deals with the design of an inference scheme in dynamic domains, adapted to parallel architectures and able to operate in hybrid domains.
 


This website uses cookies. Cookies are only used for handling login and identifying registrered users, not for visitor tracking.