WP 4 - Learning AMIDST models from data


OBJECTIVES

  • To develop algorithms for automatically learning AMIDST models from data streams. The algorithms should support a seamless integration of data and domain knowledge provided by system experts.
  • The development of the learning algorithm will be closely linked to the inference algorithms developed in WP 3 as inference will play a key role in the learning algorithms.
  • The algorithmic developments in this work package will be put into a distributed context, where parallel algorithms can be used to exploit multiple cores or CPUs. This is done to provide additional support for the scalability of the algorithms.
 

PLANNED WORK

This work package looks at how to learn models from the AMIDST framework from data. Firstly, dedicated learning algorithms for learning the structure of the models will be defined, taking advantage of the parallel architectures of modern computers as well as distributed computer network architectures.

Next, we will consider how to learn the parameterization of fixed structure models, where both batch-learning and online adaption will be considered.

Further, we will also delve into how to solve the feature-selection problem when learning from streaming data. 
 


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