< Back to previous page

Project

Learning directed probabilistic logical models from incomplete data.

In the field of machine learning and data mining there is an increasing interest in probabilistic logical models. This is because such models have two advantages: their ability to deal with relational data (by means of the logic-component), and their ability to deal with noisy data (by means of the probabilistic component). In this work we will focus on a particular kind of probabilistic logical models, namely directed models based on the concept of conditional probability (since such models are easy to interpret). There already exist a number of methods for learning such models from data. However, usually it is assumed that the data is complete (that is, that hte values of all relevant random variables are observed). The aim of this project is to extend the existing methods for learning probabilistic logical models to incomplete data.
Date:29 Dec 2008 →  30 Dec 2009
Keywords:Data mining, Machine learning, Artificial intelligence
Disciplines:Artificial intelligence, Cognitive science and intelligent systems