Learning Declarative Bias (Learning Constraints)

In these papers, we present an approach to learning constraints from results of previous machine learning sessions. The learned constraints can be applied as declarative bias to constrain the space of hypotheses in related learning tasks from the same or similar domains. We empirically evaluated the approach in the context of learning models of dynamic systems from time-series data and process-based knowledge (i.e., process-based modeling).


Ljupco Todorovski Created: January, 2010
Updated: May, 2010