An Incremental Algorithm for Tree-shaped Bayesian Network Learning

Incremental learning is a very important approach to learning when data is presented in short chunks of instances. In such situations, there is an obvious need for improving the performance and accuracy of knowledge representations or data models as new data is available. It would be too costly, in computing time and memory space, to use the batch algorithms processing again the old data together with the new one. We present in this paper an incremental algorithm for learning tree-shaped Bayesian Networks. We propose a heuristic able to trigger the updating process when data invalidates, in some sense, the current structure. The algorithm rebuilds the network structure from the branch which it is found to be invalidated. We will experimentally demonstrate that the heuristic is able to obtain almost optimal tree-shaped Bayesian Networks while saving computing time.