Dynamically adding symbolically meaningful nodes to knowledge-based neural networks

Traditional connectionist theory-reenement systems map the dependencies of a domain-speciic rule base into a neural network, and then reene this network using neural learning techniques. Most of these systems, however, lack the ability to reene their network's topology and are thus unable to add new rules to the (reformulated) rule base. Therefore , on domain theories that are lacking rules, generalization is poor, and training can corrupt the original rules, even those that were initially correct. We present TopGen, an extension to the Kbann algorithm, that heuristically searches for possible expansions to Kbann's network. TopGen does this by dynamically adding hidden nodes to the neural representation of the domain theory, in a manner analogous to adding rules and conjuncts to the symbolic rule base. Experiments indicate that our method is able to heuristically nd eeective places to add nodes to the knowledge bases of four real-world problems, as well as an artiicial chess domain. The experiments also verify that new nodes must be added in an intelligent manner. Our algorithm showed statistically signiicant improvements over Kbann in all ve domains. network-growing algorithm theory reenement the Kbann algorithm computational biology Submitted to the journal Knowledge-Based Systems.