ABSTRACT
Since the resurgence of interest in neural networks in the mid-80’s, the possibility to profit from the respective advantages of the symbolic and neural paradigms in integrated connectionist-symbolic systems has been a persistent research goal. These systems all strive to achieve both symbolic and neural functionalities. Combining the advantages of both paradigms has a clear benefit: human cognitive capabilities are not within the reach of either symbolic or connectionist artificial intelligence because of the limitations of each approach. Combining their powerful properties could be a better way to model these fundamental capabilities.
