The reference neuron model of memory: Explorations of its capabilities

Connectionist models are computer simulations based on highly These simulations are an attempt to explain the sensory input, perception processing, memory and learning abilities of the brain. Unfortunately, by reducing the many layers of neurological structure present in the human brain, connectionist models eliminate some of the inherent information processing of the real system. For example, most connectionist models cannot represent and recall a scene from the past, visualize a face when an associated name is mentioned or understand mathematical formulas as a sequence of steps.; The Reference Neuron Model (RNM) is an example of an advanced model of neurological structure. The model is based on the principle of superposition-free memory, i.e., the requirement that the acquisition of new memories by a neural network does not degrade or hybridize the previously acquired memories. The RNM is a hybrid system that has both the features found in procedural systems as well as the distributed pattern recognition capabilities found in connectionist neural network approaches. The memory manipulation mechanisms inherent in the model support the development of temporal as well as associative memory structures through trial and error learning. The results of which contribute to the development of knowledge by the model about its environment.; Earlier work demonstrated that the RNM was capable of learning to play a game Tic-tac-toe with a reasonable level of performance. The previous work restricted experiments to 3x3 board games and used only content addressable memory. In this research, the size of the search space is expanded to increase the difficulty of the problem and enhanced memory techniques namely temporal, associative and artificial imagination memories are introduced.; The results demonstrate that with the additional memory techniques the RNM is a scalable approach and is capable of learning winning strategies when pitted against a Random Player, as well as, the ability to play to a draw versus a Minimax player.