9511674 Northmore The processing of information by neurons in the brain depends upon their ability to respond discriminatively to the multitude of signals impinging upon them from sense organs and other neurons. The most important component of a neuron for this function is the branched dendritic tree that collects signals at numerous synapses distributed over its surface membrane. In previous work, silicon microcircuits were developed that modeled neurons with extensively branched dendrites with simple, passive properties. These electronic neurons successfully discriminate particular patterns of input when delivered to the appropriate synapses. The proposed research will build into the dendrites active mechanisms capable of local amplification of signals. This will be achieved by mimicking certain chemical- and voltage-dependent processes occurring at synapses in biological neurons, processes that are thought to be the basis for learning. As more accurate models, these circuits will improve our understanding of information processing by neurons, both individually and in networks. They will also result in artificial neurons of greater processing power that could be harnessed for robotics. ***
