A neural net for the recall of sequences
Rand Corp.; available from the Defense Documentation Center, 1963 - Artificial intelligence - 23 pages
The net discussed here is based on the requirement that all the elements in a sequence, presented as stimuli during a short interval of time, must remain present as a serial memory trace and be available for serial recall in the order given. Time may be represented by space storage along a chain of neurons. Consequently, a mechanism equivalent to scanning is needed to lay down the trace and to replay the sequence. (Author).
What people are saying - Write a review
We haven't found any reviews in the usual places.
670 MD Playback Cell Firing Learning central nervous system computations connections MAJJ continue firing continue to fire DATA FOR RUN facilitation inputs facilitation on JJ fire JJ fires three firing of SX firings of MB forgetting inactive inhibit input cells J. W. Smith JJ fires JJ will fire knob is sufficient l0 milliseconds l0 mv learning knob learning synapses MAJJ0l MD Playback MDJJ0l MEMORANDUM RM-3872-PR DECEMBER motor neuron MB Neural number of cell number of firings P. M. Milner Phase Time Interval play back Playback l 390 PROJECT RAND promotes either serial RAND Corporation RECALL OF SEQUENCES recall phase rectangular pulse resting potential rH CM rH rH S2 and Tl SANTA MONICA SB stops SB to MB scanning chain self-learning knob serial learning square wave starting S2 stopping S2 sufficient to fire SX is firing synaptic knob telephone number three firings threshold function threshold value United States Air