We have developed a new floating-gate silicon MOS transistor for analo
g learning applications, The memory storage is nonvolatile; hot-electr
on injection and electron tunneling permit bidirectional memory update
s, Because these updates depend on both the stored memory value and th
e transistor terminal voltages, the synapse can implement a learning f
unction, We have derived a memory-update rule from the physics of the
tunneling and injection processes, and have investigated synapse learn
ing in a prototype array, Unlike conventional EEPROM devices, the syna
pse allows simultaneous memory reading and writing, Synapse transistor
arrays can therefore compute both the array output, and local memory
updates, in parallel, The synapse is small, and typically is operated
at subthreshold current levels; it will permit the development of dens
e, low-power silicon learning systems.