TY - JOUR
T1 - Training a Network with Ternary Weights Using the CHIR Algorithm
AU - Abramson, S.
AU - Marom, E.
PY - 1993/11
Y1 - 1993/11
N2 - A modification of the binary weight CHIR algorithm is presented, whereby a zero state is added to the possible binary weight states. This method allows obtaining solutions with reduced connectivity, by offering disconnections in addition to the excitatory and inhibitory connections. The algorithm was examined via extensive computer simulations for the restricted cases of parity, symmetry, and teacher problems, which show similar convergence rates to those presented for the binary CHIR2 algorithm, however, with reduced connectivity. Moreover, this method expands the set of problems solvable via the binary weight network configuration with no additional parameter requirements.
AB - A modification of the binary weight CHIR algorithm is presented, whereby a zero state is added to the possible binary weight states. This method allows obtaining solutions with reduced connectivity, by offering disconnections in addition to the excitatory and inhibitory connections. The algorithm was examined via extensive computer simulations for the restricted cases of parity, symmetry, and teacher problems, which show similar convergence rates to those presented for the binary CHIR2 algorithm, however, with reduced connectivity. Moreover, this method expands the set of problems solvable via the binary weight network configuration with no additional parameter requirements.
UR - http://www.scopus.com/inward/record.url?scp=0027696208&partnerID=8YFLogxK
U2 - 10.1109/72.286901
DO - 10.1109/72.286901
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:0027696208
SN - 1045-9227
VL - 4
SP - 997
EP - 1000
JO - IEEE Transactions on Neural Networks
JF - IEEE Transactions on Neural Networks
IS - 6
ER -