Abstract
Summary form only given. The author considers neural networks with N binary neurons and symmetric lth-order synaptic connections, in which m randomly chosen N-bit patterns are stored and retrieved with a small fraction delta of bit errors allowed. The analysis yields rigorous lower bounds for the maximum possible value of l factorial multiplied by alpha , the number of stored bits per distinct synapse: 0. 11 for l equals 2 (compared to 0. 29 as estimated by Hopfield and Amit et al. ), 0. 22 for l equals 3 and 0. 16 for l equals 4.
Original language | English (US) |
---|---|
Pages | 50 |
Number of pages | 1 |
State | Published - 1987 |
ASJC Scopus subject areas
- General Engineering