| CM-1K Neural Network Chip |
Patent parallel architecture, 1024 parallel neurons, Vector data of up to 256 byte, 10 µs learning time (maximum), 10 µs recognition time (maximum), No limit to neuron expansion, Trained by example, Direct digital video recognition with CogniSight stage, RCE (Restricted Coulomb energy), L1 and LSup distance norms, Radial Basis Function (RBF) or K- Nearest Neighbor (KNN) classifier, 500 mW @ 15 MHz, 3.3 V I/O operation 1.2 V core supply, 100-pin TQFP package, 0.13 µM technology - die size 8 x 8 mm |
|
