| IP개요 |
We present a low power Artificial Intelligence (AI) processor for embedded system. It is based on k-Nearest Neighbor (k-NN) algorithm and Radial Basis Function (RBF) algorithm. In order to reduce the operation workload in calculating distance, this processor adopts L1 (Manhattan distance) instead of L2 (Euclidean distance). By designing the neuro-cell (N-Cell) with registers instead of using memory, the AI processor trains N-Cells rapidly and classifies test data with a less computation and low power consumption. We will fabricate the low power and embedded AI processor with Samsung 65 nm RFCMOS technology. The AI processor operates in 1.8V, 50MHz frequency and the circuit is digital type. |
