| IP개요 |
In this paper, we present a low power embedded AI system for machine learning. The embedded AI system is based on k-Nearest Neighbor (k-NN) algorithm. The AI controller adopts more efficient finite state machine (FSM) optimized with instructions to generate the concise instructions. Therefore, the embedded AI system trains dataset in AI cores rapidly and classifies test data with low power consumption. To measure the performance of the embedded AI system, we have implemented on a field-programmable gate array and used memory intellectual property (IP) in AI cores. We will fabricate the low power embedded AI system for machine learning with Samsung 28 um RFCMOS 1-poly 8-metal technology. The embedded AI system operates in 1.0V, 50MHz frequency and the circuit is digital type. |
