An analog online clustering circuit in 130nm CMOS

An analog clustering circuit is presented. It is capable of inferring the underlying pattern and extracting the statistical parameters from the input vectors, as well as providing measures of similarity based on both mean and variance. A floating-gate analog memory provides non-volatile storage. A current-mode distance computation, a time-domain loser-take-all and a memory adaptation circuit implement efficient and robust learning algorithm. We show that our analog computation element can achieve more than 10x higher energy efficiency than its digital counterpart. An 8-dimension 4-centroid prototype was fabricated in a 130 nm standard CMOS process. Measurement results demonstrate vector classification at 16 kHz, and unsupervised online clustering at 4 kHz with a power consumption of 15 μW.