A research team hailing from the Brain-Inspired Intelligence Technology Research Center at Nanjing University has put forth a highly precise analog in-memory computing approach. Leveraging complementary metal-oxide-semiconductor (CMOS) technology, they have successfully developed an analog in-memory computing chip. This chip demonstrates an exceptional root mean square error of just 0.101% during parallel vector-matrix multiplication tasks, thus establishing a new benchmark for accuracy in the realm of analog in-memory computing. The pertinent research outcomes have been duly published in the esteemed journal Science Advances.
