A Reconfigurable Linear RF Analog Processor for Realizing Microwave Artificial Neural Network

Owing to the data explosion and rapid development of artificial intelligence (AI), particularly deep neural networks (DNNs), ever-increasing demand for large-scale matrix-vector multiplication has become one major issues in machine learning (ML). Training evaluating such rely on heavy computational resources, resulting significant system latency power consumption. To overcome these issues, analog computing using optical interferometric-based linear processors recently appeared as promising candidates accelerating lowering On other hand, radio frequency (RF) electromagnetic waves can also exhibit similar advantages counterpart by performing computation at light speed with lower power. Furthermore, RF devices have extra benefits, cost, mature fabrication, analog–digital mixed design simplicity, which great potential realizing affordable, scalable, low latency, power, near-sensor network (RFNN) that may greatly enrich signal processing capability. In this work, we propose a 2 $ \times$ reconfigurable processor theory experiment, be applied matrix multiplier an (ANN). The proposed device utilized realize simple RFNN classification. An 8 notation="LaTeX">$\times $ formed 28 is four-layer ANN Modified National Institute Standards Technology (MNIST) dataset