Low Power and Efficient Re-Configurable Multiplier for Accelerator

Abstract

Deep learning is a rising topic at the edge of technology, with applications in many areas of our lives, including object detection, speech recognition, natural language processing, and more. Deep learning's advantages of high accuracy, speed, and flexibility are now being used in practically all major sciences and technologies. As a result, any efforts to improve the performance of related techniques are worthwhile. We always have a tendency to generate data faster than we can analyse, comprehend, transfer, and reconstruct it. Demanding data-intensive applications such as Big Data. Deep Learning, Machine Learning (ML), the Internet of Things (IoT), and high- speed computing are driving the demand for "accelerators" to offload work from general-purpose CPUs. An accelerator (a hardware device) works in tandem with the CPU server to improve data processing speed and performance. There are a variety of off-the-shelf accelerator architectures available, including GPU, ASIC, and FPGA architectures. So, this work focus on designing a multiplier unit for the accelerators. This increases the performance of DNN, reduced the area and increasing the training speed of the system.