Abstract: There is increasingly interest in developing embedded machine learning hardware as it can offer better performance in terms of privacy, bandwidth efficiency, and scalability.