Contents I. Introduction and Related Work
The era of the Internet of Things (IoT) is characterized by billions of devices gathering data from sensors and sending them to powerful servers, where analysis and pattern extraction is performed using intelligent algorithms. To save bandwidth and energy across the whole network infrastructure, IoT devices can pre-process sensor data with advanced algorithms like filters and compressors [1], giving birth to so-called edge computing devices. Such algorithms differ in complexity, ranging from requiring only a few integer or fixed-point operations to performing dense kernels using floating-point (FP) arithmetic.