TinyML

Machine-learning inference on microcontrollers and similar embedded processors, which typically have tens to hundreds of kilobytes of RAM and often no operating system, within a power budget that the tinyML community puts at around a milliwatt or less. Typical applications are always-on wake-word and keyword spotting, gesture and activity recognition from accelerometers, anomaly detection on vibration sensors, and simple vision such as person detection. Models are small convolutionalconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects. or recurrent networks of thousands to a few million parameters, compressed with 8-bit or lower quantizationquantization, n.: Storing an AI model's numbers with fewer digits, like rounding prices to the nearest dollar, so the model takes less memory and runs faster, usually with only a small loss of accuracy. and pruningpruning, n.: Cutting away the parts of an AI model that contribute least, as a gardener trims branches, so the model becomes smaller or faster while still working nearly as well., often designed by hardware-aware neural architecture searchneural architecture search, n.: Using an automated search, rather than human trial and error, to design the structure of a neural network: which layers to use, how many, and how to connect them., and run by interpreters or compilers such as TensorFlow Lite for Microcontrollers that avoid dynamic memory allocation. TinyML is the extreme end of edge AIedge AI, n.: Doing AI's computing on or near the devices that collect the data, such as cameras, cars, factory machines, phones or local servers, instead of sending everything off to distant data centers., trading capability for cost, privacy and battery life.

The keyword spotter uses 20 KB of RAM and runs for a year on a coin cell.