Always-on sensors are increasingly expected to embark a variety of tiny neural networks and to continuously perform inference on time-series of the data they sense. In order to fit lifetime and energy consumption requirements when operating on battery, such hardware uses microcontrollers (MCUs) with tiny memory budget e.g., 128kB of RAM. In this context, optimizing data flows across neural network layers becomes crucial. In this paper, we introduce TinyDéjàVu, a new framework and novel algorithms we designed to drastically reduce the RAM footprint required by inference using various tiny ML models for sensor data time-series on typical microcontroller hardware. We publish the implementation of TinyDéjàVu as open source, and we perform reproducible benchmarks on hardware. We show that TinyDéjàVu can save more than 60% of RAM usage and eliminate up to 90% of redundant compute on overlapping sliding window inputs.

I've been fortunate enough to attend the first meeting of what then became the Trusted IoT consortium, held in Berkeley in late 2016. The key idea was, combining properties of blockchain such as identity and immutable ledger with sensor data where provenance, authenticity, location, and governance are crucial for actions taken on the basis of the sensor data stream. It still takes an effort to wrap your mind around the way the two areas, Blockchain and IoT, interoperate. Various objects need to be ascertained as being something, belonging to someone, and being somewhere. One example is Supply Chain -- when a container is shipped from say Shanghai to Oakland, the buyer, the seller and the shipper all need to know, and to be able to prove, that it follows a certain state machine -- with loading, sailing, arrival, customs, and unloading, sequenced one after the other in time.