Towards the end of March 2021, we took part in the first virtual tinyML Summit as both Gold sponsor, and speaker.
Running over five days, the summit brought together senior level technical experts and decision makers, representing the fast-growing tinyML community.
Our first session – ‘Performing inference on binarized neural networks (BNNs) with xcore®.ai’ – was delivered by Senior Technologist, Laszlo Kindrat, in partnership with Plumerai. During the session, Lazslo shared the detail of our specialised instructions for the vector unit of our economical crossover processor, xcore.ai. He went on to provide an overview of our machine learning model deployment toolchain that seamlessly integrates with Larq – a popular open-source framework for training binarized neural networks, before presenting performance benchmarks on image classification models with various combinations of binarized and 8bit quantized layers. We are delighted to share the recording of this session, now available here.
In our second speakership of the Summit, CEO, Mark Lippett explored the power / cost conundrum of tinyML and introduced an improved way of measuring energy consumption in tinyML processors. He also examined the system trade-offs required to minimise energy consumption and lower cost, bringing value to the end user. Please follow this link to watch the session in full.
Presentation decks for both sessions can also be found on the tinyML Summit website.
