Machine learning on 1 cm2 - Tweakers Dev Summit
2 likes•449 views
This document discusses machine learning on microcontrollers. It notes that microcontrollers are small, cheap, efficient but slow with limited memory. It then covers reinforcement learning using a board game as an example. It discusses how machine learning can be done efficiently on microcontrollers through techniques like classification rather than training. Examples shown include keyword spotting and object detection on video, achieving inferences per second while using limited RAM and battery life over a year. The document encourages people to get started with machine learning on microcontrollers by running tutorials and examples on development boards.
Machine learning on 1 cm2 - Tweakers Dev Summit
- 1. Machine learning on one one square centimeter Jan Jongboom, Arm Tweakers Dev Summit 14 February 2019
- 11. 11 Microcontrollers Small (1cm²) Cheap (~1$) Efficient (years on ba=ery) Slow (max. 100 MHz) Limited memory (max. 256K RAM) Downsides 8 cm
- 12. 12 Reinforcement learning X X O O Content of lucifer box Every box resembles a state of the board
- 37. 37 Battery life 1 coin cell 1 image every 5 minutes 100 ms. per inference > 1 year of battery life
- 40. 40 Recap 1. Machine learning is cool 2. Machine learning on the edge is even cooler 3. We're just getting started, so get hacking 4. ??? 5. PROFIT!!!