Machine Learning with Dynamics
SMRTR summary
Physical systems like gyroscopes, springs, and rods can be used as computers by encoding information in their movement. A network of 100 gyroscopes connected by 300 springs, trained using standard machine learning methods, achieved 83.4% accuracy classifying handwritten digits — outperforming a basic linear classifier.
SMRTR provides this summary for quick context. The original article belongs to Hacker News.
Read the original article