• Ported an Online Machine Learning model from Python to Rust, achieving a 97% reduction in execution time and attaining a processing capability of 10,000 records per second for IoT devices.
  • Optimized memory usage to a minimal 20MB, facilitating deployment on robots with stringent memory limits and resulting in a 55% enhancement in positioning accuracy.
  • Specialized implementation of Mondrian Forests, streamlining the model for real-time learning in performance-sensitive applications.

Thesis — Paper

📜 Twente Publication 📜 Aalto Publication (mirror)
📜 Download PDF (mirror)

Presentation

🎤 Slides

GitHub

🧑‍💻️ LightRiver-Cache

🧑‍💻️ LightRiver