What changed
We retrained the 2-M16 model with updated class weights, increasing the REST weight from 0.2 to 0.8. The goal was to reduce REST confusion without sacrificing finger separation on non-REST windows.
Before vs after (2-M16)
| Metric | Before (REST=0.2, 2026-02-19) | After (REST=0.8, 2026-02-24) | Δ |
|---|---|---|---|
| Test action accuracy | 89.31% | 89.71% | +0.39 pp |
| Test finger accuracy on non-REST windows | 89.90% | 90.38% | +0.48 pp |
| Train avg loss | 0.5174 | 0.5020 | -0.0154 |
| Test windows | 2,040 | 2,040 | 0 |
| Test non-REST windows | 1,871 | 1,871 | 0 |
Notes
- The reweighting produced small but consistent gains in both test action accuracy and finger accuracy on non-REST windows.
- We will continue exploring class-weighting and calibration to improve REST vs movement separation without increasing false positives.