pub fn accuracy_quantized( rff: &Rff, w: &[f64], x: &[Vec<f64>], y: &[f64], bits: u32, ) -> f64
Accuracy using bits-bit qFHRR features (the edge-deployable path).
bits