Quantum Feature Encoding

The feature weighting mechanism is differentiable and soft rather than a hard thresholding or position-based weight-sharing method. Importance-aware weighting uses learnable Ry and Rz rotations to scale feature amplitude and phase contributions. The quantum feature encoding maps each feature to an Ry rotation on the zero state. Selected slices are resized to 16 by 16, vectorized into 256 features, and Z-score normalized before PCA. PCA reduces dimensionality while preserving 95% explained variance under an upper qubit limit. PCA-reduced features are scaled to the interval from negative pi to pi for compatibility with rotation gates.