Machine learning classification of quantum coherence quality in two-qubit systems using experimentally accessible features from a generated subset of the QDataSet.
Develops ML models to classify quantum coherence into three bands (High, Medium, Low) based on observable metrics: state fidelity, Z-expectation change, transverse coherence, and computational basis populations. Compares logistic regression and MLP approaches using Leave-One-Group-Out validation across different physical configurations.
- Strong generalisation across control waveforms and distortion conditions (>95% accuracy)
- Asymmetric transferability between Hamiltonian complexities - models trained on complex systems generalise to simpler ones, but not vice versa
- Medium coherence represents transient states that are challenging to classify consistently
- Fidelity dominates classification decisions, creating inherent circularity in the prediction problem
Completed as coursework for undergraduate degree