The anomaly detection implementation originates from PennyLane tutorial and the related research paper Quantum variational rewinding for time series anomaly detection.
The generative model was implemented during the Hackathon following the paper A quantum generative model for multi-dimensional time series using Hamiltonian learning.
Next steps:
- Both of the algorithms share similar elements. Maybe one advantage would be to train a single model for both use cases.
- Encoding graph structure to the diagonal Hamiltonian
- Comparing models to the classical algorithms
- Does the generative model work as a predictor?