AIPHY General searches with (GNN) AutoEncoders
Objectives:
- Develop general Graph Neural Network (GNN) reconstruction algorithm(s) for all event types in experiment.
- Based on this GNN architecture, use Variable AutoEncoder (VAE) to encode the dataset in a latent space (LS).
- Apply the VAE to simulated data, also including theoretically motivated but unobserved physics signatures.
- Compare real and simulated data LS overlaps, and search for populations outside known physics regions.
- VAE shifts between real and simulated data of known physics can also be used to calibrate simulation
Expected Results:
- Ultrafast and highly performant classification and reconstruction algorithm of events in irregular detectors.
- Innovative new method for general physics searches, but also applicable to other sciences, medicine and industry.
- General and highly automated method for calibration of simulated data to better match real data.
- World leading search potential for several highly motivated physics theories (Dark Matter, Monopoles, etc.).
