Address the challenge that high energy physicists currently face using ML techniques. Indeed, the extrapolation capabilities of these techniques ranges from impossible to unreliable, in particular when describing the tails of distributions for LHC processes
In order to tackle this extrapolation problem, we intend to propose new analytic descriptions in the ML model (differential equations, partial differential equations…).
Further push the logic of infusing more physics into ML in the LHC domain by incorporating leading or next to leading order information (like a Taylor series…) to catch better description of processes before smearing effect of the detector
Expected Results:
Improved analytic description of the physical processes, through ML-based prediction and incorporating higher order physical effects
Reconstruction of a physical constraint verified by data by matching ML and physics
Implementation in the LHC domain of the most recent approaches of physics-informed ML, solving differential equations and reconstructing a Hamiltonian operator
