Scientific Computing & Simulation / Independent exploration
Neural-Surrogate Monte Carlo Collision Simulator
A simplified particle-collision simulator paired with a neural regression surrogate and an explanatory dashboard.
Problem
A surrogate can approximate simulator outputs, but its usefulness depends on the simulator's physical assumptions and evaluation protocol.
Architecture
NumPy-based kinematics and event simulation feed a dataset pipeline. A PyTorch surrogate trains on saved arrays, and a dashboard presents simulations and results.
Implementation
The physics module computes momentum magnitudes and invariant mass and provides a simplified two-body energy calculation. The training script uses MSE, Adam, shuffled minibatches, and a held-out array for evaluation.
Results
VERIFIED: the kinematic functions and surrogate training loop are implemented. No independently reproduced speedup, detector realism, or physical-discovery result is claimed.
Limitations
The module explicitly leaves full frame transformations and detector noise as future work. Clamping final-state energies is not a general conservation proof for forbidden configurations. Training reproducibility and out-of-distribution validation require further work.
Reproducibility
Reviewed at commit 99574db97977 on 2026-11-09. Source inspection establishes implementation structure; repository-authored results are not independent measurements. The upstream application and benchmark suite were not rerun during this review.
Next investigation
Resolve the limitations above, preserve the reviewed commit and exact experiment configuration, and reproduce correctness checks before promoting performance or deployment claims.