Neural network-based framework for extracting Compton Form Factors from particle scattering data. Developed over summer research internship to support nucleon structure research at the University of Virginia Department of Physics.
This tool uses TensorFlow/Keras ANNs to:
- Extract CFFs (ReH, ReE, ReHtilde) from DVCS pseudo-data.
- Optimize network hyperparameters for accuracy across kinematic sets.
- Quantify uncertainties using the Replica Method on Rivanna HPC cluster.
The framework implements physics-informed neural networks that learn the mapping from kinematic variables to Compton Form Factors. The Replica Method generates uncertainty statistics by training multiple network instances on data variations, producing mean predictions and error bands for model evaluation.
Python, TensorFlow/Keras, NumPy, Pandas, Matplotlib, SciPy, PyWavelets, Jupyter, Rivanna HPC
Particle Physics Research Intern | UVA Department of Physics (2024): Developed to optimize ANN architectures for CFF extraction and determine precision metrics for nucleon structure analysis.