DER/SPRC/LEPh - Parameter reduction on nuclear mass models

Goal: investigate statistical parameter-reduction methods on the DZ33 nuclear mass model to handle poorly constrained/redundant parameters and improve extrapolative predictive power.

• Context: calibration against observables may leave parameters sloppy; need to identify/remove or reduce influence of weakly constrained ones.
• Tasks: develop a simple Python package to test PCA, SVD, Ridge regression and related methods for dimensionality reduction; evaluate accuracy vs complexity; prepare methodology for extension to more complex LEPh nuclear-structure models.
• Tools: Python; basic statistics knowledge.
• Outcome: insights on parameter relevance, reduced models preserving predictive accuracy.