Credit Scoring Model Trainer: Watch Logistic Regression Learn (2D)

This 2D companion turns the case-study's fixed threshold into a model you train yourself: instead of sliding a cutoff over a pre-scored population, you watch a logistic-regression classifier learn the boundary from scratch through full-batch gradient descent, with a live loss curve, a training-accuracy readout and the three learned weights updating every step. Learning rate and L2 regularization are both exposed as sliders, so it is possible to see, directly, why a rate that is too high oscillates instead of converging and why a large regularization term shrinks the decision boundary into a flatter, less confident line.