Pick k Honestly

Choosing k by checking scores on the test rows lets the test set quietly teach the model. Instead, carve a validation slice out of the training rows, choose k there, and only then look at the test rows once.

Write honest_k(ks) for the breast cancer data.

• Load the data with load_breast_cancer(return_X_y=True, as_frame=True) and split it with train_test_split(X, y, test_size=0.25, random_state=0, stratify=y).
• Split the training rows again with train_test_split(X_train, y_train, test_size=0.25, random_state=0, stratify=y_train). Call the parts fit rows and validation rows.
• For each k in ks, fit make_pipeline(StandardScaler(), KNeighborsClassifier(n_neighbors=k)) on the fit rows and score it on the validation rows.
• Choose the k with the highest validation score, comparing scores rounded to 3 places. On a tie, choose the smallest k.
• Refit a fresh pipeline with that k on the whole training set, and score it on the test rows.

Return a dict:

• "k": the chosen k, as an int.
• "validation": its validation score, rounded to 3 places.
• "test": the refitted model's test score, rounded to 3 places.