Standard scaler without leakage

The problem, from first principles

Scaling features is one of the easiest places to leak information into a model. If you compute the mean and standard deviation over the full dataset before splitting, the test set has already shaped the training features. The fix is to learn the scaling statistics from the training rows only, then apply those same numbers to the test rows, even when the test rows look different.

Implement standard_scale_no_leakage(X_train, X_test), which returns (scaled_train, scaled_test). Both are scaled with the per-column mean and standard deviation computed from X_train alone.

Constraints

• A column with zero standard deviation in the training data is scaled as if its standard deviation were 1, so it becomes all zeros.
• Inputs are 2-D arrays with the same number of columns.
• Do not modify the inputs.

Hints

Open one at a time. Each gives away a little more than the last.