Why is feature scaling (preprocessing) needed?

When features differ widely in range, models sensitive to distance or coefficient size can be swayed by the large-valued feature. Bringing features onto a comparable scale before modeling reduces that imbalance.

How do I verify the preprocessing result?

Feed in data and check directly: the output array has the same shape as the input, the original is left unchanged, and each feature's statistics land near the intended values.

What libraries does this problem use?

Python and numpy. It uses numpy's default standard deviation (np.std); scikit-learn's StandardScaler is the standard tool that provides the same kind of preprocessing.

feature standardization

Your task

Before feeding data to a model, you often need to bring the features (columns) onto a comparable scale. If one column ranges 0–1 and another ranges 0–1000, the large-valued column alone dominates distances and coefficients, and the other features get drowned out. Complete standardize(X). X is a 2-D float array whose rows are samples and columns are features. Return a new array in which each feature (column) has mean 0 and standard deviation 1. Example X = [[1, 100], [2, 200], [3, 300]] → both columns become the same shape: mean 0, standard deviation 1.

Ground rules

•X is a 2-D float numpy array; rows are samples and columns are features.
•use NumPy's default standard deviation (np.std).

Not required

•training, prediction, or an actual model.
•handling missing values (NaN) or other scalings such as min-max.

▶About this challenge

This challenge asks you to complete standardize(X) in numpy, a preprocessing helper that brings features onto a comparable scale before data goes into a model. The input is a 2-D float array whose rows are samples and columns are features, and the function must return a new array — without mutating the original — where each feature has mean 0 and standard deviation 1. It practices the fundamentals of feature scaling as a preprocessing step.

Real datasets often mix features on very different ranges — one might span 0–1 while another spans 0–1000. When ranges differ this much, models that rely on distances or are sensitive to coefficient size can be dominated by the single large-valued feature. That is why bringing features onto a comparable scale is a common preprocessing step before modeling. This problem implements that step directly in numpy, while practicing the habit of verifying the result by inspecting the actual numbers.

Why is feature scaling (preprocessing) needed?: When features differ widely in range, models sensitive to distance or coefficient size can be swayed by the large-valued feature. Bringing features onto a comparable scale before modeling reduces that imbalance.
How do I verify the preprocessing result?: Feed in data and check directly: the output array has the same shape as the input, the original is left unchanged, and each feature's statistics land near the intended values.