missing-value-mean-imputation — Microsoft case

The team needs a reliable implementation for feature cleaning in a production-oriented ML workflow. A small implementation error can change the model input or metric enough to make a downstream result misleading.

Replace NaN entries in each feature column with that column's mean computed from the observed values.

Process each column independently: compute the mean of its observed (non-NaN) values and put it in place of every NaN in that column. A column with no observed value stays NaN.

Implement solve(X).

Returns. Return a new float NumPy matrix of the same shape. The input is not modified.

Examples

Example 1

Input:

solve([[1.0, float('nan')], [3.0, 5.0]])

Output:

[[1.0, 5.0], [3.0, 5.0]]

Example 2

Input:

solve([[1.0, 2.0], [float('nan'), 4.0], [5.0, float('nan')]])

Output:

[[1.0, 2.0], [3.0, 4.0], [5.0, 3.0]]