NumPy: Standardise each column

Many models need every feature on the same scale. Compute (X − X.mean(axis=0)) / X.std(axis=0) for a small data matrix.

What to build

A program that standardises every column of a matrix.

Requirements

  • Print means followed by the column means, each to two decimals, separated by spaces.
  • Then print each row of the standardised matrix, values to two decimals separated by spaces.
  • Use the population standard deviation. Where a column's standard deviation is 0, print 0.00 for its values (NumPy would give nan with a warning).

Input

Line 1: r and c. Then r lines of c numbers.

Output

1 + r lines.

Constraints

2 ≤ r ≤ 100; 1 ≤ c ≤ 10.

Example

Input:

3 2
1 10
2 20
3 30

Output:

means 2.00 20.00
-1.22 -1.22
0.00 0.00
1.22 1.22

Why it works

Subtracting the column mean centres each feature at 0 and dividing by its standard deviation sets its spread to 1, which is what broadcasting the two axis-0 reductions across the rows does.

Examples

Example 1 · Input
3 2
1 10
2 20
3 30
Expected output
means 2.00 20.00
-1.22 -1.22
0.00 0.00
1.22 1.22
Example 2 · Input
2 1
0
4
Expected output
means 2.00
-1.00
1.00

Prompts inside input() are ignored when checking, so input("n: ") is fine.

Hints

Hint 1

Work column by column: zip(*rows) gives the columns.

Hint 2

Use the population standard deviation (ddof=0), NumPy's default.

Hint 3

Rebuild the rows from the standardised columns with zip again.

Solution

Show one correct solution (try it yourself first)
r, c = map(int, input().split())
rows = [[float(x) for x in input().split()] for _ in range(r)]
cols = [list(col) for col in zip(*rows)]
means = [sum(col) / r for col in cols]
stds = [(sum((x - m) ** 2 for x in col) / r) ** 0.5 for col, m in zip(cols, means)]
print("means", " ".join(f"{m:.2f}" for m in means))
for row in rows:
    print(" ".join(f"{(x - m) / s:.2f}" if s else "0.00" for x, m, s in zip(row, means, stds)))

Lesson: Statistics with NumPy →