Module 4 · Lesson 1
Aggregations & the axis Parameter
np.sum, np.mean, np.std, np.min, np.max all reduce an array to
fewer numbers. Called with no axis, they collapse the whole array to
a single scalar. With axis=0 they reduce along rows — one result per
column. With axis=1 they reduce along columns — one result per
row.
| Call | Meaning |
|---|---|
arr.sum() | one total, whole array |
arr.sum(axis=0) | one total per column (walks down each column) |
arr.sum(axis=1) | one total per row (walks across each row) |
Pass keepdims=True to keep the reduced axis as length-1 instead of
dropping it — handy when you want the result to broadcast back against the
original array.
grid = np.array([[1, 2, 3],
[4, 5, 6]])
print("whole array sum:", grid.sum())
print("column sums (axis=0):", grid.sum(axis=0))
print("row sums (axis=1): ", grid.sum(axis=1))
print("row means, keepdims: ", grid.mean(axis=1, keepdims=True).ravel())
# keepdims=True lets you broadcast the result back against the original:
centered = grid - grid.mean(axis=1, keepdims=True)
print(centered)Output:
whole array sum: 21
column sums (axis=0): [5 7 9]
row sums (axis=1): [ 6 15]
row means, keepdims: [2. 5.]
[[-1. 0. 1.]
[-1. 0. 1.]]Column-normalize a matrix
Write normalize_columns(grid) that returns a copy of grid where each
column has been divided by that column's sum, so every column sums to
1.0. grid is a 2-D array of positive floats.
💡 Hint
Divide grid by grid.sum(axis=0) — broadcasting lines up a length-n_cols array against each row automatically.
✅ Solution
def normalize_columns(grid):
return grid / grid.sum(axis=0)
Row-wise range
Write row_ranges(grid) that returns a 1-D array with, for each row, the
difference between that row's max and min value (max - min).
💡 Hint
grid.max(axis=1) - grid.min(axis=1).
✅ Solution
def row_ranges(grid):
return grid.max(axis=1) - grid.min(axis=1)