entropy-of-clicks — Google case

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

Compute Shannon entropy for a probability vector representing user-click outcomes.

Use base-2 logarithms so the answer is in bits. The input must be a non-empty 1-D vector of non-negative values that sums to 1, otherwise ValueError is raised; zero probabilities contribute nothing.

Implement solve(p).

Returns. Return the entropy as a non-negative Python float.

Examples

Example 1

Input:

solve([0.5, 0.5])

Output:

1.0

Example 2

Input:

solve([0.9, 0.1])

Output:

0.468996

Example 3

Input:

solve([1.0, 0.0])

Output:

0.0

Example 4

Input:

solve([0.5, 0.6])

Output: Raises ValueError.