compute_log_expected_power

compute_log_expected_power(k_values: ndarray, powers: ndarray, memory_limit_elements: int = 10000000) → float[source]

Compute \(sum( ln( E[X_i^p_i] ) )\).

Does so efficiently by using sorted batching and vectorization.

Parameters:

  • k_values (ndarray) – shape: (n,) Upper bounds.

  • powers (ndarray) – shape: (n,) Exponents.

  • memory_limit_elements (int) – Max number of float elements in the temporary matrix buffer. 10^7 elements ~ 80MB RAM.

Returns:

The scalar log-value.

Return type:

float