tanner

caf.distribute.cost_functions.tanner(base_cost, alpha, beta, min_return_val=1e-150, max_return_val=1e+100)

Apply the tanner cost function.

Parameters:

• base_cost (np.ndarray) – Array of the base costs.

• alpha (float) – Parameters of the tanner cost function, see Notes.

• beta (float) – Parameters of the tanner cost function, see Notes.

• min_return_val (float) – The minimum value allowed in the return. Avoid return arrays with values such as 1e-300 which lead to overflow errors when divisions are made.

• max_return_val (float) – The maximum value allowed in the return. Avoid return arrays with values such as np.inf which lead to errors.

Returns:

Output from the tanner equation, same shape as base_cost.

Return type:

tanner_costs

Notes

Formula used for this function is:

\[f(C_{ij}) = C_{ij}^\alpha \cdot \exp(\beta C_{ij})\]

where:

• \(C_{ij}\): cost from i to k.

• \(\alpha, \beta\): calibration parameters.