spapros.ut.plateau_penalty_kernel

spapros.ut.plateau_penalty_kernel(var, x_min=None, x_max=None)

Return penalty function.

The kernel can be one or two sided (one-sided if either x_min or x_max is None). The kernel is 1 between x_min and x_max. If one-sided it’s 1 from x_min or till x_max. Outside the defined range the kernal decays with a gaussian function with variance = var.

Parameters:

• var (Union[float, List[float]]) – Outside the defined range, the kernel decays with a gaussian function with variance = var.

• x_min (Optional[ndarray]) – Lower border above which the kernel is 1.

• x_max (Optional[ndarray]) – Upper border below which the kernel is 1.

Returns:

Penalty function.

Return type:

Callable

Example

import numpy as np
import matplotlib.pyplot as plt
import spapros as sp

penalty_fcts = {
    "left"  : sp.ut.plateau_penalty_kernel(var=0.5, x_min=2, x_max=None),
    "right" : sp.ut.plateau_penalty_kernel(var=2, x_min=None, x_max=5),
    "dual"  : sp.ut.plateau_penalty_kernel(var=[0.5,2], x_min=2, x_max=5),
}

x = np.linspace(0,10,100)
_, axs = plt.subplots(nrows=1,ncols=3,figsize=(10,2))
for i, (title, penalty_fct) in enumerate(penalty_fcts.items()):
    axs[i].plot(x,penalty_fct(x))
    axs[i].set_title(title)
plt.show()