monte_carlo_local

circuit_analyzer.all.monte_carlo_local(circuit_model, wavelengths, samples, parallel=False, parallel_option='number', save_path=None, seed=None)

Run Monte Carlo simulations where each occurrence is modified.

The changes are done on each occurence (hence ‘local’), as opposed to globally, where the same parameter is modified for all occurences on the circuit. This is based on a description provided by _parameter_distribution in the CircuitModelView.

This requires the CircuitModel to have a ‘_parameter_distribution’ method. That method returns a number of samples taken from a distribution, for each parameter in the CompactModel.

Parameters:

circuit_model: CircuitModel

IPKISS CircuitModel used in the simulation.

wavelengths: array-like

An array of wavelengths, for instance: np.linspace(1.54, 1.56, 1001).

samples: int

Number of simulations.

parallel: bool, optional

Parallelize the calculations to potentially reduce the simulation time.

parallel_option: str

String of what to parallelize, either over ‘wavelengths’ or over the ‘number’ of simulations. Options are ‘wavelengths’ or ‘number’.

save_path: str/path, optional

Path where the save file will be stored, for instance: save_path = “simulation_results.data”.

seed: int or None

A seed value for the random number generator to ensure reproducibility. Defaults to None (no seed).

Returns:

an array of S matrices, one S matrix per sampling.

Examples

This example shows how to use the function to generate parameter samples for a given number.

>>> import ipkiss3.all as i3
>>> from circuit_analyzer.statistical_modeling import monte_carlo_local
>>>
>>>
>>> class Cell(i3.PCell):
>>>     class CircuitModel(i3.CircuitModelView):
>>>         insertion_loss = i3.NumberProperty(default=1.23)
>>>         bandwidth = i3.PositiveNumberProperty(default=2.23)
>>>         center_wavelength = i3.PositiveNumberProperty(default=1.55)
>>>
>>>         def _parameter_distributions(self, number: int, seed: None):
>>>             '''Returns a dict with a `number` of samples for each parameter of the CompactModel.
>>>
>>>             This is a required method for Statistical CircuitModels!!!
>>>             '''
>>>             from scipy import stats
>>>             import numpy as np
>>>
>>>             insertion_loss = np.asarray([1.22, 1.23, 1.21, 1.20, 1.211, 1.266, 1.25])
>>>             bandwidth = np.asarray([1.22, 1.23, 1.21, 1.20, 1.211, 1.266, 1.25]) + 2
>>>             center_wavelength = np.asarray([1.50, 1.53, 1.55, 1.56, 1.57, 1.55, 1.54])
>>>
>>>             dist = stats.gaussian_kde([insertion_loss, bandwidth, center_wavelength])
>>>             samples = dist.resample(number, seed=seed)
>>>
>>>             result = {
>>>                 "insertion_loss": samples[0],
>>>                 "bandwidth": samples[1],
>>>                 "center_wavelength": samples[2],
>>>             }
>>>             return result
>>>
>>>         def _generate_model(self):
>>>             return SomeModel(
>>>                 insertion_loss=self.insertion_loss,
>>>                 bandwidth=self.bandwidth,
>>>                 center_wavelength=self.center_wavelength,
>>>             )
>>>
>>> # We can now use this statistical CircuitModel to run Monte Carlo simulations:
>>> smatrices = monte_carlo_local(
>>>     circuit_model=Cell().CircuitModel(),
>>>     wavelengths=np.linspace(1.5, 1.6, 200),
>>>     parallel=False,
>>>     samples=100,
>>> )
>>>