Simulation Wrapper

class EarlyStopping.SimulationWrapper(design, true_signal=None, true_noise_level=None, monte_carlo_runs=5, noise=None, response_noiseless=None, critical_value=None, interpolation=False, computation_threshold=1e-08, cores=5)

[Source] A wrapper class for collecting montecarlo simulation data.

Parameters

design: ndarray. The design matrix of the simulation, representing the model matrix.

true_signal: ndarray. The true signal or target vector used for generating simulated response values.

true_noise_level: float. The standard deviation of the normally distributed noise applied to the response.

monte_carlo_runs: int. Defines the number of Monte-Carlo runs to perform in the simulation.

noise: ndarray, optional. Specifies an initial noise matrix; defaults to None, in which case noise will be generated as needed.

response_noiseless: ndarray, optional. Represents the noiseless response vector, if available. Default is None.

critical_value: float, optional. Critical value to be used for early stopping. Default is None leading to theoretical values being used.

interpolation: bool, default=False. Specifies whether to use interpolation within the simulation.

computation_threshold: float, default=10 ** (-8). A small threshold to control numerical computations in the simulation procedures.

Methods

run_simulation_truncated_svd(max_iteration, diagonal, data_set_name)	Run montecarlo simulation for truncated SVD.
run_simulation_landweber(max_iteration, learning_rate,``data_set_name``)	Run montecarlo simulation for the Landweber iteration.
run_simulation_conjugate_gradients(max_iteration, data_set_name)	Run montecarlo simulation for conjugate gradients.

SimulationWrapper.run_simulation_truncated_svd(max_iteration, diagonal=False, data_set_name=None)

Runs a simulation for an inverse problem using truncated Singular Value Decomposition (SVD). The function generates a noisy response based on the specified noise level and performs a Monte-Carlo simulation to collect various metrics related to the estimator’s performance.

Parameters

max_iteration: int. Specifies the maximum number of iterations to use within the simulation.

diagonal: bool, optional. Specifies whether to treat the matrix as diagonal. Default is False.

data_set_name: str, optional. If specified, the results are saved to a CSV file with this name.

Returns

results_df: pd.DataFrame. DataFrame containing the results of the Monte-Carlo simulation.

The resulting DataFrame can be saved to a CSV file if data_set_name is provided.

SimulationWrapper.run_simulation_conjugate_gradients(max_iteration, data_set_name=None)

Runs a simulation for an inverse problem using the Conjugate Gradients method. This function generates a noisy response based on the specified noise level and performs a Monte-Carlo simulation to collect various metrics related to the estimator’s performance.

Parameters

max_iteration: int. Specifies the maximum number of iterations to use within the simulation.

data_set_name: str, optional. If specified, the results are saved to a CSV file with this name.

Returns

results: list. A list of results from the Monte-Carlo simulation.

SimulationWrapper.run_simulation_landweber(max_iteration, learning_rate=1, data_set_name=None)

Runs a simulation for an inverse problem using the Landweber iteration method. The function generates a noisy response based on the specified noise level and and performs a Monte-Carlo simulation to collect various metrics related to the estimator’s performance.

Parameters

max_iteration: int. Specifies the maximum number of iterations to use within the simulation.

learning_rate: float or str, optional. Specifies the learning rate for the Landweber iteration. If set to “auto”, the method will search for an optimal learning rate. Default is 1.

data_set_name: str, optional. If specified, the results are saved to a CSV file with this name.

Returns

results_df: pd.DataFrame. DataFrame containing the results of the Monte-Carlo simulation.