ModelCalibrator

Class Description

The ModelCalibrator class is designed to calibrate a stochastic model using the Sinkhorn divergence. It utilizes the geomloss library for calculating the loss.

Parameters

• model: The stochastic model to be calibrated.

• observed_data: The observed market data used for calibration.

• S0 (optional): Initial value. Default is None.

• T (float): Time horizon. Default is 1.0.

• loss_type (str): Type of loss function. Default is “sinkhorn”.

• p (int): Power parameter for the loss function. Default is 2.

• blur (float): Blur parameter for the loss function. Default is 0.05.

• optimizer_cls: Optimizer class. Default is torch.optim.Adam.

• lr (float): Learning rate for the optimizer. Default is 0.01.

Methods

__init__(self, model, observed_data, S0=None, T=1.0, loss_type=”sinkhorn”, p=2, blur=0.05, optimizer_cls=optim.Adam, lr=0.01)

Initialize the ModelCalibrator with the given parameters.

calibrate(self, num_epochs=1000, batch_size=None, steps=100, verbose=True)

Perform the calibration process.

get_calibrated_params(self)

Retrieve the calibrated parameters after calibration.