Kohmodel
KOHModel(model_parameters, kohdataset, obs_stddev=None, jitter=1e-06)
Bases: Module
Class for a KOH model.
Parameters:
model_parameters: ModelParameters The model parameters for the KOH model. kohdataset: KOHDataset The dataset containing the field and simulation observations. obs_stddev: gpx.parameters.Static The standard deviation of the observations. If not None, it will be static and not estimated. jitter: float The jitter to add to the covariance matrix for numerical stability.
Source code in .tox/docs/lib/python3.12/site-packages/kohgpjax/kohmodel.py
GP_posterior(GPJAX_params)
Constructs the GP posterior using the GPJAX parameters. Args: GPJAX_params: The GPJAX parameters in the same shape as prior_dict. Returns: A KOHPosterior object.
Source code in .tox/docs/lib/python3.12/site-packages/kohgpjax/kohmodel.py
get_KOH_neg_log_pos_dens_func()
Returns a function which calculates the negative log posterior density of the model.
Source code in .tox/docs/lib/python3.12/site-packages/kohgpjax/kohmodel.py
k_delta(params_constrained)
abstractmethod
Returns the delta kernel, which is used to model the structure of the calibration parameters. To be implemented by subclasses. Args: params_constrained: The constrained parameters of the model. Returns: A GPJAX kernel.
Source code in .tox/docs/lib/python3.12/site-packages/kohgpjax/kohmodel.py
k_epsilon(params_constrained)
Returns the epsilon kernel, which defaults to a white noise kernel. This is used to model the observation variance. Args: params_constrained: The constrained parameters of the model. Returns: A GPJAX white noise kernel with the observation variance.
Source code in .tox/docs/lib/python3.12/site-packages/kohgpjax/kohmodel.py
k_eta(params_constrained)
abstractmethod
Returns the eta kernel, which is used to model the structure of the field observations and simulation outputs. To be implemented by subclasses. Args: params_constrained: The constrained parameters of the model. Returns: A GPJAX kernel.
Source code in .tox/docs/lib/python3.12/site-packages/kohgpjax/kohmodel.py
likelihood(num_datapoints, GPJAX_params)
Constructs the likelihood for the KOH model. Args: num_datapoints: The number of data points in the dataset. GPJAX_params: The GPJAX parameters in the same shape as prior_dict. Returns: A GPJAX likelihood object.