A simple demonstration of the Bayesian Regression models using PyMC3.
Demonstrates the implementations of linear regression models based on Bayesian inference.
GP regression with ARD.
Can select between the MAP inference and MCMC sampling.
To select MAP inference initiate the model as follows.
model = BayesianGPRegression(is_MAP=True)
To select MCMC sampling set is_MAP to False during the model initiation.
The feature relevance can be retrieved as follows.
ard_scores = model.ard_coefficients()
Notice that if is_MAP is set to True then this method will return only the scores. However, if MCMC sampling used, then the uncertainty of the ARD scores will be returned with the scores as follows.
model = BayesianGPRegression(is_MAP=False)
model.fit(X, y)
ard_scores, ard_uncertainty = model.ard_coefficients()