gaussian processes for data regression and sampling

This code showcases the different parameters used for gaussian process regression.

Following Machine Learning and Pattern Recognition (Bishop) I implemented GPR in python along with a small widget using matplotlib to be able to see the effect of the hyperparameters on the regression.

Sampling from the computed distribution:

Using a different kernel (and effectively modelling the stock market):

Requirements:

numpy
matplotlib

Using gpr.py:

import, fit:

import gpr
gp = gpr.GP()
gp.fit(data_x, data_y)

predict:

gp.predict(pred_x)