Fit Gaussian Process Regression Model

Introduction

In this lab, we will learn how to use Gaussian Process regression to fit a model to a dataset. We will generate a synthetic dataset and use Gaussian Process regression to fit a model to it. We will use the scikit-learn library to perform Gaussian Process regression.

VM Tips

After the VM startup is done, click the top left corner to switch to the Notebook tab to access Jupyter Notebook for practice.

Sometimes, you may need to wait a few seconds for Jupyter Notebook to finish loading. The validation of operations cannot be automated because of limitations in Jupyter Notebook.

If you face issues during learning, feel free to ask Labby. Provide feedback after the session, and we will promptly resolve the problem for you.

Skills Graph

Dataset Generation

We will generate a synthetic dataset. The true generative process is defined as f(x) = x sin(x).

import numpy as np

X = np.linspace(start=0, stop=10, num=1_000).reshape(-1, 1)
y = np.squeeze(X * np.sin(X))

import matplotlib.pyplot as plt

plt.plot(X, y, label=r"$f(x) = x \sin(x)$", linestyle="dotted")
plt.legend()
plt.xlabel("$x$")
plt.ylabel("$f(x)$")
_ = plt.title("True generative process")

Noise-free Target

In this step, we will use the true generative process without adding any noise. For training the Gaussian Process regression, we will only select a few samples.

rng = np.random.RandomState(1)
training_indices = rng.choice(np.arange(y.size), size=6, replace=False)
X_train, y_train = X[training_indices], y[training_indices]

from sklearn.gaussian_process import GaussianProcessRegressor
from sklearn.gaussian_process.kernels import RBF

kernel = 1 * RBF(length_scale=1.0, length_scale_bounds=(1e-2, 1e2))
gaussian_process = GaussianProcessRegressor(kernel=kernel, n_restarts_optimizer=9)
gaussian_process.fit(X_train, y_train)
gaussian_process.kernel_

Predictions and Confidence Intervals

After fitting our model, we see that the hyperparameters of the kernel have been optimized. Now, we will use our kernel to compute the mean prediction of the full dataset and plot the 95% confidence interval.

mean_prediction, std_prediction = gaussian_process.predict(X, return_std=True)

plt.plot(X, y, label=r"$f(x) = x \sin(x)$", linestyle="dotted")
plt.scatter(X_train, y_train, label="Observations")
plt.plot(X, mean_prediction, label="Mean prediction")
plt.fill_between(
    X.ravel(),
    mean_prediction - 1.96 * std_prediction,
    mean_prediction + 1.96 * std_prediction,
    alpha=0.5,
    label=r"95% confidence interval",
)
plt.legend()
plt.xlabel("$x$")
plt.ylabel("$f(x)$")
_ = plt.title("Gaussian process regression on noise-free dataset")

Noisy Targets

We can repeat a similar experiment adding an additional noise to the target this time. It will allow seeing the effect of the noise on the fitted model.

noise_std = 0.75
y_train_noisy = y_train + rng.normal(loc=0.0, scale=noise_std, size=y_train.shape)

gaussian_process = GaussianProcessRegressor(
    kernel=kernel, alpha=noise_std**2, n_restarts_optimizer=9
)
gaussian_process.fit(X_train, y_train_noisy)
mean_prediction, std_prediction = gaussian_process.predict(X, return_std=True)

plt.plot(X, y, label=r"$f(x) = x \sin(x)$", linestyle="dotted")
plt.errorbar(
    X_train,
    y_train_noisy,
    noise_std,
    linestyle="None",
    color="tab:blue",
    marker=".",
    markersize=10,
    label="Observations",
)
plt.plot(X, mean_prediction, label="Mean prediction")
plt.fill_between(
    X.ravel(),
    mean_prediction - 1.96 * std_prediction,
    mean_prediction + 1.96 * std_prediction,
    color="tab:orange",
    alpha=0.5,
    label=r"95% confidence interval",
)
plt.legend()
plt.xlabel("$x$")
plt.ylabel("$f(x)$")
_ = plt.title("Gaussian process regression on a noisy dataset")

Summary

In this lab, we learned how to use Gaussian Process regression to fit a model to a dataset. We generated a synthetic dataset and used Gaussian Process regression to fit a model to it. We used the scikit-learn library to perform Gaussian Process regression and plotted the mean predictions and 95% confidence intervals.