Quan Nguyen - But what is a Gaussian process? Regression while knowing how certain you are

Python

Discover the power of Gaussian processes in quantifying uncertainty and making predictions with confidence intervals. Learn how to implement GP regression in Python, model joint distributions, and visualize results with parallel coordinates plots.

Key takeaways

The speaker starts by explaining the mathematical intuition behind Gaussian processes.
Uncertainty is quantified by the confidence interval, which increases gradually to reflect increasing uncertainty.
The speaker uses a simple motivating example of regression while knowing how certain you are.
The example illustrates how kernel risk regression, a common machine learning model, can be used to solve real-world problems.
The speaker explains that uncertainty quantification can lead to more trust in machine learning models.
The speaker describes how to implement Gaussian processes in Python using the gpytorch library.
The speaker discusses how to model the joint distribution of multiple random variables using multivariate Gaussian distributions.
The speaker talks about the limitations of traditional machine learning models in quantifying uncertainty.
The speaker explains that Gaussian processes can be used to model the unknown target function.
The speaker shows examples of how to use parallel coordinates plots to visualize the individual distributions and correlations between variables.
The speaker discusses how to update the mean and covariance functions to obtain posterior distributions.
The speaker explains how uncertainty quantification can be used to make predictions with confidence intervals.
The speaker shows how to use a toy scenario to illustrate how Gaussian processes can be used to model unknown housing house prices.
The speaker summarizes the key idea of the talk, which is that Gaussian processes are used to quantify uncertainty and make predictions with confidence intervals.
The speaker provides examples of how to use Gaussian processes to solve real-world problems and provides resources for further learning.

Quan Nguyen - But what is a Gaussian process? Regression while knowing how certain you are

More talks