Lecture on Gaussian processes - Brown University, 2016

Andreas Damianou, 16 Feb. 2016

(Quick link to the PYTHON NOTEBOOK we'll cover.)

General Info: Accompanying code / tutorial in Python

During the tutorial I will run a few examples of Gaussian processes implemented using the Jupyter Notebook (a Python UI which runs and displays results within a browser). You might like to run the examples yourself beforehand, afterwards or even during the tutorial using your laptop. This is entirely optional, but playing with the code yourself will probably help you get a much better feeling of what's going on!

The (current version) of the python notebook can be viewed here and can be downloaded (and run) from here. If you want to download and run it locally, you can obtain the notebook here, but you will first have to install Python (see next section). For the last few steps of the tutorial we'll also use the GP package GPy, see last section of this page.

The code snippets of the notebook can be copied and pasted directly in the python terminal, but you can actually run them within the Jupyter Notebook from your browser (check here about how to it, but basically you just need to open Jupyter and load the ipynb file).

More details on installing Python

It is recommended to install an integrated Python environment such as Anaconda. We will need also the python libraries numpy, scipy (recommended: ver. 0.16) and the convenient python interpeter ipython. Within anaconda, installing libraries is as easy as running conda install numpy from the terminal. For a specific version, try conda install scipy=0.16.
After installing Anaconda (follow specific instructions according to your OS on the software's website) you can start programming in Python through an iPython console. E.g. in Unix you can just type ipython in the console and in Windows you can navigate to Applications/Accessories/cmd and type ipython or you can open "anaconda Command Prompt" from "windows start".
Similarly, to open a specific notebook (such as the one provided for the lecture) you can launch jupyter notebook file_name.ipynb from the console.

More details on installing GPy, a Gaussian process framework in Python

GPy can be installed following the instructions here. To summarize, after installing Python (e.g. Anaconda), type this in the terminal: pip install GPy. That's it! To make sure you installed it correctly, open an IPython terminal and type import GPy followed by GPy.tests().

Advertisment: Gaussian process Summer School & Workshop on UQ.

If you enjoy the lectures, want to learn more and like beer, you might consider registering for one of the future Gaussian Process Summer Schools organized in Sheffield, England, by Prof. Neil Lawrence and his group. Furthermore, this year's workshop (accompanying the summer school) is on Uncertainty Quantification. Details on GPSS website.