Skip to content

willtebbutt/PseudoPointStateSpace-UAI-2021

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Combining Pseudo-Point and State Space Approximations for Sum-Separable Gaussian Processes

Implements the experiments in [1]. This repo is helpful for the sake of reproducing the experiments, but a better starting point for using the approximation developed in [1] is TemporalGPs.jl.

All experiments should be run from the top level of this repo -- don't navigate into the scripts directory.

You must have a working version of Julia installed to run this code -- experiments were run using version 1.5.3, a copy of which can be obtained from https://julialang.org/downloads/ . I recommend running these experiments using version 1.5.3.

There are no additional binary dependencies other than Julia itself.

This project comes equipped with a Manifest.toml, which is a specification of the exact commit of each dependency which was used in this project. Consequently, when running any of the scripts below, you should wind up with precisely the same versions of all dependencies originally used to produce the results from the paper.

Benchmarking

Run the scripts/run_benchmarking_experiments.sh script. Results will be output to results/benchmarking.

Global Historical Climatology Network

Firstly, you'll need to get hold of the NASADEM (NASA Digital Elevation Map) data set. First sign up for an account here: https://urs.earthdata.nasa.gov/users/new to get credentials. Then navigate to the scripts/nasa_data_grabbing directory and run the get_all_of_nasadem.sh script. Running this script will prompt you for the credentials you obtained from NASA. It will download the data associated with the region studied in this work. It will produce a number of zip files, which you should unzip into scripts/nasa_data_grabbing/nasa_data.

Once the above has been achieved, run

julia scripts/global_historical_climatology_network.jl

to run all experiments and reproduce results. Running this script will prompt you to download a large amount of data from NOAA's Global Historical Climatology Network daily dataset, which is necessary to run these experiments.

Results will be output into results/global_historical_climatology_network.

Apartment Prices

Run

julia scripts/apartment_prices.jl

Results will be output into results/apartment_prices. Running this script will prompt you to download a dataset from HM Land Registry database, and a database from Camden council containing a mapping between postcodes and lat-lon coordinates.

Something Doesn't Work?

Please open an issue.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published