thomaspinder / gpviz

A Python package for visualising GPJax Gaussian processes.

Hi there 👋

I'm an Applied Scientist at Amazon working on causal analysis. Previously, I completed a PhD in Statistics titled Developments in Gaussian processes with applications to climate science and network problems. I am interested in Gaussian processes, network models, and causal inference.

• 🔭 I’m currently working on non-stationary causal analysis
• 🙌 I’m looking to collaborate on GPJax

gpviz's Issues

y-axis scaling

The y-axis limits should be controlled in a similar manner to the way x-axis limits are set currently.

Create resolver functions

Currently the functions for plotting GPs are quite large and, in the future, cause lots of code duplication. There should be individual sub-functions made such as plot_samples(), plot_mean(), plot_data(), and plot_bands(). This issue should be taken into account when resolving #2 .

Treat colours as a param type object

Currently the colours object is loaded in and called in a way that the does not enable the user to customise the colour palette. This should be made a non-required function argument for plots.

3D plot

Hi, I'd like to plot two columns of X against y. Should I just add condition into plot methods to e.g. ax.plot_surface() when data.X has two columns?

Regards,

Data plotter

Create an object similar to kernels and GPs that will quickly plot the training and/or test data. Plots could either be a scatter plot or a histogram/density plot of the observed values.