No Color Bar

[yywe]

When I fit a Gaussian Process Regression model by calling 'GPy.models.GPRegression'. then I want to plot the model so I call 'm.plot(property='3d'). I know the highlight area usually have a larger value, but how can I show the related color bar for the colours. any help? thanks so much.

[mzwiessele]

When you plot you get back the full dictionary of all matplotlib plots. Use the plot for the image and add a colorbar to the figure yourself.

fig.colorbar(cm, ax=ax)

Where cm is the image you want the colorbar for, ax is the axis you want the colorbar to be added and fig is the figure which contains cm.

[yywe]

The returned plot is fig = m.plot(). The problem is from the fig object, I did not find the image, and also what do I pass for the second parameter ax?

[yywe]

fig.figure.images is empty

[lionfish0]

Solved: This wasn't obvious : )

Using help from this stackoverflow answer:

Immediately after you call m.plot(), call:

ax = plt.gca() #get current axes
mappable = ax.collections[0] #this is specific for what the surface call returns
plt.colorbar(mappable)

More notes/comments on GPy's code: Relevant code is line 214 in GPy/plotting/gpy_plot/gp_plots.py - an issue is this returns a dictionary, with gpmean set to a list of plots, which is what we really need for setting the colorbar properly. However, this method's plots list is fed into the add_to_canvas method by plot (line 351). The add_to_canvas method doesn't use the plots argument, so its content is, I guess, not accessible later. Hence why you're having trouble I guess finding it.

[lionfish0]

Addendum: This doesn't work if you've set the plot_data parameter to False.