Scatter plot tick labels too large for bbob and bbob-largescale data?

[brockho]

I thought that the axis labels in our scatter plots show the exponent of the log10(#f-evals), however, when comparing for example BIPOP-CMA-ES (on bbob) with LBFGS (on bbob-largescale), the scatter plots show for some functions surprisingly high values on the axes:

Is this to be considered a bug or are these values really that high (I imagine it is the former)? It almost looks to me as if the dimensions (from 2 to 640) play a role here for the axes labels, but that would be even weirder.

[nikohansen]

This doesn't look like the scatter plots that I have in mind. The above plot looks like generated with loglog(data), while the original scatter plots are generated like plot(log10(data)) and use log10(data) as tick annotations.

[nikohansen]

OK, I think the problem is that we see the automatically generated minor ticks, because the overall scale of the data is small. Disabling minor ticks should be a quick fix.

[brockho]

Just checked, the problem still appears in the current version (development branch):

[nikohansen]

    l = matplotlib.ticker.FixedLocator([])
    gca().xaxis.set_minor_locator(l)

[nikohansen]

We added a marker at log10(0, 0) (by adding a target) and set the lower axis limit to be at least log10(0, 0), which fixes both problem.

[brockho]

Actually, the current version has the marker at (2,2) but nevertheless, the issue can be closed. Note that a new issue to adapt the LaTeX templates---if needed---has been added (#1962).