broken correspondence between method labels on the x-axis and the actual values used

[sergeyplis]

In experiments with the latest polyssifier we've noticed that the mean accuracy values saved in files are different from the bar heights in the generated figure. After looking through the code I conclude that the values in the filenames are correct. It remains unclear why the labels get reordered.
Looks like a bug to me, either in polyssifier or seaborn/polyssifier connection

[rdevon]

Are the labels reordered or just incorrect? Can you provide some instructions to reproduce? Thanks

[alvarouc]

I get the same error. You can observe this by reading the saved results and those in the plot. They are different.
I will also look into this but for now only trust the saved results.

[sergeyplis]

I thought we've fixed it. You're using the latest version, right? And your
seaborn is 0.6 and above?
On Jun 5, 2015 3:49 PM, "Alvaro Ulloa" [email protected] wrote:

I get the same error. You can observe this by reading the saved results
and those in the plot. They are different.
I will also look into this but for now only trust the saved results.

—
Reply to this email directly or view it on GitHub
#5 (comment).

[alvarouc]

How do I get seaborn 0.6? Is it the one at github?
the latest I see on the webpage is 0.5.1

[alvarouc]

You were right, installing the dev version of seaborn solves the problem. So it is a seaborn bug.