Closed danbraunai closed 2 years ago
@leesharkey thoughts?
Hmm actually this legend is only really accurate for the Bar chart, because there is no white (=0 saliency) in the saliency maps.
It's a great idea. I think we'd need to use a different spectrum (the current one gets too dark) but otherwise looks great. I don't think we should worry about the whiteness issue. So long as the very middle of the spectrum is the same colour as the background of the webpage, then it's possible to interpret it as 'middle is transparent'
On Tue, 19 Apr 2022 at 00:08, Dan Braun @.***> wrote:
Hmm actually this legend is only really accurate for the Bar chart, because there is no white (=0 saliency) in the saliency maps.
— Reply to this email directly, view it on GitHub https://github.com/interpretingrlbehavior/interpretingrlbehavior.github.io/issues/71#issuecomment-1101840675, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABYCJ2NUL7IE3G2HOWNF6DTVFXTO5ANCNFSM5S76E25A . You are receiving this because you were mentioned.Message ID: <interpretingrlbehavior/interpretingrlbehavior.github. @.***>
Legend.
Maybe somewhere near the 'Saliency step: 4' text, that indicates that red is neg saliency and blue is positive saliency