Closed chrismessenger closed 6 years ago
Shall try a version with all the curves and post a link here when ready.
Here's a plot without SNR 10 and 12 (plus same axis ranges) and a single plot with SNRs 2-8: Thoughts?
I like the single plot but you could drop the snr 8 results.
Cheers
Chris
Sent from my iPhone
On 28 Oct 2017, at 10:20, Michael Williams notifications@github.com<mailto:notifications@github.com> wrote:
Here's a plot without SNR 10 and 12 (plus same axis ranges) and a single plot with SNRs 2-8: [roc_curves_single]https://user-images.githubusercontent.com/25609742/32133117-5e4327b4-bbc9-11e7-9a97-d61d6db02683.png [roc_curves]https://user-images.githubusercontent.com/25609742/32133121-64728792-bbc9-11e7-896a-f4663af15a12.png Thoughts?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/hagabbar/matching_cnn_paper/issues/30#issuecomment-340164403, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AIoeJcTGr7UPCr4abZVKOy4v0eyHnRyAks5swvHngaJpZM4QJODe.
Removed SNR 8 and uploaded the plot to the repository (also fixed the SN typo). I'm also looking into adding the errors using the binomial distribution.
Seeing as the ROC curves have been brought up in Tom's comments, I think updates to the plots should be discussed there.
We need to decide on a better way to show the ROC results. We certainly don't need to show plots for SNR 10 and 12.
It would be good to have the same axis ranges on all plots but may not be practical.
One possibility is to have 1 plot with many curves but that could get very messy.