Skip to main content

Timeline for Sandbox for Proposed Challenges

Current License: CC BY-SA 3.0

9 events
when toggle format what by license comment
Jun 17, 2020 at 9:03 history edited CommunityBot
Commonmark migration
Aug 24, 2016 at 21:07 history edited user45941 CC BY-SA 3.0
added 112 characters in body
Aug 24, 2016 at 21:02 comment added user45941 @Fatalize I wasn't sure about how people felt about the scoring set not being public, but I will go along with that suggestion.
Aug 24, 2016 at 15:32 comment added AdmBorkBork I agree with Fatalize regarding the hidden test set. Also, that plate #14 is blowing my mind (since I have no problems with vision, I've never experienced the "other" side of these sorts of tests).
Aug 24, 2016 at 12:03 comment added Fatalize As usual with those kinds of classification challenge, the optimal way to provide data should be: a training set, a validation set, and a test set. You only provide the training and validation set, so that people can use the training set to build their algorithms and the validation set to check that they generalize properly. You keep the test set to yourself and use it to give a score to each answer, as such no one can optimize their algorithm to get the best score.
Aug 24, 2016 at 10:11 comment added user45941 @PeterTaylor Thank you for reminding me to add the footnote. As for point 3, disallowing optimizing for the scoring cases takes care of that.
Aug 24, 2016 at 10:09 history edited user45941 CC BY-SA 3.0
added 122 characters in body
Aug 24, 2016 at 10:03 comment added Peter Taylor 1. There seems to be a reference to a footnote, but I don't see the footnote. 2. The descriptions of the plates seem to me ("normal") to match the images. 3. You do realise that the golfiest approach is likely to be to find a single pixel which differs in all the images in the test battery? I'd be rather surprised if any answer actually does OCR.
Aug 24, 2016 at 9:15 history answered user45941 CC BY-SA 3.0