Hi, I am hopefully very close to submitting some first results. I would just like to veryify something about the overal approach. Is it correct that the feedback (and thus training) aspect to this challenge comes purely from submitting multiple entries and seeing which approach works best, then, presumably, taking the best-scoring approach(es) and try to refine/develop them further? I hope this question makes sense!
Created by Stijn van Dongen speedwell In the navigation bar on the left, it's a sub-page of "4 - Leaderboards" (maybe it wasn't expanded :)
Here's the link: [Leaderboard Sub-challenge 1](syn6156761/wiki/402586) OK ... I was going to wait until I've looked at my submission with further questions, but for the life of me I cannot find any 'leaderboard' page other than the one describing what it is. Undoubtebly it is staring me in the face, but where is it? (I understand it will be a while before my submission will show up, but I just haven't found the place where it is supposed to show up).
Hi
Yes that's right. A given method typically has one or more parameters to set. The leaderboard allows to tune your settings to the networks / problem. A key issue is to find a good granularity of the modular decomposition (module size / number of modules). You can use different settings for each network. See also [this post](syn6156761/discussion/threadId=706) and the [Preliminary Results](syn6156761/wiki/400653) for discussion.
Best, Daniel In the absence of ground truth, I guess that is the only available option.