Many of us had a bit of a lazy weekend, but we're back, and we're antsy to catch stalls! 🐜
Daily total vessel count has jumped over 124% since yesterday, and while some teams have gained the lead while others were cracking Easter eggs, the leaderboards could shuffle again soon! Read the detailed report on #CrushALZ Day 11 below.
Day 11 Start of day: April 16, 8pm ET End of day: April 17, 8pm ET Hours: 24
Teams: 21 Total vessels annotated: 5330 ↑124.42 Real vessels annotated: 3361↑143.73 -- Proportion of Real vessels: 63% Progress toward research goal: 11.58% Equivalent lab time: 2 to 6 weeksLeading teams:
Stall Destroyers are still leading the way in the daily fight against stalls! Breathing down their neck are CitSciGamers, and - bounced back after the holiday break, most likely - is our EyesOnALZ "home" team! Middle School STEM are getting back in the game too in 4th place, closely followed by Alzheimer's Research UK.All time results
I think we are likely to see some reshuffling of the All time leaderboard later this week!
How much lab work have we achieved so far?
As you may recall from this teaser video about the competition with Pietro, it could take up to one year to analyse a dataset such as the one uploaded to Stall Catchers for the #CrushALZ competition in the lab:
And we have already analyzed more than 11% in 11 days, which could have otherwise taken up to 5 weeks in the lab! Of course, the exact numbers can be tricky to calculate - as you well know if you play Stall Catchers, some vessels are trickier than others, and same goes for whole experiments!
In any case, we have already saved significant research time, and will continue to do so until the dataset is finished and the current research question has been answered.
When are we set to finish? Our current dataset consists of 18 thousand vessels, and to get a confident crowd answer we currently need to get 20 answers per vessel. This means that using our current methods for crowd answers, we need to carry out 360 thousand annotations to finish the dataset - by the end of Day 11 we had done 41 675.
If we stay on the current track we are set to finish the dataset in about 90 days. While this means we might not make it in time for the #CrushALZ competition finale, we will keep going & still achieve it many times faster than would be possible without Stall Catchers!
On the other hand, there could still be ways we could speed up the analysis, for example...
- Get new catchers! Yep - if you haven't yet, invite everyone to join your team on Stall Catchers!
- Devise dynamic methods for determining crowd score. We are secretly working on an algorithm that could help reduce the amount of annotations we need to get per vessel (and still end up with a confident answer), which would save catcher effort and speed the analysis. But more about that soon...
Thank you for all your help so far - let's keep crushing ALZ for the next 2.5 weeks!
This is a companion discussion topic for the original entry at https://blog.hcinst.org/crushalz-day-eleven/