Over the summer, we put together a paper on the HiggsHunters project and your performance, and it’s now been submitted to a journal and the arxiv preprint server – have a look!
One highlight is a discussion of your ability to identify Off-Centre Vertices compared to reconstruction algorithms, and in a substantial fraction of cases you do a better job! Out of nine different types of simulated “baby bosons”, you outperform the computers for three and come close on a further three. Below is an example of one of the cases where you do best compared to the reconstruction algorithms, with different coloured points representing different ways of clustering your clicks, for example varying N_clicks = the number of you who clicked on a given location. It’s clear you can identify vertices more efficiently (points further to the right) and with fewer false identifications (further down) than the algorithm – marked by a black star – so well done! We also talk about your abilities to spot interesting and unexpected features, for example the muon jet.
This bodes well for the next steps, where with the help of researchers in schools we’ll further optimise the combination of your clicks and then apply the lessons learned to see what we can uncover in the actual data!
Here are a few (of many!) nice examples where you’ve collectively outperformed the reconstruction algorithms to find some Off-Centre Vertices in simulation.
Image 1 – AHH00002qz ‘zoom’
7 clicks from 7 people (marked with pale blue dots) form a cluster (orange dot at (420,590)) right on top of one the ‘truth’ (i.e. real, from simulation) OCVs (yellow dots) – well done!
Image 2 – AHH00002rw ‘normal’
Here 3 people have together identified both OCVs – the orange dots are hidden underneath the yellow ones since they’re spot on!
Image 2 – AHH00006xk ‘slice’
Saving the best for last: a more complicated image, with a few ‘stray’ tracks complicating things. However, when combined the 28 clicks from 7 people form clusters bang on the simulated decay points! As an aside: the ‘weird’ clicks are marking the little coloured dots you often see in slice views – these are completely expected, and show the positions of all the proton-proton collisions that happened at the same time but at different positions along the beam pipe (which runs horizontally across the centre of the image).
It’s great to see you all doing such a good job at finding these OCVs, more analysis to follow!
With the 2016 LHC restart well under way, having successfully recovered from attacks by small mammals (BBC article), it’s time to start analysing your clicks in earnest! Over the past year and a half you’ve made more than half a million classifications (thank you!), and flagged a whole host of interesting features. Over the next few months we’ll be extending the work done by Thomas (e.g. here) on this enlarged dataset to see what we can find out! Stay tuned for more details, and keep on clicking – every image you look at is potentially an as-yet unseen event!