# What is known about Higgs LHC machine learning algorithm for identifying Higgs events?

Recently many LHC-affiliated organizations and otherwise announced the Higgs ML learning challenge (in May) running over the summer. There are many competing teams and significant results posted already.

Is this contest duplicating analysis that has been done inside LHC teams already, looking for better Higgs-related statistical classification than "in-house" algorithms? If so, what is known about the nature/performance of "in-house" algorithms? Is any of that published? Especially wondering how it compares with the same contest metric.

-
Pointedly there isn't a "the LHC algorithm". Each experiment runs it's own analysis and will have their own algorithm (or more likely algorithms). The field shares the basic skillset for writing and running these things, but the state of the art is always pressing forward and a better Higgstrap can mean being first to publish some important new measurement, so the details come out after the experiment has milked some value our of each improvement (this is also how you show that it was an improvement...). – dmckee Jun 22 '14 at 1:32
understood ofc there are many algorithms/experiments. however the competition is for a very specific scenario of something like Higgs decay into two taus. has that analysis been published by insiders (LHC teams)? is that different than detection of the Higgs? what algorithm did they use there? etc – vzn Jun 22 '14 at 1:48
I'm not a collider guy and have moved on from the job I had recently where I worked next to a couple and saw a bunch of their talks, so I'm not up on the details of what has been released. I'm a little out of my depth on anything at those energies, but I would expect that mode to be a bit tricky because there is a lot of variability in the missing $p_T$ and many ways the $\tau$s can manifest in the detector. – dmckee Jun 22 '14 at 2:19