Find the Higgs boson may each!

CERN announced on May 12 « Higgs Boson Machine Learning Challenge », a competition for the best algorithm to search for events with бозона Higgs in a set of experimental data. The contest will run until September 15, the winners will get cash prizes from $ 2,000 to $ 7,000. A successful solution can be integrated into the actual process data from the detector ATLAS. To participate in the contest does not need any special knowledge in elementary particle physics.

Higgs boson at the Large Hadron Collider is detected directly, but from the decay products. Enormous energy protons collide in the center of the detector. In the course of the collision can be born Higgs boson, which in a short time decay into other particles. According to the predictions of the standard model is the most popular channel of decay - a pair of quarks B and anti-B. The competition is proposed to focus on more rare events when the Higgs boson decays into tau lepton and lepton antitau. Since these leptons too quickly break through various channels, the detector "sees" only the products of their decomposition. However, a similar set of decomposition products can come in many other ways, so many events and form a background to study exactly the Higgs boson, must be distinguished from the boson events from the background.

In Collider is a huge number of collisions, so it is important to quickly and accurately distinguish the interesting events from uninteresting according to the detector. This and offered to do the contestants.

Each event is described thirty numbers, 17 of which - the immediate data from the detector, and 13 - derivative values ​​calculated from the raw data, which experts believe could be useful for prediction. Among the raw data, for example, PRI_tau_pt - the perpendicular component of the detected pulse "of the hadronic tau" (tau lepton, hadron channel recovery decay). Among the derivatives, for example, DER_mass_MMC - estimated mass of the Higgs boson, which would most likely generate the event (if there ever was a Higgs boson). A complete theoretical description of the parameters given in a special article , though, may not be worth it Read on to approach the task with unblinkered view.

Participants offered training set of 250 thousand events for which it is known, they are a signal or noise, and are asked to classify 550,000 pre-eminent milestones. The results will be evaluated by formula , taking into account the number of correct and incorrect answers. To complicate the adjustment results, you do not tell you the exact result of the test: to end the contest check is conducted on a random subset of 18% of the control sample.

Participants can join a team of up to four people and send up to five solutions per day. You can discuss approaches to solving at forum . To check your solution is enough to send a file with the predictions: you can download the source code only then, if you qualify for the prize.

The authors of the three best solutions will receive cash prizes: $ 7,000, $ 4,000 and $ 2,000. Also ATLAS collaboration will choose a winning team, whose decision will be well suited for use in the experiment (including performance, reliability, and other parameters). This team will be invited to meet with CERN collaboration ATLAS (with travel costs).

Source: habrahabr.ru/post/225591/