Boosted decision trees as an alternative to artificial neural networks for particle identification

Abstract

The efficacy of particle identification is compared using artificial neutral networks and boosted decision trees. The comparison is performed in the context of the MiniBooNE, an experiment at Fermilab searching for neutrino oscillations. Based on studies of Monte Carlo samples of simulated data, particle identification with boosting algorithms has better performance than that with artificial neural networks for the MiniBooNE experiment. Although the tests in this paper were for one experiment, it is expected that boosting algorithms will find wide application in physics.

Antiproton Flux, Antiproton-to-Proton Flux Ratio, and Properties of Elementary Particle Fluxes in Primary Cosmic Rays Measured with the Alpha Magnetic Spectrometer on the International Space Station

“…To distinguish antiprotons from charge confusion protons, that is, protons which are reconstructed in the tracker with negative rigidity due to the finite tracker resolution or due to interactions with the detector materials, a charge confusion estimator Λ CC is defined using the boosted decision tree technique [26]. The estimator combines information from the tracker such as the track χ 2 =d:o:f:, rigidities reconstructed with different combinations of tracker layers, the number of hits in the vicinity of the track, and the charge measurements in the TOF and the tracker.…”

Measurement of theB¯s0Effective Lifetime in theJ/ψf0(980Aaij1, Beteta2, Adametz3 et al. 2012Phys. Rev. Lett.182110View full textAdd to dashboardCite

“…Furthermore, the J=c þ À candidate must have a vertex with a fit 2 less than 10, flight distance from production to decay vertex greater than 1.5 mm, and the angle between the combined momentum vector of the decay products and the vector formed from the positions of the primary and the " B 0 s decay vertices (pointing angle) is required to be consistent with zero. Events satisfying this preselection are then further filtered using requirements determined using a boosted decision tree (BDT) [18]. The BDT uses nine variables to differentiate signal from background: the identification quality of each muon, the probability that each pion comes from the primary vertex, the transverse momentum of each pion, the " B 0 s vertex fit quality, flight distance from production to decay vertex, and pointing angle.…”

Observation of the B s 0 → J/ψϕϕ decay

“…Different multivariate selection algorithms, based on a boosted decision tree (BDT) [28,29] with the AdaBoost algorithm [30], are used to select the signal and the reference channel samples. The BDT is trained with simulated B 0 s samples for the signals, while for the background, a sample of 40 millions simulated events containing inclusive B → J/ψ X decays is used.…”