AI in HEP: Can “Evolvable Discriminate Function” discern Neutral Pions and Higgs from background? James Cunha Werner Christmas Meeting 2006 – University of Manchester

Neutral Pion Reconstruction Neutral Pions decays into 2 Gammas (in the same way Higgs does!), detected by BaBar’s Electromagnetic Calorimeter How to DISCRIMINATE background from real neutral pions? 2 gammas from background can reconstruct a neutral pion just by chance!

Previous papers using Genetic Programming for event selection in HEP: Cranmer,K.; Bowman,R.S.; "PhysicsGP: A genetic programming approach to event selection" Computer Physics Communications 167 (2005) Focus Collaboration, "Application of genetic programming to high energy physics event selection" Nuclear instruments and methods in physics research A 551 (2005) Focus Collaboration; "Search for L+c -> pK+p- and D+s -> K+K+p- using genetic programming event selection" Physics letters B 624 (2005) Mjahed, M.; "Search for Higgs boson at LHC by using genetic algorithms" Submitted to Nuclear Instruments and Methods in Physics Research. My approach is original because…it uses genetic programming to obtain a discriminate function to discern between neutral pions and background.

Discriminate Functions Mathematical model obtained with GP maps the variables hyperspace to a real value through the discriminator function, an algebraic function of kinematics variables. Applying the discriminator to a given pair of gammas: –if the discriminate value is bigger than zero, the pair of gammas is deemed to come from pion decay. –Otherwise, the pair is deemed to come from another (background) source.

How Genetic Programming works… AI algorithm that mimics evolution: –Initial random population. –Each individual is one problem solution. Its chromosome codes the solution using functions and variables. –Chromosome represents a mathematical model. –Fitness evaluates solution’s economic function. GP is underlined by Markov chain theory. For more information see

Methodology MC data Training data Test data Discriminate function Raw data Select Real / background events MC data 1. Obtaining Discriminate Function (DF): 2. Test DF accuracy: 3. Selecting events for superposition: GP

Criteria and Events Selection in this study I will focus in neutral pion decaying from Rho(770) resonance.

Training Genetic Programming (GP) to obtain NPDF Monte Carlo (MC) generators integrates particle decays models with detector’s system transfer function. MC events contain all information from each track particle and gamma radiation, which allows select high purity training and test datasets (96%+). Events with real neutral pion were selected and marked as “1”. Events without real pions into MC truth and invariant mass reconstruction in the same region of real neutral pions where also selected and marked as “0”.

Energy cuts cases all gammas without energy cut (60,000 real and background records for training, and 60,000 real and background for test), more energetic than 30 MeV electronics’ noise threshold (32,000 real and background records for training and test), more energetic than 50 MeV (15,000 real and background records for training and test), more energetic than 30MeV, lateral moment between 0.0 and 0.8, and have hit more than one crystal in the electromagnetic calorimeter - the conventional cut for neutral pion(16,000 real and background records for training and test).

NPDF Final results for several energy cuts -α: Sensitivity or efficiency. -β: specificity or purity. -γ: accuracy.

Superposition of all NPDF NPDF obtained from different selection conditions produce the same energy distributions.

Hadronic tau decays results:

All gammas: 30 MeV cut: 50 MeV cut: Conventional Cuts: Discriminate functions obtained by GP

Further development in LHC: Higgs to  +0j,  +1j and  +2j H  +1j H  +0j ATLFAST/DC1 Signal: VBF Signal: gg Fusion EW+DPS  jj QCD  jj  jjj+jjjj H  +2j L = 10 fb -1

Conclusion Genetic programming approach obtains neutral pion discriminate function to discern between background and real neutral pion particles with an average 80% accuracy, 87% sensitivity (efficiency), and 84% specificity (purity). Further development: understand what NPDF model means. What is its relationship with physics laws and properties. Merry Christmas and Happy New Year!!!