1. 3rd ROOT Workshop, M. Kunze1 The Neural Network Objects RhoNNO: A ROOT package for general use Marcel Kunze Institut für Experimentalphysik 1 Ruhr-Universität Bochum

2. 3rd ROOT Workshop, M. Kunze2 Why ? Neural Network Objects are around us and tested in various applications since several years (“NNO”) ROOT seems very appealing wrt. object persistence and interactivity NNO has been overhauled to re-use ROOT functionality (“RhoNNO”) Integration of other ROOT applications (Neural Network Kernel of J.P.Ernenwein) RhoNNO complies to ROOT‘s coding standards

3. 3rd ROOT Workshop, M. Kunze3 Neural Network Models Supervised Training Models Multi-Layer Perceptron (TMLP, TXMLP) Fisher Discriminant (TFD) Supervised Growing Cell Structure (TSGCS) Supervised Growing Neural Gas (TSGNG) Neural Network Kernel (TNNK) Unsupervised Training Models Learning Vector Quantization (TLVQ) Growing Cell Structure (TGCS) Growing Neural Gas (TGNG) Implementation of the most popular supervised and unsupervised network models

4. 3rd ROOT Workshop, M. Kunze4 Growing Neural Gas (GNG) Example: Adaptation to a multi-dimensional PDF (1D-2D-3D) Fractal growth process: Density and topology conservation

5. 3rd ROOT Workshop, M. Kunze5 Architecture RhoNNO class hierarchy

6. 3rd ROOT Workshop, M. Kunze6 Interactive Control The plotter highlights training progress - Output functions - Error functions Positive Samples (“pro”) Negative Samples (“con”)

7. 3rd ROOT Workshop, M. Kunze7 Management of Data Sets TDataServe A mini database to support management of input/output vector relations - Add vector pairs to data sets - Partition data sets for training/testing - Shuffle data sets - Serve vector pairs - Save and load data sets using ROOT persistence

8. 3rd ROOT Workshop, M. Kunze8 VNeuralNet Interface Abstract base class of all network models // Abstract interface for all networks virtual void AllocNet() = 0; virtual void InitNet() = 0; virtual void WriteText() = 0; virtual void WriteBinary() = 0; virtual void ReadText() = 0; virtual void ReadBinary() = 0; virtual Double_t* Recall(NNO_INTYPE* in,NNO_OUTTYPE* out=0) = 0; virtual Double_t Train(NNO_INTYPE* in,NNO_OUTTYPE* out=0) = 0; // Training and testing Double_t TrainEpoch(TDataServe *server, Int_t nEpoch=1); Double_t TestEpoch(TDataServe *server); void BalanceSamples(Bool_t yesNo = kTRUE); virtual void SetMomentumTerm(Double_t f); virtual void SetFlatSpotElimination(Double_t f);

9. 3rd ROOT Workshop, M. Kunze9 NetworkTrainer A RhoNNO sample application to train and test neural networks - Assemble training and test data sets out of ROOT trees, based on TFormula - Define network architecture and transfer functions - Define and execute a training schedule - Persist networks - Generate C++ code to perform network recall

10. 3rd ROOT Workshop, M. Kunze10 Example: PID Tagging Identify charged tracks using dE/dx,  C etc. P [GeV/c] p K   e cc

11. 3rd ROOT Workshop, M. Kunze11 The ROOT Training File Arbitrary standard tree to provide training and test sample MC-Truth: 1 = electron 2 = muon 3 = pion 4 = kaon 5 = proton Measurement and shape variables Likelihood values

12. 3rd ROOT Workshop, M. Kunze12 Steering File  # Example: Training of PIDSelectors with NNO #define the network topology and training schedule xmlp 7 15 10 1 # MLP with 2 hidden layers transfer TR_FERMI # Transfer function momentum 0.2 # Momentum term balance true # Assure same statistics for pro and con samples plots true # Show updating error plots on training progress test 10000 # Number of test vector pairs to reserve start 1 # First training epoch stop 200 # Last training epoch #define the data source, take two input files datapath../Data # Directory to look up data files networkpath../Networks # Directory to persist network files file PidTuple1.root # First file to get input from file PidTuple2.root # Second … (ad infinitum) #set up the input layer (use branch names) tree PidTuple # This is the tree to look up data cut mom>0.5&&dch>0&&dch<10000 # Preselection of samples input mom:acos(theta):svt:emc:drc:dch:ifr #Input layer autoscale true # Apply a scale to assure inputs are O(1) #set up the output layer (use branch names) #Particles pid = {electron=1,muon,pion,kaon,proton} output abs(pid)==3 # Perform training for pions

13. 3rd ROOT Workshop, M. Kunze13 Generation of Recall Code // TXMLP network trained with NNO NetworkTrainer at Fri Apr 27 // Input parameters mom:acos(theta):svt:emc:drc:dch:ifr // Output parameters abs(pid)==3 // Training files: //../Data/PidTuple1.root //../Data/PidTuple2.root #include "RhoNNO/TXMLP.h" Double_t* Recall(Double_t *invec) { static TXMLP net("TXMLP.net"); Float_t x[7]; x[0] = 0.76594*invec[0];// mom x[1] = 2.21056*invec[1];// acos(theta) x[2] = 0.20365 *invec[2];// svt x[3] = 2.2859*invec[3];// emc x[4] = 1.75435*invec[4];// drc x[5] = 0.00165*invec[5];// dch x[6] = 0.85728*invec[6];// ifr return net.Recall(x); }

14. 3rd ROOT Workshop, M. Kunze14 Execute the Example NetworkTrainer

15. 3rd ROOT Workshop, M. Kunze15 Summary The re-use of ROOT functionality improves NNO a lot TFormula works great to pre-process samples Run training either from CINT C++ or from ASCII steering file A RhoNNO GUI and/or Wizard is still missing RhoNNO comes as part of the Rho package, but can be used independently: Installation of the shared lib plus the headers is sufficient.Rho package The NetworkTrainer application runs standalone Documentation: http://www.ep1.ruhr-uni-bochum.de/~marcel/RhoNNO.html http://www.ep1.ruhr-uni-bochum.de/~marcel/RhoNNO.html