AVardanyan

Multivariate Methods of Data Analysis in Cosmic Ray Astrophysics A. Chilingarian, A. Vardanyan: 

Multivariate Methods of Data Analysis in Cosmic Ray Astrophysics A. Chilingarian, A. Vardanyan Cosmic Ray Division, Yerevan Physics Institute, Armenia

Topics: 

Topics Main tasks to be solved in cosmic ray astrophysics Analysis methods Preprocessing and indication of the best parameters Neural Networks for the main data analysis Multi Start Random Search learning algorithm Training, Validation and Generalization errors Overtraining control

Slide3: 

Individual event weights Results of NN classification and estimation Examples of applications

The MAGIC telescope for detecting  -rays from point sources: 

The MAGIC telescope for detecting  -rays from point sources

The MAKET-ANI installation for the registration of Extensive Air Showers: 

The MAKET-ANI installation for the registration of Extensive Air Showers

The development of an extensive air shower induced by primary cosmic ray particle in the atmosphere: 

The development of an extensive air shower induced by primary cosmic ray particle in the atmosphere

The Monte-Carlo Simulation is the key problem of any physical inference in indirect experiments: 

The Monte-Carlo Simulation is the key problem of any physical inference in indirect experiments

What tasks we want to solve measuring EAS characteristics?: 

What tasks we want to solve measuring EAS characteristics? An inverse problem to be solved: Experimental data Simulated data ?,?(Ne,Nμ,Nh,S…)  E,A(Ne,Nμ,Nh,S…) Identification of primary particle type Estimation of primary particle energy

Why Neural Networks?: 

Why Neural Networks? Neural Networks belong to the general class of nonparametric methods that do not require any assumption about the parametric form of a statistical model they use Are appropriate technique for classification and estimation tasks Are able to treat multidimensional input data

The Neural information techniques: 

The Neural information techniques The central issue of Neural Networks is a bounded mapping of n-dimensional input to m-dimensional output: The functional form of is accumulated in -NN parameters (weights) during the NN training process. The NN training process consists in iterative processing of simulated events, The aim of the training process consists in finding that provides the minimum of error (quality) function:

A Feed-Forward Neural Network: 

A Feed-Forward Neural Network

An example of the NN output distribution in case of classification task: 

An example of the NN output distribution in case of classification task

Common drawbacks in NN training process: 

Common drawbacks in NN training process Training only one network can lead to the suboptimal generalization Insufficient training events and a risk of overtraining

Multi start random search algorithm: 

Multi start random search algorithm The Random search learning algorithm implements the following steps: The initial values of NN weights are chosen randomly from Gaussian distribution with =0 and 0.01 The random step in the multidimensional space of NN weights is performed from initial point to modify the weights, the alternation of weights is done according to: where is the NN weight vector at th iteration, is the step size, RNDM is a random number from [0,1] interval, and the term introduces and controls the degree of dependence of the random step value on the already achieved quality function

Slide15: 

The quality function  is calculated at each iteration by presenting all the training events to NN If i ≤ i-1 , then the vector is kept as new weights of NN and the next step is initializing from that point in space of NN weights, otherwise – return to the previous point is implemented and a new random step is performed. Multi start technique consists in training many Neural Nets starting from different initial weights and using different step size parameters, allowing to scan many points in the multidimensional space of NN weights

Training and Validation errors,Overtraining control: 

Training and Validation errors, Overtraining control

An acceptable procedure to avoid the overtraining: 

An acceptable procedure to avoid the overtraining after each successful iteration of the learning process the net error is calculated for the validation sample if the validation error is less than the one obtained at previous iteration, then the NN weights obtained at the current training iteration are memorized else, the NN weights obtained at the previous iteration are stored At the end of training process the weights which provide the minimal error on the validation sample are found and used as the final best weights for NN

The multi start RS technique provides a possibility to select the NN with best performance on the control data set: 

The multi start RS technique provides a possibility to select the NN with best performance on the control data set

Results of energy estimation by NN: 

Results of energy estimation by NN

Results of mass classification by NN: 

Results of mass classification by NN

Application of NN for gamma/hadron separation task in gamma-ray astronomy: 

Application of NN for gamma/hadron separation task in gamma-ray astronomy

Cosmic Ray differential energy spectra obtained by NN classification and estimation: 

Cosmic Ray differential energy spectra obtained by NN classification and estimation