夸克胶子喷注的人工神经网络识别研究

On the Identification of Quark and Gluon Jets Using Artificial Neural Network Method 

为了将人工神经网络用于高能物理中对喷注的分类识别 ,用从高能正负电子对撞的蒙特卡洛模拟中得到的不对称三喷注事件中的夸克喷注和胶子喷注的平均多重数、平均横动量和两类喷注所对的夹角的平均值作为输入BP神经网络的 3个特征参量 ,对 2 .5— 2 2 .5GeV能区的 8个能量间隔进行等精度的训练 .用训练好的神经网络模型对不对称三喷注事件中的夸克喷注和胶子喷注样本进行检验判定 ,并对混合喷注样本进行分类识别 .所得结果表明 ,有望将人工神经网络用于高能正负电子对撞产生的喷注的分类分析 .

The identification of quark and gluon jets produced in e +e -collisions using the artificial neural network method is addressed.The structure and the learning algorithm of the BP(Back Propagation)neural network model is studied.Three characteristic parameters——the average multiplicity and the average transverse momentum of jets and the average value of the angles opposite to the quark or gluon jets are taken as training parameters and are inputed to the BP network for repeated training.The learning process is ended when the output error of the neural network is less than a pre-set precision(σ=0.005).The same training routine is repeated in each of the 8 energy bins ranging from 2.5—22.5 GeV,respectively.The finally updated weights and thresholds of the BP neural network are tested using the quark and gluon jet samples,getting from the non-symmetric three-jet events produced by the Monte Carlo generator JETSET 7.4.Then the pattern recognition of the mixed sample getting from the combination of the quark and gluon jet samples is carried out through applying the trained BP neural network.It turns out that the purities of the identified quark and gluon jets are around 75%—85%,showing that the artificial neural network is effective and practical in jet analysis.It is hopeful to use the further improved BP neural network to study the experimental data of high energy e +e - collisions.