The Electromagnetic Environment Evaluation Based On Algebra Algorithm Of Neural Network

With the development of science and technology, radio communication technology has got great progress, which make people’s work and life more and more convenient. However, the electromagnetic environment becomes more and more complex at the same time. Now, radio signal is almost omnipresent, electromagnetic space is more and more crowded, and electromagnetic radiation is serious day by day. If electromagnetic environment wasn’t monitored, evaluated and managed in time, electromagnetic wave would interference each other seriously, spectrum resources would become scarce day by day, and the electromagnetic environment would be polluted seriously, which may even bring huge blow to the country and people’s life and property. Therefore, in order to protect the electromagnetic environment and understand the current state of electromagnetic environment timely, monitoring and evaluation of electromagnetic environment has a great practical significance. Based on this, firstly analyze the factors which influence the electromagnetic environment, and then select a reasonable evaluation indexes, finally mathematical evaluation model of electromagnetic environment is proposed in this paper.The neural network is composed of neurons that connect each other in a certain way. The function of the network is decided by its connectivity, i.e. its topology structure. The tasks for which different topology structures are suitable are also different. Feedforward three-layer neural network is suitable for classification problems. The complexity of electromagnetic environment is divided into five grades in this paper. This is to say the complexity of electromagnetic environment is divided into five cases. So, determining the complexity of electromagnetic environment can be regarded as a classification problem essentially. Therefore, feedforward three-layer neural network can be used to establish the mathematical model of electromagnetic environment evaluation.The main work of this paper are as following:1. Improve the definition of similar support function. Defining similar support function as the transform function of the hidden layer neurons, hoping that when the connection weight matrix between the input layer and the hidden layer satisfies certain conditions, the output matrix of hidden layer is strictly diagonally dominant matrix in algebraic algorithm of the feedforward three-layer neural network. But not all so defined similar support functions can meet the above requirements. A counterexample is given, then the definition of support function is improved.2. Improve algebraic algorithm of the feedforward three-layer neural network such that the output matrix of hidden layer is upper triangular matrix whose main diagonal elements are not zero, and those matrix is nonsingular. Then determine the connection weights between hidden layer and output layer by solving the linear system of equations, and solving the linear system of equations which coefficient matrix is upper triangular matrix is easier and more convenient than solving the linear system of equations which coefficient matrix is strictly diagonally dominant matrix.3. Give the comprehensive evaluation method based on improved algebraic algorithm of feedforward three-layer neural network.4. Analyze the factors which influence the electromagnetic environment, and then select a reasonable evaluation indexes. Finally, establish electromagnetic environment evaluation model by improved algebra algorithm of neural network.