الفهرس | Only 14 pages are availabe for public view |
Abstract In our study, we are interested in artificial neural network, the difference between the artificial and natural neural networks, fields of using neural network, and how neural network are important in our real life. Further more the algorithms of neural networks, and their applications. Finally, we built an error backpropagation neural network and training it in order to get the solutions of four problems. In all of these problems, we satisfied a higher speed by using an error backpropagation feedforward network in comparable with the other methods. In third problem (robot), we compared our results with the results given by Elettreby research and we satisfying a higher accuracy and higher speed. First problem is an equation of second degree and we solve it with error 1.0e08, Second problem is an equation of two variables and we find solve it with error 0.0001.Third problem we takes a robot kinematics problem as an application on neural network we take the set of angel values ( 2 1 (Sk k (Band ) as input to the network then having the end effectors point as outputs with squared error 1.0000e 07. Forth problem is differential equations, in first differential equation we get the values of y and y? for a given t with an error 1.0000e08.In second differential equation is a ?Van Der Pol? differential equation, we get the values of y and y? for a given t with an error 1.0000e07. |