Title :
Measuring the time needed for training a neural network based on the number of training steps
Author :
Stoica, M. ; Calangiu, G.A. ; Sisak, F.
Author_Institution :
Electr. Eng. Dept., Transilvania Univ., Brasov, Romania
Abstract :
Artificial neural networks play an important role in robot programming by demonstration. In this paper we present a method for artificial neural network training. The main idea of this method is to train the artificial neural network with all of the data, before the current training step, and at a certain step the network is already trained a huge number of times. Some features of the quality of neural network training, using this method, were presented in. Because the method uses all of the data before the current training step, in this paper, we are concerned about training time and computing time comportment of the neural network. A software application for obtaining training time based on the number of training steps was designed. This software application implements the training method on an unidirectional multi-layer neural network and prints into a graph the training time and computing time. The results obtained using the software application and important conclusions towards the training and computing time comportment are also presented.
Keywords :
graph theory; learning (artificial intelligence); neural nets; robot programming; artificial neural networks; graph; neural network training; robot programming; training steps; unidirectional multi-layer neural network; Artificial neural networks; Biological neural networks; Biological system modeling; Control systems; Humans; Neural networks; Robot kinematics; Service robots; Sliding mode control; Time measurement;
Conference_Titel :
Robotics in Alpe-Adria-Danube Region (RAAD), 2010 IEEE 19th International Workshop on
Conference_Location :
Budapest
Print_ISBN :
978-1-4244-6885-0
DOI :
10.1109/RAAD.2010.5524599