International Journal of Sciences: Basic and Applied Research (IJSBAR) ISSN 2307-4531 (Print & Online) http://gssrr.org/index.php?journal=JournalOfBasicAndApplied --------------------------------------------------------------------------------------------------------------------------- 112 Performances of Various Back-propagation Learning Algorithms of Neural Network Using Matlab Md. Ashek-Al-Aziz a *, Abdullah-Hil Muntakim b a Associate Professor, University of Development Alternative (UODA), Dhaka-1209, Bangladesh b Assistant Professor and Assistant Director, University of Development Alternative (UODA), Dhaka-1209, Bangladesh a Email: ashek3000@gmail.com b Email: faculty.ahmuntakim@gmail.com Abstract There are plenty of back-propagation learning algorithms of artificial neural network. Performances of various back-propagation learning algorithms have been checked using few portions of Australian Rain Dataset. Polak- Ribiere conjugate gradient back-propagation and Levenberg-Marquardt back-propagation have showed good performance than others. Keywords: Neural Network; Back-propagation; Training; Testing. 1. Introduction Artificial Neural Networks are the artificial mimic of human brain [1]. Human beings learn with the presence of teacher or guide which is a common learning paradigm [2]. Whatever the inputs received by the receptor of human being, another person tells him/her what the objects should be that is output is defined by the teacher. While this paradigm is subject to be mimicked artificially, the target output is assigned by supervisor for each corresponding inputs to the neural network. Computed output also called actual output is not same as given output or target output or desired output because inputs are multiplied by some random weight values in the neural network. In that case, weight values are changed by back-propagation [3]. ------------------------------------------------------------------------ * Corresponding author.