This project made with MATLAB. It's about machine learning. Artificial Neural Network in the form of Multilayer Perceptron. Also it is using backpropagation algorithm.
neural.m - This is the file that i used to create mine. It was my source. neural2.m - This is the matlab file that i edited as my requirements. untitled6.m - Reading dataset for train. untitled7.m - Reading dataset for test.
n = 2.6; nbrOfNodes = 10; nbrOfEpochs = 50;
ORIGINAL ASSIGNMENT
1) Build an artificial neural network in the form of a Multilayer Perceptron. 2) Implement the bakpropogation algorithm given below to train the network. 3) Train the network using the training dataset given. Train for at least 50 epochs. 4) Test the trained network on the provided test file.
Dataset: 150 training samples and 60 test samples with 7 attributes, belonging to 3 classes (1,2,3). The data is gathered from the seeds dataset in http://archive.ics.uci.edu/ml/datasets/seeds The training data and training class labels are given in tab delimited format in “train_data.txt” and “train_class.txt” files. The test data and test class labels are given in tab delimited format in “test_data.txt” and “test_class.txt” files. Evaluation: The accuracy (correct prediction ratio) should be reported by after test phase by the program.
Important issues: • This is a classification program, so your network should round the real number output value to an integer as the prediction result. • Any multilayer perceptron architecture (number of layers, hidden neurons, output neuron numbers, etc.) can be used. You are supposed to design your network to accept datasets with variable number of attributes. • The homework can be done in groups of maximum 3 people, or individually. • You are asked to give a working demo of your program on the final exam hours. In this demo, the provided data, and also two different datasets will be used to test your neural network. So, each group should bring a personal computer to the exam. • The program can be implemented in any programming language of choice. • NO LIBRARY OR TOOLBOX CONTAINING NEURAL NETWORK FUNCTIONS SHOULD BE USED!!! THE ALGORITHM SHOULD BE CODED BY THE PROJECT GROUP, OTHERWISE THE GRADE YOU WILL GET FROM THE FINAL EXAM WILL BE MINIMAL.