{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

# HW1 - d What other critical parameters impacted the results...

This preview shows page 1. Sign up to view the full content.

ME 537 Learning Based Control Fall 2010 HW #1: Neural Networks Due: 10/8/2010 at noon Use your favorite programming language to implement a one hidden-layer feed forward neural network. Directory Homework/hw1.data contains four data files. Each file has the number of data points listed on the first line, followed by one data point on each line where the data points have two input (x1, x2) and two outputs (y1, y2): x1, x2, y1, y2 d1.train contains 200 training patterns d1.test, d2.test, d3.test contain 100 test patterns each. Use the gradient descent algorithm to train a two input, two output (one for each class) neural network using file d1.train . Write a report addressing the following questions (you should run experiments to support each of your answers): 1- Describe the training performance of the network: a. How does the number of hidden units impact the results? b. How does the training time impact the results? c. How does the learning rate impact the results?
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: d. What other critical parameters impacted the results? Note, this is a classification problem, meaning that each data pattern (x1, x2) belongs to one of two classes (y1 or y2). Consequently, use correct classification percentage (instead of MSE) to report your results. You will still use MSE to train the neural networks; you will simply report the classification percentage (or classification error) to assess the performance of the neural networks. 2- Use d1.test to test the performance of the trained neural network. Answer question a-d from above for the test set. What conclusions can you draw from your results? 3- Use d2.test to test the performance of the trained neural network. Answer question a-d from above for the test set. What conclusions can you draw from your results? 4- Use d3.test to test the performance of the trained neural network. Answer question a-d from above for the test set. What conclusions can you draw from your results?...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online