1. Define one or two decreasing functions of time for updating the learning rate η(t) and make corresponding changes in the lvq_update_learning_rate() routine. Compile and run the LVQ and compare the performance with the original version.
2. Modify the LVQ source code for the implementation of a SOFM with a onedimensional output slab.
3. Based on the implementation from exercise 8, implement the SOFM with a two-dimensional output slab.
4. Modify the evolutionary back-propagation implementation so that it also evolves the slope of the sigmoid function.
5. Run the back-propagation implementation on the web site using the run files provided for training and testing on the Iris dataset. Keeping everything else constant in the run files, try training for 10,000, 1,000, and 100 iterations. Discuss your results with respect to both training and testing.
Already registered? Login
Not Account? Sign up
Enter your email address to reset your password
Back to Login? Click here