CECS 551
Assignment 4
Total: 25 Points
General Instruction
Submit uncompressed file(s) in the Dropbox folder via BeachBoard (Not email).
1. Design a neural network and test regularization methods using Keras library.
(a) Find Assignment 4 regularization.html and data.mat.
(b) The main objective of the assignment is reproducing the last two figures in
Assignment 4 regularization.html
(c) (10 points) Design your own neural network and implement it.
Intentionally, consider a complex network with high epochs to build a overfitted non-linear decision boundary.
train X and test X have two features x1 and x2.
train y and test y includes the classes of (x1, x2), 0 or 1.
The objective of the network is classifying (x1, x2) as 0 or 1.
(d) (5 points) Implement a network by adding L2 regularization to the previous one
without modifying the network design.
(e) (10 points) Draw decision boundaries based on training data set between two classes.
You can refer this site.
(f) Submit your Assignment 4 regularization.ipynb file.
