Final project:
I have enclosed two different sets of training and testing data (A and B) for the final project.
The training data has classifications in the last column. There are no classifications for the testing data.
Data A has classes 1,2 and 3.
Data B has classes 2 and 4.
Use the training data to train the neural network; have your trained network evaluate the test data.
For both A and B explain how to construct your network architectures and how you trained them.
Plot the MSE as a function of training Epoch.
Make a table showing the classification accuracy after training. For example, for data A, let each row of the table (1,2,3) indicate when a particular class is presented to the network, let each column (1,2,3) tally the number of classifications made for each class.
Create tables using the testing data. Each row will specify a test instance. For data set A, the first 4 columns will specify the elements of the test instance, the fifth column will specify the class for that instance made by your network.
For data set B, the first 9 columns will specify the elements of the test instance, the tenth column will specify the class for that instance made by your network.
- For the Final Project, for each data set use the following methods:
Rosenblatt's Perceptron
Multilayer Perceptron
Radial Basis Function
Support Vector Machine
More expaination:
Final Project Information:
Both data sets need to be trained using the architectures below.
Rosenblatt's Perceptron
Multilayer Perceptron
Radial Basis Function
Support Vector Machine
There are two different sets of training and testing data (A and B) for the final project.
The training data has classifications in the last column. There are no classifications for the testing data.
Data A has classes 1,2 and 3.
Each of the four architectures will have three different output neurons, one for each class.
Data B has classes 2 and 4.
Each of the four architectures will have only one output neuron to identify the class.
Use the training data to train the neural network; have your trained network evaluate the test data.
For both A and B explain how to construct your network architectures and how you trained them.
Rosenblatt's Perceptron What are the final weights and biases? What activation function?
Multilayer Perceptron What are the final weights and biases? What activation function?
Radial Basis Function What are the final weights and biases? What are the radial basis centers?
Support Vector Machine What kernel and its parameters? What are the support vectors, their desired values and alphas?
Plot the MSE as a function of training Epoch.
Make a table showing the classification accuracy after training.
When doing the final project, keep in mind that for EVERY CLASS there are four possibilities that should be answered in the report.
Example for class 1 output node.
1 Given class 1 instance is presented, it is classified as belonging to class 1.
2 Given class 1 instance is presented, it is classified as not belonging to class 1.
3 Given class 2 or 3 instance is presented, it is classified as not belonging to class 1.
4 Given class 2 or 3 instance is presented, it is classified as belonging to class 1.
Examples for showing the four possibilities above.
Example 1, for data A, let each row of the table (1,2,3) indicate when a particular class is presented to the network, let each column (1,2,3) tally the number of classifications made for each class.
Example 2:
Do the following for each class; class 1 is used as an example.
Given a class 1 instance is presented, how many were classified correctly (classified as class 1), how many classified incorrectly (classified as NOT class 1).
Given a NOT class 1 instance is presented (i.e. class 2 or 3), , how many were classified correctly (classified as NOT class 1), how many classified incorrectly (classified as class 1).
Create tables using the testing data. Each row will specify a test instance.
For data set A, the first 4 columns will specify the elements of the test instance, the fifth column will specify the class for that instance made by your network.
For data set B, the first 9 columns will specify the elements of the test instance, the tenth column will specify the class for that instance made by your network.
Attachment:- Dataset.rar