1. Train a single neuron perceptron to classify the Iris dataset provided with this homework. The dataset consists of a 150x3 matrix. Columns 1 and 2 of the data represent the two-dimensional input features, and column 3 contains the class labels. Each of the data samples belongs to one of two varieties of the Iris plant. 
2. 
   
3. a. Is this dataset linearly separable? Show your result graphically. 
4. 
   
5. b. Implement this network in MATLAB without using the neural network toolbox. Separate the data into two sets, and use one set for training the network and the other for testing the trained network. You can use a 70:30 split where 70% of the data is used for training and 30% for testing the network. 
6. 
   
7. c. Plot the mean squared error curve also called the learning curve. 
8. 
   
9. d. Compute the percentage of misclassified testing samples. 
10. 
    
11. e. Plot the 2-dimensional error surface for this problem by varying each of the weights between [-100,100]. f. Study the impact that varying the initial weight vector has on the learning curve and the number of interactions it takes the algorithm to converge. Explain your observations with respect to the error surface you plotted for part d.