In this folder, we have several helper functions for building a k-nearest neighbor classifier:

(1) conf_matrix.m:  Builds a confusion matrix and will also output the error of the classifier.
(2) fitknn.m:  Takes in X, targets and new data with number of neighbors, and outputs the predicted classes.
(3) irisdata.mat:  Sample classification data.  See below for more details on what the data represents.
(4) KfoldCV.m:  Splits a given data set into k-folds, ready for cross validation.
(5) StandardScaler.m:  Scales the data by subtracting the mean and dividing by the standard deviation.
(6) TrainTestSplit.m:  Splits a given data set randomly into a training and testing sets (give it a percentage for test, like 0.3).

Our goal is to build two drivers:

knnapp1.m:  Split the data (70-30), then use the "fixed" training set, and loop through the number of nearest neighbors,
getting the error for each.  Plot the error to see what the best number is. 

knnapp2.m:  Same idea as before, but take the training set and further divide into 5 folds, and find the best "k".

For this homework, we'll use the iris data which really doesn't seem to need scaling.