Plan of the lab experience

• Introduction (5 min.)

• The k-NN algorithm (5 min.)

• Splitting a test dataset in pure Python (10 min.)

• Code Euclidean norm in pure Python (5 min.)

• Develop a classifier for Iris (10 min.)

• Solution (5 min.)

Development of a k-NN classifer in Python

• Split the data-set into training(75%) and test(25%)

• Run the k-NN classification algorithm

• Plot output from the classifier

• Find the basic machinery from iris_k-NN_base.py

Task 1: Data-set split

def train_and_test_split(N, test_fraction=None):
    #Hint:use np.random function to generate an array of True and False
    inputs=np.array([[1, 2], [3, 4],[5,6]])
    train_part=[True, False, True]
    train_inputs=inputs[train_part,:]
    print(train_inputs)

KNN classification algorithm

Display from Tan et al., 2014.

Task 2: Find distance between two vectors

def euclidean_distance(v1, v2, n):
    #distance-- a scalar value of euclidian distance between two vectors
    distance = 0
    %%TODO: write codes to find distances between two vectors
    return distance

Task 3: Get prediction

def classify(neighbors, train_targets):
    predicted_class=2.0
    # %% Write codes to classify
    #return the majority class
    return predicted_class

Output

Actual classification	Predicted classification

To complete...

Use the basic entropy module seen in class to compute the quality of your solution

File entropy.py is for download from the class repository

import entropy as ent
# ...
myeta = ent.H(myclass, target)