FE520讲解、辅导manuscript、讲解Python语言、辅导Python设计

FE520 Assignment 5

Dan Wang, Zhiyuan Yao

November 2018

1 Linear Regression Practice (40 Pts)

1. Create two random arrays x1 and x2, the values of array are from 0 to 100, and

its size = 1000, sort these arrays from small to large.

2. Create the corresponding y, y = x1 3 + x2 4 + , where  ～ N(0, 2).

3. Combine x1 and x2 as x using pandas or numpy.

4. Let the combining x as your input x, y is your target response to x. Solve the

coefficients Θ using mathematics way and output the coefficients. (Referring to

my manuscript)

5. Using sklearn linear regression model to solve the coefficients. Compare the

difference with last question.

2 Logic Regression Practice (40 Pts)

1. Look at the documentation of (make classification) , to make a dataset with binary

class, sample size = 1000.

2. Randomly set 80% of your data set as training set, and the rest as your test set.

3. Training your data set using logic regression.

4. Test your model using the trained model with your test set, and output the accuracy.

5. (Bonus 10pts) Data visualization: plot (scatter plot) different classes using different

colors, plot the regression to cut off the two classes.

6. (Bonus 10 pts) Using mathematics method to compute the coefficients which is

derived in my uploaded paper.

3 Softmax Regression Practice (20 Pts)

Change the two classes in Q2 into 10 classes, and repeat the steps in Question 2.

1

4 More Practice in sklearn (0 pt)

Due to time limitation, we can’t cover most of detail in class. You are encouraged

to learn more about those algorithms from online resources like books, blogs, and

MOOCs. Then practise with Python.

Actually, the interfaces for implementations of majority of machine learning algorithms

in scikit-learn have the similar processes with linear regression and logistic

regression:

1. Processing the data (dividing the training data and testing data)

2. Importing different models from sklearn, (i.e. SVM, SVR, Decision Tree, Random

Forest, etc.)

3. Training your model.

4. Test your model.

5. See the performance of different classifiers and regressors.

Submission Requirement:

For all the problems in this assignment you need to design and use Python 3, output

and present the results in nicely format. Please submit a written report (pdf), where you

detail your results and copy your code into an Appendix. You are required to submit

a single python file and a brief report and the output as csv format. Your grade will

be evaluated by combination of report and code. You are strongly encouraged to write

comment for your code, because it is a convention to have your code documented all

the time. In your python file, you need contain both function and test part of function.

Python script must be a ’.py’ script, Jupyter notebook ’.ipynb’ is not allowed. Do

NOT copy and paste from others, all homework will be firstly checked by plagiarism

detection tool.