Machine Learning Lab Assignment 1

• OBJECT OF THE ASSIGNMENT:
  To solve and understand linear regression problems by using gradient descent algorithm.
• PROBLEM:
  Implement linear regression with one or multiple independent variables to predict a dependent (target) variable.
• INPUT OF THE PROBLEM:
  Training dataset / testing examples
• OUTPUT OF THE PROBLEM:
  (a) Display the coefficients of the hyperplane, i.e., w = (w0, w1, …, wM).
  (b) Predict values of the dependent variable for testing examples.
• TESTING CASES:

  1. The file data1.txt contains a training set of Auto Insurance (汽車保險) in Sweden. The first column (= x) is the number of claims (索賠) and the second column (= y) is the total payment for all the claims in thousands of Swedish Kronor (瑞典克朗).
     (a) Find the regression line ŷ = w0 + w1x. Thus, use the gradient descent algorithm to find the weight w = (w0, w1).
     (b) Once you have found the regression equation, you can use the model to make predictions.
     (i) What is the predicted value of y when x = 45?
     (ii) What is the predicted value of y when x = 25?

  2. The file data2.txt contains a training set of Basketball. Columns 1 – 4 are the feature variables x1, x2, x3, x4, and column 5 is the target variable y. The following data (x1, x2, x3, x4, y) are for each player:
     x1 = height in feet
     x2 = weight in pounds
     x3 = percent of successful field goals (out of 100 attempted)
     x4 = percent of successful free throws (out of 100 attempted)
     y = average points scored per game
     (a) Find the hyperplane ŷ = w0 + w1x1+ w2x2+ w3x3+ w4x4. Thus, use the gradient descent algorithm to find the weight w = (w0, w1, w2, w3, w4).
     (b) Once you have found the hyperplane, you can use the model to make predictions.
     (i) What is the predicted value of y when (x1, x2, x3, x4) = (6.8, 210, 0.402, 0.739)?
     (ii) What is the predicted value of y when (x1, x2, x3, x4) = (6.1, 180, 0.415, 0.713)?

• APPENDIX:
  
• REMARKS:
  Stopping criteria of the Gradient descent algorithm usually includes:
  (a) Stop when a maximum number of iterations has been exceeded.
  (b) Stop when some error measure on the training set is small enough.
  　　For example: