Exercise: Linear Regression

转载 2016年05月30日 18:03:24

Exercise: Linear Regression

This course consists of videos and programming exercises to teach you about machine learning. The exercises are designed to give you hands-on, practical experience for getting these algorithms to work. To get the most out of this course, you should watch the videos and complete the exercises in the order in which they are listed.

This first exercise will give you practice with linear regression. These exercises have been extensively tested with Matlab, but they should also work in Octave, which has been called a "free version of Matlab." If you are using Octave, be sure to install the Image package as well (available for Windows as an option in the installer, and available for Linux from Octave-Forge ).


Download ex2Data.zip, and extract the files from the zip file.

The files contain some example measurements of heights for various boys between the ages of two and eights. The y-values are the heights measured in meters, and the x-values are the ages of the boys corresponding to the heights.

Each height and age tuple constitutes one training example $(x^{(i)}, y^{(i)}$ in our dataset. There are $m = 50$ training examples, and you will use them to develop a linear regression model.

Supervised learning problem

In this problem, you'll implement linear regression using gradient descent. In Matlab/Octave, you can load the training set using the commands

x = load('ex2x.dat');
y = load('ex2y.dat');

This will be our training set for a supervised learning problem with $n=1$ features ( in addition to the usual $x_0 = 1$, so $x \in {\mathbb R}^2$ ). If you're using Matlab/Octave, run the following commands to plot your training set (and label the axes):

figure % open a new figure window
plot(x, y, 'o');
ylabel('Height in meters')
xlabel('Age in years')

You should see a series of data points similar to the figure below.


Before starting gradient descent, we need to add the $x_0 = 1$ intercept term to every example. To do this in Matlab/Octave, the command is

m = length(y); % store the number of training examples
x = [ones(m, 1), x]; % Add a column of ones to x

From this point on, you will need to remember that the age values from your training data are actually in the second column of x. This will be important when plotting your results later.

Linear regression

Now, we will implement linear regression for this problem. Recall that the linear regression model is

\begin{displaymath} h_{\theta}(x) = \theta^Tx = \sum_{i=0}^n \theta_i x_i, \nonumber \end{displaymath}

and the batch gradient descent update rule is

\begin{displaymath} \theta_j := \theta_j - \alpha \frac{1}{m} \sum_{i=1}^m (h_{\... ...{(i)}) x_j^{(i)} \;\;\;\;\;\mbox{(for all $j$)} \nonumber \par \end{displaymath}

1. Implement gradient descent using a learning rate of $\alpha = 0.07$. Since Matlab/Octave and Octave index vectors starting from 1 rather than 0, you'll probably use theta(1) and theta(2) in Matlab/Octave to represent $\theta_0$ and $\theta_1$. Initialize the parameters to $\theta = \vec{0}$ (i.e., $\theta_0=\theta_1=0$), and run one iteration of gradient descent from this initial starting point. Record the value of of $\theta_0$ and $\theta_1$ that you get after this first iteration. (To verify that your implementation is correct, later we'll ask you to check your values of $\theta_0$ and $\theta_1$ against ours.)

2. Continue running gradient descent for more iterations until $\theta$ converges. (this will take a total of about 1500 iterations). After convergence, record the final values of $\theta_0$ and $\theta_1$ that you get.

When you have found $\theta$, plot the straight line fit from your algorithm on the same graph as your training data. The plotting commands will look something like this:

hold on % Plot new data without clearing old plot
plot(x(:,2), x*theta, '-') % remember that x is now a matrix with 2 columns
                           % and the second column contains the time info
legend('Training data', 'Linear regression')

Note that for most machine learning problems, $x$ is very high dimensional, so we don't be able to plot $h_\theta(x)$. But since in this example we have only one feature, being able to plot this gives a nice sanity-check on our result.

3. Finally, we'd like to make some predictions using the learned hypothesis. Use your model to predict the height for a two boys of age 3.5 and age 7.

Debugging If you are using Matlab/Octave and seeing many errors at runtime, try inspecting your matrix operations to check that you are multiplying and adding matrices in ways that their dimensions would allow. Remember that Matlab/Octave by default interprets an operation as a matrix operation. In cases where you don't intend to use the matrix definition of an operator but your expression is ambiguous to Matlab/Octave, you will have to use the 'dot' operator to specify your command. Additionally, you can try printing x, y, and theta to make sure their dimensions are correct.

Understanding $J(\theta)$

We'd like to understand better what gradient descent has done, and visualize the relationship between the parameters $\theta \in {\mathbb R}^2$ and $J(\theta)$. In this problem, we'll plot $J(\theta)$ as a 3D surface plot. (When applying learning algorithms, we don't usually try to plot $J(\theta)$ since usually $\theta \in {\mathbb R}^n$ is very high-dimensional so that we don't have any simple way to plot or visualize $J(\theta)$. But because the example here uses a very low dimensional $\theta \in {\mathbb R}^2$, we'll plot $J(\theta)$ to gain more intuition about linear regression.) Recall that the formula for $J(\theta)$ is

\begin{displaymath} J(\theta)=\frac{1}{2m}\sum_{i=1}^{m}\left(h_{\theta}(x^{(i)})-y^{(i)}\right)^{2} \nonumber \end{displaymath}

To get the best viewing results on your surface plot, use the range of theta values that we suggest in the code skeleton below.

J_vals = zeros(100, 100);   % initialize Jvals to 100x100 matrix of 0's
theta0_vals = linspace(-3, 3, 100);
theta1_vals = linspace(-1, 1, 100);
for i = 1:length(theta0_vals)
	  for j = 1:length(theta1_vals)
	  t = [theta0_vals(i); theta1_vals(j)];
	  J_vals(i,j) = %% YOUR CODE HERE %%

% Plot the surface plot
% Because of the way meshgrids work in the surf command, we need to 
% transpose J_vals before calling surf, or else the axes will be flipped
J_vals = J_vals'
surf(theta0_vals, theta1_vals, J_vals)
xlabel('\theta_0'); ylabel('\theta_1')

You should get a figure similar to the following. If you are using Matlab/Octave, you can use the orbit tool to view this plot from different viewpoints.

Surface plot

What is the relationship between this 3D surface and the value of $\theta_0$ and $\theta_1$ that your implementation of gradient descent had found?

Programming Exercise 1: Linear Regression Machine Learning

大家好,今天总结Coursera网课上Andrew Ng MachineLearning 第一次作业 1.computeCost.mfunction J = computeCost(X, y, th...
  • qq_21275321
  • qq_21275321
  • 2016年12月27日 20:57
  • 289

Programming Exercise5:Regularized Linera Regression and Bias v.s Variance

大家好,本文主要和大家分享coursera网站上斯坦福大学机器学习公开课(吴文达老师)第六周Reguirized Linera Regression and Bias v.s Variance的课后编...
  • a1015553840
  • a1015553840
  • 2016年03月01日 21:27
  • 1376

斯坦福大学机器学习公开课---Programming Exercise 1: Linear Regression

斯坦福大学机器学习公开课---Programming Exercise 1: Linear Regression 1  Linear regression with one variable ...
  • E_pen
  • E_pen
  • 2015年02月03日 23:10
  • 3355

[Exercise 1] Linear Regression

  • sherlockzoom
  • sherlockzoom
  • 2015年07月10日 15:41
  • 521

coursera机器学习课程——Programming Exercise 1:Linear Regression(选做部分)

  • sinat_38930882
  • sinat_38930882
  • 2017年09月11日 16:37
  • 105

Stanford 机器学习 Week6 作业:Regularized Linear Regression and Bias v.s. Variance

linearRegCostfunctionm = length(y); J = 0; grad = zeros(size(theta));J = 1.0 / 2 / m * ( sum( (X * ...
  • Baoli1008
  • Baoli1008
  • 2016年03月04日 12:28
  • 979

Machine Learning week 6 quiz: programming assignment-Regularized Linear Regression and Bias/Variance

一、ex5.m %% Machine Learning Online Class % Exercise 5 | Regularized Linear Regression and Bias-Var...
  • GarfieldEr007
  • GarfieldEr007
  • 2015年11月26日 13:26
  • 4987

Andrew Ng机器学习week6(Regularized Linear Regression and Bias/Variance)编程习题

Andrew Ng机器学习week6(Regularized Linear Regression and Bias/Variance)编程习题linearRegCostFunction.mfuncti...
  • u010043538
  • u010043538
  • 2017年06月02日 00:30
  • 770

coursera机器学习课程——Programming Exercise 1:Linear Regression

  • sinat_38930882
  • sinat_38930882
  • 2017年09月11日 16:10
  • 175

Coursera Machine Learning 第二周 quiz Programming Exercise 1: Linear Regression

warmUpExercise.m A = eye(5); computeCost.m
  • mupengfei6688
  • mupengfei6688
  • 2016年11月09日 19:47
  • 2563
您举报文章:Exercise: Linear Regression