This document discusses machine learning and linear regression. It provides examples of supervised learning problems like predicting housing prices and classifying cancer as malignant or benign. Unsupervised learning is used to discover patterns in unlabeled data, like grouping customers for market segmentation. Linear regression finds the linear function that best fits some training data to make predictions on new data. It can be extended to nonlinear functions by adding polynomial features. More complex models may overfit the training data and not generalize well to new examples.

1. Machine Learning & Linear
Regression
Faculty of Computing and
Information Technology
CPIS-703: Intelligent Information Systems
and
Decision Support
Department of Information Science

2. Machine Learning
- Grew out of work in AI
- New capability for computers
Examples:
- Database mining
Large datasets from growth of automation/web.
E.g., Web click data, medical records, biology,
engineering
- Applications can’t program by hand.
E.g., Autonomous helicopter, handwriting recognition, most
of Natural Language Processing (NLP), Computer Vision.

3. Machine Learning
- Grew out of work in AI
- New capability for computers
Examples:
- Database mining
Large datasets from growth of automation/web.
E.g., Web click data, medical records, biology,
engineering
- Applications can’t program by hand.
E.g., Autonomous helicopter, handwriting recognition, most
of Natural Language Processing (NLP), Computer Vision.
- Self-customizing programs
E.g., Amazon, Netflix product recommendations
- Understanding human learning (brain, real AI).

4. • Arthur Samuel (1959). Machine Learning:
Field of study that gives computers the ability
to learn without being explicitly programmed.
• Tom Mitchell (1998) Well-posed Learning
Problem: A computer program is said to learn
from experience E with respect to some task
T and some performance measure P, if its
performance on T, as measured by P,
improves with experience E.
Machine Learning definition

5. Classifying emails as spam or not spam.
Watching you label emails as spam or not spam.
The number (or fraction) of emails correctly classified as spam/not spam.
None of the above—this is not a machine learning problem.
Suppose your email program watches which emails you do or
do not mark as spam, and based on that learns how to better
filter spam. What is the task T in this setting?
“A computer program is said to learn from experience E with respect
to some task T and some performance measure P, if its performance
on T, as measured by P, improves with experience E.”

6. Machine learning algorithms:
- Supervised learning
- Unsupervised learning
Others: Reinforcement learning,
recommender systems.
Also talk about: Practical advice for applying
learning algorithms.

7. Machine learning Review

16. Supervised Learning

17. 0
100
200
300
400
0 500 1000 1500 2000 2500
Housing price prediction
Price ($)
in 1000’s
Size in feet2
Regression: Predict
continuous valued
output (price)
Supervised Learning
“right answers” given

18. Cancer (malignant, benign)
Classification
Discrete
valued output
(0 or 1)
Malignant?
1(Y)
0(N)
Tumor Size
Tumor Size

19. Tumor Size
Age
- Clump Thickness
- Uniformity of Cell
Size
- Uniformity of Cell
Shape
…

20. Treat both as classification problems.
Treat problem 1 as a classification problem, problem 2 as a regression
problem.
Treat problem 1 as a regression problem, problem 2 as a classification
problem.
Treat both as regression problems.
You’re running a company, and you want to develop learning algorithms to address
each of two problems.
Problem 1: You have a large inventory of identical items. You want to predict how
many of these items will sell over the next 3 months.
Problem 2: You’d like software to examine individual customer accounts, and for
each account decide if it has been hacked/compromised.
Should you treat these as classification or as regression problems?

21. Unsupervised Learning

22. x1
x2
Supervised Learning

23. Unsupervised Learning
x1
x2

26. [Source: Daphne Koller]
Genes
Individuals

27. [Source: Daphne Koller]
Genes
Individuals

28. Organize computing clusters Social network analysis
Image credit: NASA/JPL-Caltech/E. Churchwell (Univ. of Wisconsin, Madison)
Astronomical data analysis
Market segmentation

29. Of the following examples, which would you address
using an unsupervised learning algorithm? (Check
all that apply.)
Given a database of customer data, automatically
discover market segments and group customers into
different market segments.
Given email labeled as spam/not spam, learn a spam
filter.
Given a set of news articles found on the web, group
them into set of articles about the same story.
Given a dataset of patients diagnosed as either having
diabetes or not, learn to classify new patients as having
diabetes or not.

30. Supervised learning
• Notation
– Features x
– Targets y
– Predictions ŷ
– Parameters q
Program (“Learner”)
Characterized by
some “parameters” θ
Procedure (using θ)
that outputs a prediction
Training data
(examples)
Features
Learning algorithm
Change θ
Improve performance
Feedback /
Target values Score performance
(“cost function”)

31. Linear regression
• Define form of function f(x) explicitly
• Find a good f(x) within that family
0 10 20
0
20
40
Target
y
Feature x
“Predictor”:
Evaluate line:
return r

32. More dimensions?
0
10
20
30
40
0
10
20
30
20
22
24
26
0
10
20
30
40
0
10
20
30
20
22
24
26
x1 x2
y
x1 x2
y

33. Notation
Define “feature” x0 = 1 (constant)
Then

34. Measuring error
0 20
0
Error or “residual”
Prediction
Observation

35. Mean squared error
• How can we quantify the error?
• Could choose something else, of course…
– Computationally convenient (more later)
– Measures the variance of the residuals
– Corresponds to likelihood under Gaussian model of “noise”

36. MSE cost function
• Rewrite using matrix form
(Matlab) >> e = y’ – th*X’; J = e*e’/m;

37. Visualizing the cost function
-1 -0.5 0 0.5 1 1.5 2 2.5 3
-40
-30
-20
-10
0
10
20
30
40
-1 -0.5 0 0.5 1 1.5 2 2.5 3
-40
-30
-20
-10
0
10
20
30
40
θ1
J(θ)

38. Supervised learning
• Notation
– Features x
– Targets y
– Predictions ŷ
– Parameters q
Program (“Learner”)
Characterized by
some “parameters” θ
Procedure (using θ)
that outputs a prediction
Training data
(examples)
Features
Learning algorithm
Change θ
Improve performance
Feedback /
Target values Score performance
(“cost function”)

39. Finding good parameters
• Want to find parameters which minimize our error…
• Think of a cost “surface”: error residual for that θ…

40. Linear regression: direct minimization
+

41. MSE Minimum
• Consider a simple problem
– One feature, two data points
– Two unknowns: µ0, µ1
– Two equations:
• Can solve this system directly:
• However, most of the time, m > n
– There may be no linear function that hits all the data exactly
– Instead, solve directly for minimum of MSE function

42. SSE (Sum of squared errors) Minimum
• Reordering, we have
• X (XT X)-1 is called the “pseudo-inverse”
• If XT is square and independent, this is the inverse
• If m > n: overdetermined; gives minimum MSE fit

43. Matlab SSE
• This is easy to solve in Matlab…
% y = [y1 ; … ; ym]
% X = [x1_0 … x1_m ; x2_0 … x2_m ; …]
% Solution 1: “manual”
th = y’ * X * inv(X’ * X);
% Solution 2: “mrdivide”
th = y’ / X’; % th*X’ = y => th = y/X’
“matrix-right-divide”

44. 0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
16
18
Effects of MSE choice
• Sensitivity to outliers
16 2 cost for this one datum
Heavy penalty for large errors
-20 -15 -10 -5 0 5
0
1
2
3
4
5

45. L1 error (minimum absolute error )
0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
16
18 L2, original data
L1, original data
L1, outlier data

46. Cost functions for regression
“Arbitrary” functions can’t be
solved in closed form…
- use gradient descent
(MSE)
(MAE)
Something else entirely…
(???)

47. Linear regression: nonlinear features
+

48. Nonlinear functions
• What if our hypotheses are not lines?
– Ex: higher-order polynomials
0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
16
18
Order 1 polynom ial
0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
16
18
Order 3 polynom ial

49. Nonlinear functions
• Single feature x, predict target y:
• Sometimes useful to think of “feature transform”
Add features:
Linear regression in new features

50. Higher-order polynomials
• Fit in the same way
• More “features”
0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
16
18
Order 1 polynom ial
0 2 4 6 8 10 12 14 16 18 20
-2
0
2
4
6
8
10
12
14
16
18
Order 2 polynom ial
0 2 4 6 8 10 12 14 16 18 20
0
2
4
6
8
10
12
14
16
18
Order 3 polynom ial

51. Features
• In general, can use any features we think are useful
• Other information about the problem
– Sq. footage, location, age, …
• Polynomial functions
– Features [1, x, x2, x3, …]
• Other functions
– 1/x, sqrt(x), x1 * x2, …
• “Linear regression” = linear in the parameters
– Features we can make as complex as we want!

52. Higher-order polynomials
• Are more features better?
• “Nested” hypotheses
– 2nd order more general than 1st,
– 3rd order “ “ than 2nd, …
• Fits the observed data better

53. Overfitting and complexity
• More complex models will always fit the training data
better
• But they may “overfit” the training data, learning
complex relationships that are not really present
X
Y
Complex model
X
Y
Simple model

54. Test data
• After training the model
• Go out and get more data from the world
– New observations (x,y)
• How well does our model perform?
Training data
New, “test” data

55. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
5
10
15
20
25
30
Training data
Training versus test error
• Plot MSE as a function
of model complexity
– Polynomial order
• Decreases
– More complex function
fits training data better
• What about new data?
Mean
squared
error
Polynomial order
New, “test” data
• 0th to 1st order
– Error decreases
– Underfitting
• Higher order
– Error increases
– Overfitting