Linear

Regression
Linear Regression
• Linear regression, or ordinary least squares (OLS), is the simplest and most classic linear method
for regression.
• Linear regression finds the parameters w and b that minimize the mean squared error between
predictions and the true regression targets, y, on the training set.
• The mean squared error is the sum of the squared differences between the predictions and the
true values.
• With higher-dimensional datasets, linear models become more powerful, and there is a higher
chance of overfitting.
• We should try to find a model that allows us to control complexity.
 Linear regression
with one variable
Model
representation
Machine Learning
 500000
Housing Prices 400000
(Portland, OR)
300000

Price 200000

(in 1000s 100000

of dollars)
0
500 1000 1500 2000 2500 3000

Size (feet2)
Supervised Learning Regression Problem
Given the “right answer” for Predict real-valued output
each example in the data.
Training set of Size in feet2 (x) Price ($) in 1000's (y)
housing prices 2104 460
(Portland, OR) 1416 232
1534 315
852 178
… …
Notation:
m = Number of training examples
x’s = “input” variable / features
y’s = “output” variable / “target” variable

(x, y) – one training example

(,
 Training Set How do we represent h ?
h 𝜃 ( 𝑥 ) =𝜃 0 + 𝜃 1 𝑥

Learning Algorithm

x
x x x
Size of h Estimated y x x
house price
x hypothesis Estimated y x
Linear regression with one variable.
h maps from x’s to y’s Univariate linear regression.
 Linear regression
with one variable
Cost function

Machine Learning
 Size in feet2 (x) Price ($) in 1000's (y)
Training Set 2104 460
1416 232
1534 315 m = 47

852 178
… …

Hypothesis:
‘s: Parameters
How to choose ‘s ?
3 3 3

2 2 2

1 1 1

0 0 0
0 1 2 3 0 1 2 3 0 1 2 3
 y

Idea: Choose so that

is close to
for our training examples
 Linear regression
with one variable
Cost function
intuition I
Machine Learning
 Simplified
Hypothesis:

Parameters:

Cost Function:

Goal:
 (for fixed , this is a function of x) (function of the parameter )

3 3

2 2

y
1 1

0 0
0 1 2 3 -0.5 0 0.5 1 1.5 2 2.5
x
 (for fixed , this is a function of x) (function of the parameter )

3 3

2 2

y
1 1

0 0
0 1 2 3 -0.5 0 0.5 1 1.5 2 2.5
x
 (for fixed , this is a function of x) (function of the parameter )

3 3

2 2

y
1 1

0 0
0 1 2 3 -0.5 0 0.5 1 1.5 2 2.5
x
 Linear regression
with one variable
Cost function
intuition II
Machine Learning
Hypothesis:

Parameters:

Cost Function:

Goal:
 (for fixed , this is a function of x) (function of the parameters )

500000

400000

Price ($) 300000

in 1000’s
200000

100000

0
500 1000 1500 2000 2500 3000
Size in feet (x)
2
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
 Linear regression
with one variable
Gradient
descent
Machine Learning
Have some function
Want

Outline:
• Start with some
• Keep changing to reduce
until we hopefully end up at a minimum
J(0,1)

1
0
J(0,1)

1
0
Gradient descent algorithm

Correct: Simultaneous update Incorrect:

 Linear regression
with one variable
Gradient descent
intuition
Machine Learning
Gradient descent algorithm
If α is too small, gradient descent
can be slow.

If α is too large, gradient descent

can overshoot the minimum. It may
fail to converge, or even diverge.
 at local optima

Current value of
Gradient descent can converge to a local
minimum, even with the learning rate α fixed.

As we approach a local
minimum, gradient
descent will automatically
take smaller steps. So, no
need to decrease α over
time.
 Linear regression
with one variable
Gradient descent for
linear regression
Machine Learning
Gradient descent algorithm Linear Regression Model
Gradient descent algorithm

update
and
simultaneously
J(0,1)

1
0
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
(for fixed , this is a function of x) (function of the parameters )
“Batch” Gradient Descent

“Batch”: Each step of gradient descent

uses all the training examples.
 Linear Regression with
multiple variables

Multiple features

Machine Learning
Multiple features (variables).

Size (feet2) Price ($1000)

2104 460
1416 232
1534 315
852 178
… …
Multiple features (variables).
Size (feet2) Number of Number of Age of home Price ($1000)
bedrooms floors (years)

2104 5 1 45 460
1416 3 2 40 232
1534 3 2 30 315
852 2 1 36 178
… … … … …
Notation:
= number of features
= input (features) of training example.
= value of feature in training example.
For convenience of notation, define .

1 * (n+1)
Multivariate linear regression.
 Linear Regression with
multiple variables

Gradient descent for

multiple variables

Machine Learning
Hypothesis:
Parameters:
Cost function:

Gradient descent:
Repeat

(simultaneously update for every )

 New algorithm :
Gradient Descent
Repeat
Previously (n=1):
Repeat
(simultaneously update for

(simultaneously update )
 Linear Regression with
multiple variables
Gradient descent in
practice I: Feature Scaling

Machine Learning
Feature Scaling
Idea: Make sure features are on a similar scale.
E.g. = size (0-2000 feet2)
= number of bedrooms (1-5)
Feature Scaling
Get every feature into approximately a range.
Mean normalization
Replace with to make features have approximately zero mean
(Do not apply to ).
E.g.
 Linear Regression with
multiple variables

Gradient descent in
practice II: Learning rate
Machine Learning
Gradient descent

- “Debugging”: How to make sure gradient

descent is working correctly.
- How to choose learning rate .
Making sure gradient descent is working correctly.

Example automatic
convergence test:

Declare convergence if
decreases by less
than in one
0 100 200 300 400
iteration.
No. of iterations
Making sure gradient descent is working correctly.
Gradient descent not working.
Use smaller .

No. of iterations

No. of iterations No. of iterations

- For sufficiently small , should decrease on every iteration.

- But if is too small, gradient descent can be slow to converge.
Summary:
- If is too small: slow convergence.
- If is too large: may not decrease on
every iteration; may not converge.

To choose , try
 Linear Regression with
multiple variables

Features and
polynomial regression
Machine Learning
Housing prices prediction
 Polynomial regression

Price
(y)

Size (x)
Choice of features

Price
(y)

Size (x)
 Linear Regression with
multiple variables

Normal equation

Machine Learning
Gradient Descent

Normal equation: Method to solve for

analytically.
Intuition: If 1D

(for every )

Solve for
Examples:
Size (feet2) Number of Number of Age of home Price ($1000)
bedrooms floors (years)

1 2104 5 1 45 460
1 1416 3 2 40 232
1 1534 3 2 30 315
1 852 2 1 36 178
 examples ; features.

E.g. If
 is inverse of matrix .

Octave: pinv(X’*X)*X’*y
 training examples, features.
Gradient Descent Normal Equation
• Need to choose . • No need to choose .
• Needs many iterations. • Don’t need to iterate.
• Works well even • Need to compute
when is large.
• Slow if is very large.
Ridge Regression (L2 Regularization)
• We also want the magnitude of coefficients to be as small as possible (small Slopes).
• Intuitively, this means each feature should have as little effect on the outcome as possible, while
still predicting well.
• This constraint is an example of what is called regularization.
• Regularization means explicitly restricting a model to avoid overfitting.
• A regularization parameter alpha controls the value of coefficients.
Lasso (Least Absolute Shrinkage and
Selection Operator) (L1 Regularization)
Logistic Regression

0
Logistic Regression
• In case of a binary classification problem where 𝑦 𝜖 {0 ,1 }

• Threshold of classifier is at = 0.5

• If > 0.5, y = 1
• If < 0.5, y = 0
• Logistic regression:
Logistic Regression
• In linear regression:
• In logistic regression:
• Where (sigmoid function or the logistic function)
• Hence the name logistic regression
• But it is a classifier that is extended from linear regression
• Finally
Logistic Regression

1
• Task is to select parameters w to fit the date

0.5

0
Logistic Regression
• estimated probability that y = 1 on input x
• For example, if

And
70% chance of tumor being malignant.

• , prob that y=1 given x, parameterized by w.

• =1
• =1-
Logistic Regression
Decision Boundary
(Linear Case)
x2
Predict y = 1 if

𝑥1 + 𝑥 2 =3
A linear decision boundary obtained from
linear regression model.

x1
Logistic Regression
Decision Boundary
(Nonlinear Case)
Predict y = 1 if
Logistic Regression
Decision Boundary
(Nonlinear Case)
1
Predict y = 1 if
y=1
y=0
-1
 x2

1 y=1

y=0

-1 x1
 Logistic Regression
Cost Function
Cost that a learning algo (hypothesis) has to pay if its prediction is h(x) when the actual label is y
𝑚
1 1 𝑖 2
Linear Regression Model: 𝐽 ( 𝑤 )= ∑ ( h𝑤 ( 𝑥 ) − 𝑦 )
𝑖
(MSE: Meane Squared Error)
𝑚 𝑖=1 2
Cost ((x),y) =

If we use the same cost function for logistic regression whose hypothesis is a nonlinear
function, it will result in a nonconvex J(w).
Gradient Descent

= (old)

Parameter > 0. plays an important role in the

convergence of the algorithm.
If is too small, the correction is small and
convergence to the optimum point is slow.
If is too large, the algorithm may oscillate around
the optimum value and convergence is not possible.
Geometric interpretation of parameter optimization.
In the gradient descent scheme, the
correction of the parameters takes place
in the direction that decreases the value
of the cost function.
• If is properly chosen, the algorithm converges to
a stationary point of J() which can be
• A local minimum
• A global minimum
• Or a saddle point
• To which of the stationary points the algorithm
will converge depends on the position of the
initial point, relative to the stationary points.
• Furthermore, the convergence speed depends on
the form of the cost J().
 Logistic Regression
Cost Function for Logistic Regression
Logistic Regression Cost ((x),y) =
Model:

Y=1
Cost = 0 if y = 1, h = 1
But as h  0
Cost  infinity
That is if h = 0
P(y=1|x;w) = 0, but y = 1
We’ll penalize the learning algorithm by a large cost.

(x) 1
 Logistic Regression
Cost Function for Logistic Regression
Logistic Regression Cost ((x),y) =
Model:

Y=0

(x) 1
Logistic Regression
Cost Function for Logistic Regression
𝑚
1
𝐽 ( 𝑤 )= ∑ 𝐶𝑜𝑠𝑡(h𝑤 ( 𝑥 𝑖 ), 𝑦 𝑖)
𝑚 𝑖=1
Cost ((x),y) =

Cost ((x),y) =
1
𝐽 ( 𝑤 )= ¿
𝑚
• This cost function is derived in Statistics from the idea of maximum likelihood estimation
which helps to efficiently find parameters for different models
Logistic Regression
Cost Function for Logistic Regression
𝑚
1
𝐽 ( 𝑤 )= ∑ 𝐶𝑜𝑠𝑡(h𝑤 ( 𝑥 𝑖 ), 𝑦 𝑖)
𝑚 𝑖=1
• To fit parameter w
min 𝐽 ( 𝑤 ) 𝑡𝑜 𝑔𝑒𝑡 𝒘 𝒃𝒖𝒕 𝒉𝒐𝒘 ? ?
𝑤

• To make a prediction given new data (x)

𝟏
h𝑤 ( 𝑥)= − 𝑤
𝑇
𝒙
𝟏+ 𝒆
Logistic Regression
Cost Function for Logistic Regression

min 𝐽 ( 𝜃 ) 𝑖𝑠 𝑎𝑐h𝑖𝑒𝑣𝑒𝑑 𝑡h𝑟𝑜𝑢𝑔h 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡 𝑑𝑒𝑠𝑐𝑒𝑛𝑡

𝑤

1
𝐽 ( 𝜃)= ¿
𝑚