Introduction.

Machine learning definition.

Why machine learning.

Applications of ML.

Learning paradigms.

Linear regression.

Cost function.

Gradient decent.

Machine learning definition

**Machine Learning**

Applications of ML

1- supervised learning

a- Regression example.

Arthur Samuel (1959) defined ML as the field of study that gives computers the ability to learn without being explicitly programmed.

Tom Mitchell (1998) defined well posed learning problem as : 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

.

Suppose your email program watches which email 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?

T

E

P

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.

Why machine learning ?

Need to make machines think and learn from mistakes like human.

To notice similarities between things and so generate new ideas.

Attempt to work out why things went wrong (Explanation).

Natural language processing

Search engines

Pattern recognition

Game playing

Robot locomotion

Medical diagnosis

Classifying DNA

Stock market analysis

Learning paradigms

1- Supervised learning

2- Unsupervised Learning

3- Reinforcement learning

- Learning to predict houses prices.

Suppose you collect statistics about how much houses cost according to the square footage of the house.

100

200

300

400

500

500

1000

1500

2000

2500

3000

0

Size in (feet)^2

Price in 1000's

a- Regression

b- Classification

1- Supervised learning

Given:

Data set (input & output)

Regression

---> the output you are trying to predict is a

continuous

value of the price.

b- Classification example.

- Let's say you collect a dataset on a cancer tumor and you want the program output to predict whether the certain tumor is malignant or benign.

+1

Malignant

Tumor size

More generally there's other examples that use more than 1 input variable (multi-i/p).

we can predict whether the tumor is malignant or benign through patient age, tumor size ..... so on.

Age

Tumor size

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.

Problem2:

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?

Treat both as classification problems.

Treat problem 1 as a classification problem, problem 2 as a regression problem.

Treat problem 2 as a classification problem, problem 1 as a regression problem.

Treat both as regression problems.

2- Unsupervised learning

Given : Inputs only

Required: Cluster data into groups

Supervised learning

Unsupervised learning

**Understanding genes data**

Social network analysis

Astronomical data analysis

Market segmentation

Cocktail party problem

Speaker 2

Speaker 1

Microphone 1

Microphone 2

Microphone1:

Output 1:

Output 2:

Output 1:

Output 2:

Microphone2:

Microphone1:

Microphone2:

Of the following examples, which would you address using an unsupervised learning algorithm? (Check all that apply.)

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 database of customer data, automatically discover market segments and group customers into different market segments.

Given a dataset of patients diagnosed as either having diabetes or not, learn to classify new patients as having diabetes or not.

3- Reinforcement learning

No data set

Have only critic

Credit assignment problem

If you write a program to make an autonomous helicopter to fly, you notice that if you make a wrong decision on a helicopter, the consequence of crashing it may not happen until much later.

Good helicopter

Bad helicopter

Reward function

Example :-

Linear regression with one variable

Training set of housing prices

Area(x)

Price(y)

2104

1416

1534

852

400

232

315

178

m:-

Number of training examples.

X:-

Input variable/feature.

Y:-

Output variable/target.

In order to design a learning algorithm, the first thing we have to decide is how we want to represent the hypothesis.

X

Y

Using linear representation for the hypothesis as :

h(x)=

Ɵ + Ɵ

0

1

x

Adding another input(feature)

h(x)= Ɵ + Ɵ X+ Ɵ X

1

1

2

2

0

**Any questions?**

malignant

benign

Linear representation

**h(x)=∑ Ɵ X**

i=0

n

i

i

h(x) :- Actual output

Ɵ ,Ɵ ,Ɵ :- Learning algorithm parameters

0

1Ɵii

2

n:- Number of features (input)

for simplicity, let X =1 ,so

0

How do we choose the parameter theta so that our hypothesis will make accurate prediction?

we want to make that squared difference between the predicted price and the actual price (cost function) as small as possible

J(Ɵ)=1/2∑(h (x )-y)

Ɵ

i

i=0

m

i

2

Our goal is to minimize the cost function. The algorithm for minimizing is

gradient descent algorithm.

Next Section :)

Checker Game

The checkers example,

E

T

P

10000s games

is playing checkers

if you win or not

Learning types

Inductive

Deductive

Learning steps

1- Training

2- Validation

3- Application

T

T

V

Data set

Classification

---> output here is

discrete

values 0 or 1