Scribd
Upload
70 views3 pages

Machine Learning by Andrew NG Lecture Notes: Vinod Kumar November 7, 2016

Uploaded by

Karthik

This document provides an introduction to machine learning, including definitions of supervised and unsupervised learning. It discusses how machine learning gives computers the ability to learn without being explicitly programmed. Supervised learning uses labeled training data to predict discrete or continuous output values, while unsupervised learning finds structure in unlabeled data by clustering it. Linear regression is given as an example of supervised learning, with the goal of finding hypothesis parameters that minimize a cost function measuring error between predictions and actual values.

Copyright:

© All Rights Reserved
Download as PDF, TXT or read online from Scribd
Download as pdf or txt

Machine Learning by Andrew NG Lecture Notes: Vinod Kumar November 7, 2016

Uploaded by

Karthik
70 views3 pages
This document provides an introduction to machine learning, including definitions of supervised and unsupervised learning. It discusses how machine learning gives computers the ability to learn without being explicitly programmed. Supervised learning uses labeled training data to predict discrete or continuous output values, while unsupervised learning finds structure in unlabeled data by clustering it. Linear regression is given as an example of supervised learning, with the goal of finding hypothesis parameters that minimize a cost function measuring error between predictions and actual values.

Original Description:

ml

Original Title

5B

Copyright

© © All Rights Reserved

Available Formats

PDF, TXT or read online from Scribd

Did you find this document useful?

Is this content inappropriate?

This document provides an introduction to machine learning, including definitions of supervised and unsupervised learning. It discusses how machine learning gives computers the ability to learn without being explicitly programmed. Supervised learning uses labeled training data to predict discrete or continuous output values, while unsupervised learning finds structure in unlabeled data by clustering it. Linear regression is given as an example of supervised learning, with the goal of finding hypothesis parameters that minimize a cost function measuring error between predictions and actual values.

Copyright:

© All Rights Reserved
Download as PDF, TXT or read online from Scribd
Download as pdf or txt
70 views3 pages

Machine Learning by Andrew NG Lecture Notes: Vinod Kumar November 7, 2016

Uploaded by

Karthik
This document provides an introduction to machine learning, including definitions of supervised and unsupervised learning. It discusses how machine learning gives computers the ability to learn without being explicitly programmed. Supervised learning uses labeled training data to predict discrete or continuous output values, while unsupervised learning finds structure in unlabeled data by clustering it. Linear regression is given as an example of supervised learning, with the goal of finding hypothesis parameters that minimize a cost function measuring error between predictions and actual values.

Copyright:

© All Rights Reserved
Download as PDF, TXT or read online from Scribd
Download as pdf or txt
You are on page 1/ 3

Machine Learning by Andrew NG Lecture Notes

Vinod Kumar

November 7, 2016

Introduction
What is Machine Learning
Field of study that gives computers the ability to learn without being ex-
plicitly programmed.

Why Machine Learning


Many programs cant be hardcoded.For example Autonomous heli-
copter,handwriting recognition, most of Natural Language Processing
(NLP), Computer Vision etc.
User based recommendations cant be done without machine learn-
ing.As each user requires a different program,it is not quite possible
to write so many programs for so many people.Ex Amazon,Netflix etc

Types Of Machine Learning


Supervised Machine Learning :-

1
An Example: Consider the data set as described in the image above.To
find the price of a new house,from its size we can try to fit a straight through
the data and find out the price of the house based on this straight line.We
can also use a quadratic or a second order polynomial for the same problem
to get a different answer.This is an example of Supervised Learning.

The term supervised learning refers to the fact that we use a data set in
which the right answers were given.In the above example we used a data set
of houses in which for every example in this data set, we told it what is the
right price and the task of the algorithm was just to produce more of these
right answers for new instances.To be a bit more precise the above example
is what we call a regression problem which is a type of supervised problem
where we try to predict a continuous value output.

In case the data set is discrete then the problem is a classification.One


example is when you try to predict whether a tumour is malignant or benign
according to the tumour size using a data set with tumour sizes and the type
of tumour.

Unsupervised Learning :-
The term unsupervised learning refers to the fact that we gave the algorithm
a data set in which we only specify the features of the instances and not the
type of the instance.The algorithm should find the structure in the data and
draw conclusions from it.One way to do is to divide the given instances in
the data sets into clusters and when a new instance is given the learning
algorithm can predict which cluster this new instance belongs to.This is
known as clustering algorithm.

An Example: Google news collects many new stories on the web and
groups them together into different categories using this algorithm.This is
done so that all stories about a single incident are grouped together.Also this
technique is used in social network analysis where given knowledge about
which friends you email the most or given your facebook friends or your
Google+ circles, we can automatically identify which are cohesive groups of
friends.

An interesting observation The problems from the above examples can


also be done by supervised learning if given a proper data set ,i.e. even
supervised learning can make clusters out of labelled data similar to the

2
ones made by unsupervised learning.This is one example where supervised
learning and unsupervised learning can be used interchangeably.In fact this
is true in most cases.The type of algorithm is mostly chosen according to
the type of data available.

Model Representation of a Linear Regression


Consider the price of the house example described under Supervised Al-
gorithm.We will use the same problem with a bigger data set.The data set
used for learning in an algorithm is called as training set.In practice the ob-
jective of the learning algorithm is to find the hypothesis of the problem.In
this case a hypothesis function h (x) that can map the size of a house to
its price using the given data. For our example we will find the hypothesis
by fitting a straight line through our training set.

h (x) = 0 + 1 x

Cost Function
There can be many values for 0 and 1 .To find the best possible hypothesis
we use those values which give the least possible cost function.The cost
function generally refers to how close the approximations are to the original
values.In this case the cost function is
m
1 X
J(0 , 1 ) = (h (xi ) yi )
2m
i=1

where xi and yi represent the original size and price of the houses,h (xi )
refers to the hypothesis function.

You might also like