You are on page 1of 10

Machine Learning Session 1

Machine Learning Definition :


Arthur Samuel :
Field of study that gives computers the ability to learn without
being explicitly programmed.
Tom Mitchell:
Well Posed Learning Problem A computer 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
Supervised Learning
Lets start by talking about a few examples of supervised learning problems.
Suppose we have a dataset giving the living areas and prices of 47 houses from
Portland, Oregon:
Living area (feet2) Price (1000$s)
2104 400
1600 330
2400 369
1416 232
3000 540


Here in this case we Supervise the algorithm by providing the data so we call it as
Supervised Learning Algorithm
Mapping
Regression Problem
When the target variable we would like to predict from the learning algorithm is
continuous we call the learning problem a Regression Problem .

If the predicted value can take only a small number of discrete values we call it a
CLASSIFICATION PROBLEM
Learning Theory
It help in finding the Correct amount of data to be collected , nothing less ,
nothing more.
Most of the times we may end up with collection of less amount of data .
On the other hand we have possibility of gather too much data which is redundant
or with minimal data the conclusion can be derived, so the data is of NO USE
Unsupervised Learning
From the date we have to find the formats and structures which can derive
meaningful conclusions and working examples for us .

Example: Using the Clustering mechanism with Unsupervised Learning they developed
a 3D vision from a normal photograph using a pixel mapping algorithm
Applications
Unsupervised algorithm is also applied in many places few examples are
1) Organized Computing Clusters
2) Market Segmentation
3)Social Networking Analysis
4)Astronomical data analysis
5)Text Processing is also another area of application

Cocktail Party Problem
There are many people in the Party and out of all we would like to separate the
voice of one person from group of many people who all attended the party.
ICA algorithm in Mat Lab or Octave can solve this problem in few lines of code.
Reinforcement Learning
It is a sequence of decisions.
A sequence of good decisions will lead to success.
A sequence of bad decisions will lead to failure.
When applied on a helicopter to automate, a series of good decision will allow it to
fly on the other hand series of bad decisions will allow it to crash.
Robotics is another area of Application for Reinforcement Learning where it learns
by series and sequence of actions.
Thank you
End of Session One

You might also like