Introduction
UNIT I
BTCOC503 : Machine Learning
What is Learning?
• Learning is an ability to improve behavior and
knowledge based on experience.
BTCOC503 : Machine Learning
Machine Learning History
• 1950s:
– Arthur Samual Developed a program for
playing checkers.
– The program was able o observe the
positions of the game and learn a model that
gives better moves for the machine player.
– 1959: Samual Coined the term “Machine
Learning”
– Samual Defined machine learning as ”A field
of studies that gives computers the ability to
learn without being explicitly programmed”
BTCOC503 : Machine Learning
Machine Learning History
• 1957:
– Rosenblatt proposed the perceptron
– Perception is a simple neural network
unit.
– “Perceptron is designed to illustrate
some of the fundamental properties
of intelligent systems in general”
– Developed Linear classifier
BTCOC503 : Machine Learning
Machine Learning History
• 1969:
– Minsky ..Came up with Limitations of
perceptron.
– He showed that X-or problem could not be
represented by perceptron and such
inseparable data distribution cannot be
handled.
– And due to this up to 1980 the research on
Neural network get dormant
BTCOC503 : Machine Learning
Machine Learning History
• 1986:
– John Ross Quinlan came up with
decision tree learning ,specially
the Iterative Dichotomiser 3 ID3
algorithm.
– It was also released as a software and it
has very simplistic rules contrary to the
black box of neural network, which got
popular.
– After ID3 very interesting concepts were
developed.
– During this ‘type symbolic natural
language’ processing also become very
popular.
BTCOC503 : Machine Learning
Machine Learning History
BTCOC503 : Machine Learning
Neural
network Back
propagation
network
Machine Learning History
BTCOC503 : Machine Learning
Traditional Programming
Machine Learning
Computer
Data
Program
Output
BTCOC503 : Machine Learning
Computer
Data
Output
Program
/ Model
What is Machine Learning?
• Machine Learning is a concept which allows
the machine to learn from examples and
experience, and that too without being
explicitly programmed.
• It Explores algorithm-
– To learn and build the models for data
– Models are used for prediction ,decision making ,
solving problems etc.
BTCOC503 : Machine Learning
What is Machine Learning?
• Machine Learning is a concept which allows
the machine to learn from examples and
experience, and that too without being
explicitly programmed.
• It Explores algorithm-
– To learn and build the models for data
– Models are used for prediction ,decision making ,
solving problems etc.
BTCOC503 : Machine Learning
How Does SIRI work?
??????????
Type of Machine Learning?
• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning
• Semi-Supervised Learning
Supervised Learning
• The Model is able to Predict with the help of
labelled dataset.
• For every instance there is a corresponding
output.
• Somewhat like Learning with Teacher..
• What is labelled dataset?
– Data for which you already know the Target
answer is called as labelled dataset.
Supervised Learning
Supervised Learning
Types of Supervised Learning
• Classification
–Discrete valued Y
–Machine is trained to classify something into
some class
–Ex. Patient has disease or not
–Ex. Email is spam or not
Supervised Learning: Classification
Types of Supervised Learning
• Regression
– Continuous valued Y
– Machine is trained to predict some value like
price, weight, or height.
Supervised Learning: Classification
Unsupervised Learning
Classification Learning
• Task T:
– Input
– Output
• Performance metric P
• Experience E
How do you get data for the Machine Learning Problem?
How do you get data for the Machine Learning Problem?
Hypothesis Space
• Set of candidate outputs that you can get.
• In Supervised learning there is set of
functions that comprises hypothesis space
and you want to find out that function from
the hypothesis space which is most probable
given your training examples
introduction to machine Learning
introduction to machine Learning

introduction to machine Learning

  • 1.
  • 2.
    What is Learning? •Learning is an ability to improve behavior and knowledge based on experience. BTCOC503 : Machine Learning
  • 3.
    Machine Learning History •1950s: – Arthur Samual Developed a program for playing checkers. – The program was able o observe the positions of the game and learn a model that gives better moves for the machine player. – 1959: Samual Coined the term “Machine Learning” – Samual Defined machine learning as ”A field of studies that gives computers the ability to learn without being explicitly programmed” BTCOC503 : Machine Learning
  • 4.
    Machine Learning History •1957: – Rosenblatt proposed the perceptron – Perception is a simple neural network unit. – “Perceptron is designed to illustrate some of the fundamental properties of intelligent systems in general” – Developed Linear classifier BTCOC503 : Machine Learning
  • 5.
    Machine Learning History •1969: – Minsky ..Came up with Limitations of perceptron. – He showed that X-or problem could not be represented by perceptron and such inseparable data distribution cannot be handled. – And due to this up to 1980 the research on Neural network get dormant BTCOC503 : Machine Learning
  • 6.
    Machine Learning History •1986: – John Ross Quinlan came up with decision tree learning ,specially the Iterative Dichotomiser 3 ID3 algorithm. – It was also released as a software and it has very simplistic rules contrary to the black box of neural network, which got popular. – After ID3 very interesting concepts were developed. – During this ‘type symbolic natural language’ processing also become very popular. BTCOC503 : Machine Learning
  • 7.
    Machine Learning History BTCOC503: Machine Learning Neural network Back propagation network
  • 8.
  • 9.
    Traditional Programming Machine Learning Computer Data Program Output BTCOC503: Machine Learning Computer Data Output Program / Model
  • 10.
    What is MachineLearning? • Machine Learning is a concept which allows the machine to learn from examples and experience, and that too without being explicitly programmed. • It Explores algorithm- – To learn and build the models for data – Models are used for prediction ,decision making , solving problems etc. BTCOC503 : Machine Learning
  • 11.
    What is MachineLearning? • Machine Learning is a concept which allows the machine to learn from examples and experience, and that too without being explicitly programmed. • It Explores algorithm- – To learn and build the models for data – Models are used for prediction ,decision making , solving problems etc. BTCOC503 : Machine Learning
  • 12.
    How Does SIRIwork? ??????????
  • 14.
    Type of MachineLearning? • Supervised Learning • Unsupervised Learning • Reinforcement Learning • Semi-Supervised Learning
  • 16.
    Supervised Learning • TheModel is able to Predict with the help of labelled dataset. • For every instance there is a corresponding output. • Somewhat like Learning with Teacher.. • What is labelled dataset? – Data for which you already know the Target answer is called as labelled dataset.
  • 17.
  • 18.
  • 19.
    Types of SupervisedLearning • Classification –Discrete valued Y –Machine is trained to classify something into some class –Ex. Patient has disease or not –Ex. Email is spam or not
  • 20.
  • 21.
    Types of SupervisedLearning • Regression – Continuous valued Y – Machine is trained to predict some value like price, weight, or height.
  • 22.
  • 23.
  • 29.
    Classification Learning • TaskT: – Input – Output • Performance metric P • Experience E
  • 33.
    How do youget data for the Machine Learning Problem?
  • 34.
    How do youget data for the Machine Learning Problem?
  • 38.
    Hypothesis Space • Setof candidate outputs that you can get. • In Supervised learning there is set of functions that comprises hypothesis space and you want to find out that function from the hypothesis space which is most probable given your training examples