2. Machine Learning 2
What is Learning?
“Learning is a process that leads to change, which occurs as a
result of experience and increases the potential for improved
performance and future learning.” 1
1. Ambrose et al, 2010. How Learning Works: Seven Research-Based Principles for Smart Teaching
3. Machine Learning 3
What is Learning?
“Learning is any relatively permanent change in behavior that
occurs as a result of experience.”
(S. P. Robbins)
“Learning is the modification of behavior through experience
and training.”
(Biswanath Ghosh)
4. Machine Learning 4
Learning Examples
• Eat
• Walk
• Read
• Drive
• Recognize
Can you think of any other 5 learning examples?
5. Machine Learning 5
Learning Experience
• How humans learn?
Wheels # Wheels Doors Window
Yes 4 Yes Yes
CAR CAR
CAR
6. Machine Learning 6
What is Machine Learning?
Machine Learning is the subfield of computer science that
gives “computers the ability to learn without being explicitly
programmed .”
(Arthur Samuel, 1959)
• To improve the performance of programs based on
given data, previous results, or experiences
– Developing methods to extract knowledge from examples
– Methods for creating computer programs by the analysis of data
sets
8. Machine Learning 8
Key Ingredients
• Data
• Experience
• Learning Model
Data is cheap and abundant (data warehouses, data
marts); knowledge is expensive and scarce
10. Machine Learning 10
Types of ML Problems (1)
• Classification
Voice/Face/Fingerprint/Iris/DNA/Signature recognition
Recommendation,
Spam filter
Credit card fraud detection
• Regression
Stock market prediction
House price prediction
• Clustering
Web-search, Document & information retrieval
User segmentation
• Strategy Learning
Games
• Association
POS Analysis
11. Machine Learning 11
Types of ML Problems (2)
• Classification - 1
Each given data has its own class or label
Once a query is given, a system should tell the class of the query
For example: Security Gate
ORL dataset, AT&T Laboratories, Cambridge UK
Permitted
Persons
Query: Permitted or Not?
12. Machine Learning 12
Types of ML Problems (6)
• Regression - 2
The process of predicting continuous values
House ID Area (sq. ft) # Bed # Bath Price
1 2700 2 2 5000000
2 2000 3 4 5500000
3 2200 3 3 6000000
4 1500 2 2 3500000
5 1800 3 2 4000000
6 1200 2 1 3000000
7 2500 4 4 6500000
House Price dataset
13. Machine Learning 13
Types of ML Problems (7)
• Clustering - 1
A set of un-labeled data is given
Your program should group the data (Finding hidden structure of
data)
If a query is given, your program should determine the group in
which the query belongs to
?
14. Machine Learning 14
Types of Learning Methods
• Supervised Learning
• Classification, Regression
• All given data is labeled
• We “teach the model”, then with that knowledge, it can predict the unknown
or future instances
• Unsupervised Learning
• Clustering, Dimension Reduction, Association
• Data is not labeled
• The model works on its own to discover information
• Semi-supervised Learning
• Classification, Clustering
• Some data is labeled, and some is not
• Reinforcement Learning
• Strategy Learning
• Reward is given to your behaviors