Artificial Intelligence

Chapter 5: Machine Learning

 Outline

 Why Machine Learning (ML)?

 Examples of ML applications
 Major classes of ML Algorithms
 ML Vs AI
 Questions about ML

1/24/2024 2
 Why Machine Learning ?

 It is very hard to write programs that solve complex problem.

(Defining the requirements)
 Computing the probability of a credit card transaction is
fraudulent.
 Recognizing a three-dimensional objects.

 We don’t know what program to write because we don’t know

how its done in our brain. (tacit knowledge not explicit)
 Even if we had a good idea about how to do it, the program
might be terribly complicated.
1/24/2024 3
 Why Machine Learning ?

 There may not be rules that are both simple and reliable.
 We need to combine a very large number of weak rules.

 May be the rules are changing frequently (dynamic)?

 E.g. Fraud is a moving target .
 The program needs to keep changing.

1/24/2024 4
 Why Machine Learning ?

 The program produced by the learning algorithm may look

very different from a typical hand-written program.
 If we do it right, the program works for new cases as well as the
ones we trained it on.

 If the data changes the program can change too by training on

the new data.
 Massive amount of computation are now cheaper than paying
someone to write a task-specific program.

1/24/2024 5
 Why Machine Learning ?

 Instead of writing a program by hand for each specific task, we

collect lots of examples that specify the correct output for a
given input.

 A machine learning algorithm then takes these examples and

produces a program that does the job.

1/24/2024 6
 Why Machine Learning ?

 Machine Learning is great for:

 Problems for which existing solutions require a lot of hand-tuning
or long lists of rules,
 Complex problems for which there is no good solution at all using
a traditional approach,
 Fluctuating environments: a ML system can adapt to new data,
 Getting insights about complex problems and large amounts of
data.

 Machine Learning Approaches: the program is much shorter,

easier to maintain, and most likely more accurate.
1/24/2024 7
 What is Machine Learning ?

 A more general definition of ML:

 Machine Learning: is a field of study that gives computers the
ability to learn without being explicitly programmed. (Arthur
Samuel – 1959)

 A more engineering-oriented:
 A computer program is said to learn from experience E with
respect to some class of task T and performance measure P, if its
performance at task in T, as measured by P, improves with
experience E. (Tom Mitchell – 1998)

1/24/2024 8
 What is Machine Learning ?

 What is ML: building computational artifacts that learn over

time based on experience.
 It includes:
 The mathematics behind it
 The science behind it (systematic and logical approach,
knowledge base on demonstrable)
 The engineering behind it (the application of mathematics and
science)
 The computing behind it (using computer technology)

1/24/2024 9
 Examples of ML Application

 Some examples of tasks best solved by learning algorithms.

 Recognizing patterns:
Object in real scenes
Facial identities or facial expressions
Spoken words
 Recognizing anomalies:
Unusual sequences of credit card transactions
Unusual patterns of sensor readings in a nuclear power plant
 Prediction:
Future stock price or currency exchange rates
Which movies will a person like?
1/24/2024 10
 Examples of ML Application

 Some application of ML in Google:

 Page ranking (RankBrain)
 Spam identification in email
 Speech recognition (language recognition)
 Software in Cars (Autonomous and self driving cars)
 Recommender system
 Google voice search
 Gmail inbox’s smart replay
 Google photos
 Google Cloud Vision API
 Virtual reality etc…
1/24/2024 11
 ML related to this Course

 The course can be:

 More of theoretical :
 More of practical:
 Balancing the theory and the application

1/24/2024 12
 ML related to this Course

 Some prerequisite :
 Programming (python, matlab, R etc…)
 Probability and Statistics;
 Mathematics (Linear Algebra, Matrix, Vector, differential
calculus, and integral calculus)
 Curiosity of working with data etc… (most importantly)

 ML – it is all about learning from examples (data).

 We take features (independent attributes) and produced labels
(dependent attribute).

1/24/2024 13
 Learning Path

 Generally, the learning path includes:

 Math skills
 Programming skills
 Data Engineering skills
 Machine learning algorithms
 Machine learning frameworks

1/24/2024 14
 Learning Path

 Generally, the learning path includes: Math skills

1/24/2024 15
 Learning Path

 Generally, the learning path includes: Programming skills

1/24/2024 16
 Learning Path

 Generally, the learning path includes: Data Engineering skills

1/24/2024 17
 Learning Path

 Generally, the learning path includes: ML algorithms

1/24/2024 18
 Learning Path

 Generally, the learning path includes: ML Framework

1/24/2024 19
 Machine Learning systems

 There are so many different types of Machine Learning systems

that it is useful to classify them in broad categories based on:
 Whether or not they are trained with human supervision
(supervised, unsupervised, semi-supervised, and Reinforcement
Learning)
 Whether or not they can learn incrementally on the fly (online
versus batch learning)
 Whether they work by simply comparing new data points to
known data points, or instead detect patterns in the training data
and build a predictive model, much like scientists do (instance-
based versus model-based learning)
1/24/2024 20
Major Classes of ML Algorithms

1/24/2024 21
 Supervised Learning

1/24/2024 22
 Supervised Learning

1/24/2024 23
 Supervised Learning

 Supervised learning: (function approximation) [labels data

well]

Input 1 2 3 4 5 6 9
Output 1 4 9 16 25 36 ?

output <- input2

 Learn to predict an output when given an input vector.

 E.g.: Features: age, gander, smoking, drinking etc…
 Labels : having the disease , does not have the disease
1/24/2024 24
 Supervised Learning

 Supervised Learning:
 Regression Problem: numerical / continuous value.
 Given some data, you assume that those values come from some
sort of function and try to find out what the function is.
 It is a problem of function approximation or interpolation.

 Classification Problem: nominal / discrete value.

 Grouping the data into predetermined classes.

1/24/2024 25
 Supervised Learning

 Supervised learning: making predictions using data.

 Example : is a given email “spam” or “ham”.
 There is an outcome; we are trying to predict.
 Classification

1/24/2024 26
 Unsupervised Learning

 Unsupervised learning :
 Unsupervised (concise description) [cluster scores well]
 The aim of unsupervised learning is to find clusters of
similar inputs in the data without being explicitly told that some
data points belong to one class and the other in other classes.

 The algorithm has to discover this similarity by itself.

 Discover a good internal representation of the input.

1/24/2024 27
 Unsupervised Learning

1/24/2024 28
 Unsupervised Learning

1/24/2024 29
 Unsupervised Learning

 Unsupervised learning: extracting structure from data.

 Example: segment grocery store shopper into clusters that exhibit
similar behaviors.
 There is no “right answer”
 Clustering

1/24/2024 30
 Reinforcement Learning

 Reinforcement learning :
 Reinforcement (stimulus/state, action, reward) [behavior scores
well] – reward maximization.

 The algorithm searches over the state space of possible inputs and
outputs in order to maximize a reward.
 Learn to select an action to maximize payoff.

1/24/2024 31
 Reinforcement Learning

1/24/2024 32
 Machine Learning Vs AI

 Machine learning : is more of induction (from specific to

generalization)
 Abebe is mortal, Almaz is mortal
 Kebede is mortal,
 Then, all humans are mortal

 Artificial Intelligence: more of deduction (from general to

specific – reasoning )
 A implies B
 Having “A”, you can infer “B”

1/24/2024 33
 Machine Learning Vs AI

 Machine learning can be:

 Unsupervised (description) -> summaries -> Supervised (function
approximation) -> Labels
 Data is the key in machine learning (ML):

 The difference b/n solving a problem as AI or ML person is:

 As a AI person: the algorithm is central,
 As a ML person: the data is central (get the priority).

1/24/2024 34
 Machine Learning Application

 Machine learning can be applied:

 What are the serious of steps I need to do in order to solve some
problem?
 If I tried to describe this problem in a particular way, is it
solvable?

1/24/2024 35
 Machine Learning

 What is Machine Learning?

 Machine learning is the semi-automated extraction of knowledge
from data?
Knowledge from data: Starts with a question that might be
answerable using data.
Automated extraction: A computer provides the insight.
Semi-automated: Requires many smart decisions by a human.

1/24/2024 36
 Machine Learning

 How does ML “work”?

 High-level steps of supervised learning:
 First, train a machine learning model using labeled data.
“Labeled data” has been labeled with the outcome.
“ML model” learns the relationship b/n the attributes of the data and
its outcome.

 Then, make predictions on new data for which the label is

unknown.
Example: segment grocery store shopper into clusters that exhibit
similar behaviors.

1/24/2024 37
 Machine Learning

 Questions about ML?

 How do I choose which attributes of my data to include in the
model?
 How do I choose which model to use?
 How do I optimize the model for best performance?
 How do I ensure that I’m building a model that will generalize to
unseen data?
 Can I estimate how well my model is likely to perform on unseen
data?

1/24/2024 38
 Question & Answer

1/24/2024 39
 Thank You !!!

1/24/2024 40