Introduction to Machine

Learning
Session 2
 Problems
Black box model
• Black Box Model

• Black box model consists of 3 components

• When one component is unknown: new problem type

3 / 20
 Black box model:
Optimisation
Optimization
• Model and desired output is known, task is to find inputs

Black box model:

Optimisation example 3: 8 queens problem
• Examples:
Time tables for university, call center, or hospital
•
• Given an 8-by-8
• Examples: • Design
chessboard
specifications
and 8 queens
• Traveling salesman problem (TSP)
• Place the 8 queens on the
• • Eight-queens
Time tables for university problem,
chessboard without any
etc.

conflict 4 / 20

• Design speci cations

• Two queens conflict if they
share same row, column or
• Travelingdiagonal
salesman problem (TSP)
• Can be extended to an n
• Eight-queens
queensproblem,
problemetc.
(n>8)
fi
 Black box model:
Modelling
Modeling
• We have corresponding sets of inputs & outputs and
seek model that delivers correct output for every known
input

• Note: modelling problems can be transformed into

optimisation problems
• Modeling Examples:
• Evolutionary machine learning
• Predicting stock exchange
• Machine Learning problems
• Voice control system for smart homes

9 / 20

• Predicting stock exchange

• Voice control for smart homes

• Bank credit prediction

 Black box model:
Simulation
Simulation
• We have a given model and wish to know the outputs
that arise under different input conditions

• Often used to answer what-if questions in evolving

• Simulation Examples:
dynamic environments
• Economics,
Evolutionary economics, Artificial Life
Arti cial Life
Weather forecast system
• Weather forecast system
Impact analysis new tax systems
• Impact analysis new tax systems
11 / 20

• Tra c Simulation
ffi
fi
Why Machine Learning?
• For many problems, it’s di cult to program the correct behavior by hand

• recognizing people and objects

• understanding human speech

• Machine learning approach: program an algorithm to automatically learn

from data, or from experience

• Why might you want to use a learning algorithm?

• hard to code up a solution by hand (e.g. vision, speech)

• system needs to adapt to a changing environment (e.g. spam detection)

• want the system to perform better than the human programmers

ffi
 What is Machine Learning?
• It’s similar to statistics…

• Both elds try to uncover patterns in data

• Both elds draw heavily on calculus, probability, and linear algebra,

and share many of the same core algorithms

• But it’s not statistics!

• Stats is more concerned with helping scientists and policymakers

draw good conclusions; ML is more concerned with building
autonomous agents

• Stats puts more emphasis on interpretability and mathematical rigor;

ML puts more emphasis on predictive performance, scalability, and
autonomy
fi
fi
 Relations to AI

• Nowadays, “machine learning” is often brought up with “arti cial intelligence” (AI)

• AI does not always imply a learning based system

• Symbolic reasoning

• Rule based system

• Tree search

• etc.

• Learning based system → learned based on the data → more exibility, good at
solving pattern recognition problems.
fi
fl
 Relations to human learning
• Human learning is:

• Very data e cient

• An entire multitasking system (vision, language, motor control, etc.)

• Takes at least a few years

• For serving speci c purposes, machine learning doesn’t have to look

like human learning in the end.

• It may borrow ideas from biological systems, e.g., neural networks.

• It may perform better or worse than humans.

ffi
fi
 Machine Learning
• Building algorithms which, to be useful, rely on a collection of
examples of some phenomenon.

• These examples can come from nature, be handcrafted by

humans or generated by another algorithm.

• Machine learning can also be de ned as the process of

solving a practical problem by

• 1) gathering a dataset, and

• 2) algorithmically building a statistical model based on that

dataset.
fi
 Types of Learning

• Supervised Learning

• Unsupervised Learning

• Reinforcement Learning
 Supervised Learning
• The dataset is the collection of labeled examples; {(xi, yi)}i=1,…,N

• Each element xi among N is called a feature vector.

• A feature vector is a vector in which each dimension contains a value that

describes the example somehow.

• Example of features for a person:

• Height

• Weight

• Gender

• …
 Supervised Learning
• The label yi can be either an element belonging to

• a nite set of classes {1, 2, . . . , C},

• a real number,

• a more complex structure, like a vector, a matrix, a tree,

or a graph.

• Example for spam detection:

• Classes = {spam, not_spam}

fi
Unsupervised Learning
• In unsupervised learning, the dataset is a collection of unlabeled examples, {xi}i=1,…,N

• The goal is to create a model that

• takes a feature vector x as input and either

• transforms it into another vector or

• into a value,

• that can be used to solve a practical problem.

Unsupervised Learning
• Examples:

• Clustering

• the model returns the id of the cluster for each feature vector in the
dataset

• Dimensionality Reduction

• the output of the model is a feature vector that has fewer features than
the input x

• Outlier Detection

• the output is a real number that indicates how x is di erent from a

“typical” example in the dataset
ff
 Reinforcement Learning
• Reinforcement learning is a sub eld of machine learning where

• the machine “lives” in an environment

• and is capable of perceiving the state of that environment as a vector of features.

• The machine can execute actions in every state.

• Di erent actions bring di erent rewards and could also move the machine to another
state of the environment.

• The goal of a reinforcement learning algorithm is to learn a policy.

ff
ff
fi
Reinforcement Learning
• Reinforcement learning solves a particular kind of problem
where decision making is sequential, and the goal is long-
term, such as

• game playing,

• robotics,

• resource management,

• or logistics
History of machine learning
• 1957 — Perceptron algorithm (implemented as a circuit!)

• 1959 — Arthur Samuel wrote a learning-based checkers program that could defeat
him

• 1969 — Minsky and Papert’s book Perceptrons (limitations of linear models)

• 1980s — Some foundational ideas

• Connectionist psychologists explored neural models of cognition

• 1984 — Leslie Valiant formalized the problem of learning as PAC learning

• 1988 — Backpropagation (re-)discovered by Geo rey Hinton and colleagues

• 1988 — Judea Pearl’s book Probabilistic Reasoning in Intelligent Systems

introduced Bayesian networks
ff
History of machine learning
• 1990s — the “AI Winter”, a time of pessimism and low funding But
looking back, the ’90s were also sort of a golden age for ML research

• Markov chain Monte Carlo

• variational inference

• kernels and support vector machines

• boosting

• convolutional networks

• reinforcement learning
 History of machine learning
• 2000s — applied AI elds (vision, NLP, etc.) adopted ML

• 2010s — deep learning

• 2010–2012 — neural nets smashed previous records in speech-to-text and

object recognition

• increasing adoption by the tech industry

• 2016 — AlphaGo defeated the human Go champion

• 2018-now — generating photorealistic images and videos

• 2020 — GPT3 language model

• now — increasing attention to ethical and societal implications

fi
ML Applications
Example for Supervised Learning

• Spam Detection

• Goal: Whether the incoming email is spam or not

• You gather the data, for example, 10,000 email

messages, each with a label either “spam” or “not_spam”

• You have to convert each email message into a feature

vector

• One common way to convert a text into a feature

vector, called bag of words
 Bag of Words
• the rst feature is equal to 1 if the email message contains
the word “a”; otherwise, this feature is 0;

• the second feature is equal to 1 if the email message

contains the word “aaron”; otherwise, this feature equals 0;

• …

• the feature at position 20,000 is equal to 1 if the email

message contains the word
“zulu”; otherwise, this feature is equal to 0.
fi
 Supervised Algorithm
• We are going to nd a 9999-dimensional line (hyperplane)
that separates examples with positive labels from
examples with negative labels

x(2)

1
2

=
b
w

—
x
w

0
=
b
—
x
w

1
—
=
b
—
x
w
x(1)
b
w
fi