IB031: Úvod do strojového učení
Tomáš Brázdil, Luboš Popelínský, Karel Vaculík
i
Kdo s kým, o cem, proč
2
Kdo s kým, o cem, proč
3
Kdo s kým, o cem, proč
ISMU
4
Kdo s kým, o cem, proč
Příklad 1: lineární model - predikce
5
Kdo s kým, o cem, proč
Příklad 2: nelineární model - detekce spamů
6
Organizace
► přednášky
► projekt
► dvoučlenné týmy;
► studium dvou nepřednášených metod,
► experimentální porovnání s klasickými metodami
► v jazyce R
► semestrální zkouška
► písemná + ústní zkouška
Závěrečné hodnocení
► projekt — intro = poster o metodách, finál = výs experimentů
► semestrální zkouška
► písemná+ústní zkouška
Co je strojové učení
Herbert Simon (1960s): "Learning is any process by which a system improves performance from experience."
Tom Mitchell (1990s, paraphrased): "Learning aims to improve task, T, with respect to performance metric, P, based on experience, E."
8
Co je strojové učení
Herbert Simon (1960s): "Learning is any process by which a system improves performance from experience."
Tom Mitchell (1990s, paraphrased): "Learning aims to improve task, T, with respect to performance metric, P, based on experience, E. "
9
Príklady
T: Playing checkers
P: Percentage of games won against an arbitrary opponent E: Playing practice games against itself
T: Recognizing hand-written words
P: Percentage of words correctly classified
E: Database of human-labeled images of handwritten words
T: Driving on four-lane highways using vision sensors
P: Average distance traveled before a human-judged error
E: A sequence of images and steering commands recorded while
observing a human driver.
T: Categorize email messages as spam or legitimate. P: Percentage of email messages correctly classified. E: Database of emails, some with human-given labels
10
Príklady
T: Playing checkers
P: Percentage of games won against an arbitrary opponent E: Playing practice games against itself
T: Recognizing hand-written words
P: Percentage of words correctly classified
E: Database of human-labeled images of handwritten words
T: Driving on four-lane highways using vision sensors
P: Average distance traveled before a human-judged error
E: A sequence of images and steering commands recorded while
observing a human driver.
T: Categorize email messages as spam or legitimate. P: Percentage of email messages correctly classified. E: Database of emails, some with human-given labels
11
Príklady
T: Playing checkers
P: Percentage of games won against an arbitrary opponent E: Playing practice games against itself
T: Recognizing hand-written words
P: Percentage of words correctly classified
E: Database of human-labeled images of handwritten words
T: Driving on four-lane highways using vision sensors
P: Average distance traveled before a human-judged error
E: A sequence of images and steering commands recorded while
observing a human driver.
T: Categorize email messages as spam or legitimate. P: Percentage of email messages correctly classified. E: Database of emails, some with human-given labels
12
Príklady
T: Playing checkers
P: Percentage of games won against an arbitrary opponent E: Playing practice games against itself
T: Recognizing hand-written words
P: Percentage of words correctly classified
E: Database of human-labeled images of handwritten words
T: Driving on four-lane highways using vision sensors
P: Average distance traveled before a human-judged error
E: A sequence of images and steering commands recorded while
observing a human driver.
T: Categorize email messages as spam or legitimate. P: Percentage of email messages correctly classified. E: Database of emails, some with human-given labels
13
Dalsí príklady?
14
Třídy úloh
► shlukování
► klasifikace a predikce
► hledání asociací
► detekce anomálií
15
Historie
► 1950s :
Alan Turing and NP-hard problems
Samuel's checker player, see Ray Mooney ML Course slides
► 1960s :
Neural networks: Perceptron Pattern recognition Learning in the limit theory
Minsky and Papert prove limitations of Perceptron
► 1970s :
Symbolic concept induction Winston's arch learner
Expert systems and the knowledge acquisition bottleneck; Scientific discovery with BACON and AM (math) Quinlan's ID3 Michalski's AQ
16
Historie
► 1980s :
Advanced decision tree and rule learning
Learning and planning and problem solving
Resurgence of neural networks (connectionism,
backpropagation)
Valiant's PAC Learning Theory
Focus on experimental methodology
► 1990s : Data mining Text learning
Reinforcement learning (RL)
Inductive Logic Programming (ILP)
Ensembles: Bagging, Boosting, and Stacking
Bayes Net learning
Web mining
Weka
17
Historie
► 2000s :
Support vector machines. Kernel methods
Statistical relational learning
Graph and Sequence mining, Link learning
Privacy-preserving data mining
Security (intrusion, virus, and worm detection)
Recommender systems; Personalized assistants that learn
Visual data mining
Stream mining
Rapid Miner
R for machine learning
► 2010s : KNIME
Big data, Big data, Big data . . Outlier detection and explanation
18