course aims in Estonian
Aine eesmärk on:
- tutvustada enamlevinud kaasaegseid masinõppe meetodeid;
- luua arusaam masinõppe meetodite baasiks olevast matemaatilisest aparatuurist;
- seostada varem õpitud matemaatilisi meetodeid masinõppe rakendustega;
- õppida programmeerima lihtsamaid masinõppe meetodeid;
- õppida tundma ja kasutama olemasolevaid masinõppe algoritme realiseerivaid pakette ja teeke.
course aims in English
The aim of this course is to:
- introduce the main methods in modern machine learning;
- establish the understanding of the mathematical basis of the machine learning methods;
- establish the connection between the previously studied mathematical courses and machine learning;
- learn to program simple machine learning models;
- get to know and use the existing tools and libraries implementing machine learning models and algorithms.
learning outcomes in the course in Est.
Õppeaine läbinud üliõpilane:
- nimetab põhilisi kaasaegseid masinõppe meetodeid;
- klassifitseerib ülesandeid ja valib nende lahendamiseks sobivaid masinõppe algoritme;
- programmeerib lihtsamaid masiõppe algoritme;
- kasutab olemasolevaid pakette ja teeke uute ülesannete lahendamiseks;
- hindab eri masinõppe meetodite sobivust uute ülesannete lahendamiseks.
learning outcomes in the course in Eng.
Upon completing this course the student:
- lists the main machine learning methods;
- classifies problems and chooses proper methods for each type of problems;
- programs simple machine learning algorithms;
- uses existing tools and libraries for solving new problems;
- estimates the suitability of different methods for solving new problems.
brief description of the course in Estonian
Põhimõisted: õppimine, üleõppimine, alaõppimine, juhendatud ja juhendamata õppimine, meetodi üldistusvõime.
Erinevad mudelid: regressioon, logistiline regressioon ja klassifitseerimine, närvivõrgud, Bayesi meetodid, klasterdamine, tugivektormasinad ja kernelid, tagasisidega õppimine, komponentanalüüsi meetodid, markovi mudelid
Meetodid: maksimaalse tõepära meetod, ootuste maksimeerimise meetod
Modelleerimise paradigmad: klassikaline, bayesiaanlik, parameetriline, mitteparameetriline
Optimeerimisalgoritmid: esimest ja teist järku meetodid.
Eeldused: Mat.analüüs I, Lineaaralgbra, Tõenäpsusteooria ja statistika. Samuti eeldatakse, et üliõpilane oskab programeerida.
brief description of the course in English
Main concepts: learning, overfitting, underfitting, supervised learning, unsupervised learning, generalization ability.
Models: regression, logistic regression and classification, neural networks, Bayes methods, clustering, support vector machines and kernels, reinforcement learning, component analysis methods, markov models
Methods: maximum likelihood estimation, expectation-maximization method
Modelling paradigms: frequentist, Bayesian, parametric, non-parametric.
Optimization: first and second order methods.
The student should possess basic knowledge of calculus, linear algebra, probability theory and statistics and computer programming before joining this course.
type of assessment in Estonian
Eksam 30%, kodutööd 70%.
type of assessment in English
Exam 30%, homework 70%.
independent study in Estonian
Programmeerimisülesanded, kodulugemine.
independent study in English
Programming exercises, home reading.
study literature
Christopher M. Bishop, “Pattern Recognition and Machine Learning”, Springer, 2007.
Kevin Murphy, “Machine Learning: A Probabilistic Perspective”, MIT Press, 2012.
Kursuse veebileht: https://courses.cs.ttu.ee/pages/ITI8565
The student should possess basic knowledge of calculus, linear algebra and computer programming before studying this course.
study forms and load
daytime study: weekly hours
4.0
session-based study work load (in a semester):