Kalender

Exportera Om schema på KTH

Logga in till din kurswebb

Du är inte inloggad på KTH så innehållet är inte anpassat efter dina val.

Ändra tidsperiod eller vy

Vecka 36 2013			Visa i Mitt schema

	Mån 2 sep 17:00-19:00	Lecture 1, Introduction
	HT 2013 Föreläsning Plats: D1	Readings: Marsland, Chapter 1 Structure of the Course What will the course cover? How are labs and examination handled? Learning Machines What do we mean by a "Learning Machine"? What can learning algorithms be used for? How can a simple learning program be constructured? Slides from this lecture: Slides lecture 1 Slides lecture 1 (part II)
	Tors 5 sep 15:00-17:00	Lecture 2, Concept Learning [Örjan]
	HT 2013 Föreläsning Plats: M1	Concept Learning Readings: Marsland, Chapter 1 What is Concept Learning? Important terminology: positive and negative examples, hypotheses The structure of the hypotheses space: General and Special hypotheses How can one find a hypothesis that conforms with data? The Find-S algorithm Why does a naive List-Then-Eliminate algorithm not work in practice? How can the choise of leaning method influence the result? "Bias-Free Learning", is it possible? Is it desirable? Slides on Concept Learning
Vecka 37 2013			Visa i Mitt schema

	Mån 9 sep 17:00-19:00	Lecture 3, Decision Trees [Atsuto]
	HT 2013 Föreläsning Plats: D1	Decision Trees Readings: Marsland, Chapter 6 What is a Decision Tree? When are decision trees useful? How can one select what questions to ask? What do we mean by Entropy for a data set? What do we mean by the Information Gain of a question? What Bias do we get if we maximize the Information Gain? What does William from Ockham (1285-1349) have to do with this? Is it possible to learn too much? How can one prevent the algorithms from learning irrelevant details? Slides on Decision Trees
Vecka 38 2013			Visa i Mitt schema

	Mån 16 sep 17:00-19:00	Lecture 4, Artificial Neural Networks [Örjan]
	HT 2013 Föreläsning Plats: D1	Artificial Neural Networks Readings: Marsland, Chapter 2-3 Why are these algorithms called Neural Networks? How does a Single Layer Perceptron work? What can a Single Layer Perceptron learn? The Perceptron Learning algorithm The Delta-rule, error minimization The Multi Layer Perceptron Error Backpropagation Slides on ANN
Vecka 39 2013			Visa i Mitt schema

	Mån 23 sep 15:00-17:00	Lecture 5, Support Vector Machines [Örjan]
	HT 2013 Föreläsning Plats: E1	Support Vector Machines Readings: Marsland, Chapter 5 How does linear separation act in high-dimensional spaces? What does empirical and structural risk refer to? Why are classification margins good for generalization performance? When are slack variables useful? How can a support vector machine be trained? Why is the dual optimization problem often easier to solve? What is a support vector? What is the advantage of using kernel functions? Slides on SVM
	Fre 27 sep 15:00-17:00	Lecture 6, Ensemble Learning [Atsuto]
	HT 2013 Föreläsning Plats: E1	Bagging and Boosting Readings: Marsland, Chapter 7 Wisdom of Crowd Characterization of classifiers Bagging Boosting Slides on Ensemble Learning
Vecka 40 2013			Visa i Mitt schema

	Mån 30 sep 16:00-18:00	Lecture 7, Probability based Learning [Atsuto]
	HT 2013 Föreläsning Plats: F2	Bayesian Learning Readings: Marsland, Chapter 8 and 15.1 Bayes Theorem MAP, ML hypotheses MAP learners Naive Bayes learner Expectation Maximization (EM) algorithm Slides on Probability based Learning
	Tors 3 okt 16:00-18:00	Lecture 8, Evolutionary Algorithms [Atsuto]
	HT 2013 Föreläsning Plats: E1	Genetic Algorithms Readings: Marsland, chapter 12 In what way can evolution be regarded as an algorithm? What can be optimized using Genetic Algorithms? How are potential solutions represented? How do we represent the goal? Chromosomes, Populations and Generations What makes GA different from other optimization methods? What do we mean by Genetic Programming? What can go wrong? Slides on Genetic Algorithm
Vecka 41 2013			Visa i Mitt schema

	Mån 7 okt 17:00-19:00	Lecture 9, Reinforcement Learning [Örjan]
	HT 2013 Föreläsning Plats: D1	Reinforcement Learning Readings: Marsland: chapter 13 Is it possible to learn when nobody tells you the correct answer? Central terms: State, Action, Reward More terms: Value function (cumulative value), Policy How can we judge the consequences of our actions? What do we mean by an optimal behavior? Is it possible to learn what is the best thing to do in each state? Is it possible to learn faster by planning ahead? Slides on Reinforcement Learning
	Tors 10 okt 10:00-12:00	Lecture 10, Graphical Models [Atsuto]
	HT 2013 Föreläsning Plats: FR4	Graphical Models Readings: Marsland, chapter 15 How can conditional probabilities be represented as a graph? What is a Bayesian network? What is a Hidden Markov Model? What is a Dynamic Bayesian network? Slides on Bayesian Networks Slides on Dynamic Bayesian Networks
Vecka 42 2013			Visa i Mitt schema

	Mån 14 okt 17:00-19:00	Lecture 11, Learning Theory [Örjan]
	HT 2013 Föreläsning Plats: D1	Learning Theory Is it possible to measure how hard a learning task is? What can go wrong during learning? What do we mean when we say that a hypothesis is "approximately correct" Is it possible to estimate the number of traning examples needed? Are there learning tasks that take exponentially long time? PAC-learnable VC-dimension Can we estimate how many errors a learner must make? Slides on Learning Theory
	Tors 17 okt 10:00-12:00	Lecture 12, Summary and Outlook
	HT 2013 Föreläsning Plats: D1	Slides, summary of Örjans lectures Slides, summary of Atsuto's lectures (extractions)

Kungliga Tekniska högskolan

Kalender

Logga in till din kurswebb

Tidsperiod

Vy

Structure of the Course

Learning Machines

Concept Learning

Decision Trees

Artificial Neural Networks

Support Vector Machines

Bagging and Boosting

Bayesian Learning

Genetic Algorithms

Reinforcement Learning

Graphical Models

Learning Theory