Probabilistic Graphical Models. Master a new way of inferring and learning in complex domains.
Suggested by: Coursera (What is Coursera?)
No prior knowledge required
No unnecessary risks
Probabilistic graphical models (PGMs) are a rich framework for encoding probability distributions across complex domains: combined (multivariate) distributions over a large number of interacting random variables. These representations lie at the intersection of statistics and computer science, and draw on concepts from probability theory, graph algorithms, machine learning, and more. They form the basis of cutting-edge methods in a wide range of applications, such as medical diagnosis, image understanding, speech recognition, natural language processing, and more. They are also a fundamental tool in formulating many machine learning problems.
This course is the first in a series of three. It describes the two basic representations of PGM: Bayesian networks, which depend on a sorted graph; and Markov networks, which use an unsorted graph. The course deals with the theoretical properties of these representations and how they are implemented in practice. The extended track (highly recommended) includes several practical assignments on how to represent real-world problems. The course will also introduce some important extensions beyond the basic PGM representation, which allow complex models to be coded in a compact way.
Probabilistic graphical models (PGMs) are a rich framework for encoding probability distributions across complex domains: combined (multivariate) distributions over a large number of interacting random variables. These representations are at the intersection of statistics and computer science, and draw on concepts from probability theory, graph algorithms, machine learning, and more. They form the basis of cutting-edge methods in a wide range of applications, such as medical diagnosis, image understanding, speech recognition, natural language processing, and more.
This course is the second in a series of three. Following the first course, which focused on representation, this course addresses the question of probabilistic inference: how PGMs can be used to answer questions. Although PGM models typically describe very high-dimensional distributions, their structure is designed so that questions can be answered efficiently. The course introduces exact and approximate algorithms for various types of inference tasks, and discusses where each can best be applied. The extended track (highly recommended) includes two practical programming assignments, in which command developers of the most common exact and approximate algorithms are implemented and applied to real problems.
Probabilistic graphical models (PGMs) are a rich framework for encoding probability distributions across complex domains: combined (multivariate) distributions over a large number of interacting random variables. These representations lie at the intersection of statistics and computer science, and draw on concepts from probability theory, graph algorithms, machine learning, and more. They form the basis of cutting-edge methods in a wide range of applications, such as medical diagnosis, image understanding, speech recognition, natural language processing, and more.
This course is the third in a series of three. Following the first course, which focused on representation, and the second, which focused on inference, this course addresses the question of learning: how to learn a PGM from a set of examples. The course discusses the central problems of parameter estimation in financial and other models, as well as the task of learning the structure for financial models. The extended track (highly recommended) includes two practical programming assignments, in which two key tasks of the two algorithms commonly used in learning are implemented and applied to real problems.
