Probabilistic graphical models (PGMs) are a rich framework for encoding probability distributions over complex domains: joint (multivariate) distributions over large numbers of random variables that interact with each other. These representations sit at the intersection of statistics and computer science, relying on concepts from probability theory, graph algorithms, machine learning, and more. They are the basis for the state-of-the-art methods in a wide variety of applications, such as medical diagnosis, image understanding, speech recognition, natural language processing, and many, many more. They are also a foundational tool in formulating many machine learning problems.
This course is the second in a sequence of three. Following the first course, which focused on representation, this course addresses the question of probabilistic inference: how a PGM can be used to answer questions. Even though a PGM generally describes a very high dimensional distribution, its structure is designed so as to allow questions to be answered efficiently. The course presents both exact and approximate algorithms for different types of inference tasks, and discusses where each could best be applied. The (highly recommended) honors track contains two hands-on programming assignments, in which key routines of the most commonly used exact and approximate algorithms are implemented and applied to a real-world problem.

Voir plus

- Week 1 -
**Inference Overview**

This module provides a high-level overview of the main types of inference tasks typically encountered in graphical models: conditional probability queries, and finding the most likely assignment (MAP inference). - Week 1 -
**Variable Elimination**

This module presents the simplest algorithm for exact inference in graphical models: variable elimination. We describe the algorithm, and analyze its complexity in terms of properties of the graph structure. - Week 2 -
**Belief Propagation Algorithms**

This module describes an alternative view of exact inference in graphical models: that of message passing between clusters each of which encodes a factor over a subset of variables. This framework provides a basis for a variety of exact and approximate inferen... - Week 3 -
**MAP Algorithms**

This module describes algorithms for finding the most likely assignment for a distribution encoded as a PGM (a task known as MAP inference). We describe message passing algorithms, which are very similar to the algorithms for computing conditional probabilitie... - Week 4 -
**Sampling Methods**

In this module, we discuss a class of algorithms that uses random sampling to provide approximate answers to conditional probability queries. Most commonly used among these is the class of Markov Chain Monte Carlo (MCMC) algorithms, which includes the simple G... - Week 4 -
**Inference in Temporal Models**

In this brief lesson, we discuss some of the complexities of applying some of the exact or approximate inference algorithms that we learned earlier in this course to dynamic Bayesian networks. - Week 5 -
**Inference Summary**

This module summarizes some of the topics that we covered in this course and discusses tradeoffs between different algorithms. It also includes the course final exam.

Coursera est une entreprise numérique proposant des formation en ligne ouverte à tous fondée par les professeurs d'informatique Andrew Ng et Daphne Koller de l'université Stanford, située à Mountain View, Californie.

Ce qui la différencie le plus des autres plateformes MOOC, c'est qu'elle travaille qu'avec les meilleures universités et organisations mondiales et diffuse leurs contenus sur le web.

Lire les avis

Trier par :

5 / 5

Terminé

Perhaps the best introduction to AI/ML - especially for those who think "the future ain't what it used to be"; the mathematical techniques covered by the course form a toolkit which can be easily thought of as "core", i.e. a locus of strength which enables a wide universe of thinking about complex problems (many of which were correctly not thought to be tractable in practice until very recently!)...

0

4 / 5

Terminé

Unfortunately, in my opinion, this course is not as well structured as the first course (PGM1: structure). There are some bugs/issues with the PAs code that should have been fixed and the course material could focus a bit more on the case of continuous random variables (which are almost ignored throughout the course). It is still a great and totally worth it course, though. Highly recommended for machine learning post-graduate students.

0

4 / 5

Terminé

Pretty good course, albeit very dense compared to the first one (which was certainly not trivial). I would give it 5 stars just based on the content, but the programming assignments don't work without significant extra effort. I completed the honors track for the first course, but gave up after spending 4 hours trying to fix HW bugs that were reported 8 months ago. Would have also been nice to have more practical examples to work on. Some of the material is very theoretical, and I find it hard to build intuitions without applying the algorithms in practice.

0

3 / 5

Terminé

Thumbs up for the course content. However, there are technical problems which no one is attending to. I could not submit my programming assignment, and after consulting every available resource, I was not dignified with an answer. It is a shame how such wonderful learning opportunity can become spoiled by some insignificant technical detail.By my opinion, the course should not be divided into 3 courses. Many technicalities were done sloppy in the process.

0

4 / 5

Terminé

it is a great class. but the presentation of the materials could be better: maybe each unit should start with a review of the key concepts we learned before? maybe a slide on motivation of the work before we dive deep into the math? but again, this is a great class! recommended 100%

0

4 / 5

Terminé

Great Course, not five stars just because probabbly it was too much content for the period of time we had the Course. I've got no complaints about the amount of content, but some of concepts were missing and the Programming Assignments were not so well described, sometimes I couldn't understand what to do.

0

2 / 5

Terminé

The lectures are fine and the book is great, but the assignments have a lot of technical problems. I spent most of my effort trying to solve trivial issues with the sample code and dealing with the auto grader.

0

5 / 5

Terminé

Thanks a lot for professor D.K.'s great course for PGM inference part. Really a very good starting point for PGM model and preparation for learning part.

0

3 / 5

Terminé

有幸能听到COＵＲＳＥＲＡ创始人的课，确实领略了一下大牛人的风采。但是从教课这个层面来看， 我相信有人能教得更好。 最可惜的是编程作业，我根本不能ｓｕｂｍｉｔ 。上课的内容和作业脱节很明显。 而且很多时候， 基本没有编程方面的支持（可以从论坛的人气就可以看出了）， 学生几乎无从下手总的来说，此课过多的侧重于抽象层面的东西。

0

4 / 5

Terminé

The course lectures are even better than PGM I, as it appears that Professor Koller has recorded some material recently that helps fill in small holes from the previously recorded lectures. Hopefully she'll have time to clean up PGM I in the near future for future students.This course is another tour-de-force for debugging, though it definitely made me a better programmer (I'm intermediate). I wish that the Discussion Boards were more active, and it's a shame that the Mentors were Missing In Action. On the one hand, the programming instructions were sometimes a bit vague, which made the assignments less like assignments are more like research projects. For these 2 reasons, the course is 4-star rather than 5-star. Still, it's a lot better than trying to learn this out of the book by oneself. Some say enrollment has dropped off since they began charging for getting access to Quizzes and Programming Assignments. Or it may be attrition, as these are pretty challenging (and well taught) courses. I'm very happy to support this course financially, as it's loads cheaper than what I'd be paying if I were back at Stanford.Like PGM I, I strongly recommend doing the Honors Programming Assignments, as it's really the way to learn the material well.

0

vous pourriez aussi être intéressé par...

MESSAGE_PLACEHOLDER