Probabilistic graphical Models: I, Introduction and overview (1, overview and motivation)

I. What PGM uses to do

1, medical diagnosis: from a variety of conditions to analyze the patient has what disease, what means to treat

2. Image segmentation: Analyze what each pixel corresponds to from a picture of megapixel

Two things in common: (1) There are very many different input variables, and (2) for the algorithm, the results are indeterminate.

Ii. What the PGM represents 1, Models

2, probabilistic (1) Probability: The design model is to analyze some of the uncertain things (uncertainty) (2) Source of uncertainty:

(3) Advantages of probability in model expression

3, Graphical (1) The form of the diagram is more suitable for the expression of complex systems, such as said above with a large number of variables

The goal is to find the joint probability distribution of things from these random variables.

(2) Example

(3) Advantages

Iii. Examples of applications

Original->superpixels (cut into large blocks of pixels), machine learning cut diagram->PGM identify image content

Iv. Overview

Probabilistic graphical Models: I, Introduction and overview (1, overview and motivation)