Scribd
Upload
9 views

3 3-BayesianNetworks

Uploaded by

Sujithra Jones
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
9 views

3 3-BayesianNetworks

Uploaded by

Sujithra Jones
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

NPTEL

Video Course on Machine Learning

Professor Carl Gustaf Jansson, KTH

Week 3 Forms of Representation

Video 3.3 Bayes (ian) Belief Networks


General characteristics
of this Representation

Abbreviation : BBN (Bayesian Belief Network)

Synonyms: Bayes (ian) network, Bayes(ian) model, Belief


network, Decision network, or probabilistic directed acyclic
graphical model.

A BBN is a probabilistic graphical model that represents a set of


variables and their conditional dependencies via a Directed Acyclic
Graph (DAG).

BBNs enable us to model and reason about uncertainty. BBNs


accommodate both subjective probabilities and probabilities based on
objective data.

The most important use of BBNs is in revising probabilities in the


light of actual observations of events.
Interdisciplinary inspiration for this
Representation

Bayesian probability is an interpretation of the concept of


probability, in which, instead of frequency of some phenomenon,
probability is interpreted as a reasonable expectation representing
- a state of knowledge or
- quantification of a personal belief.

In the Bayesian view, a probability is assigned to a hypothesis, in


contrast to the frequentist case where a hypothesis is tested without
being assigned a probability.

The Bayesian interpretation of probability can be seen as an


extension of propositional logic that enables reasoning with
hypotheses, i.e., the propositions whose truth or falsity is uncertain.
Bayes theorem

P(A) prior probability for A A


P(B) prior probability for B

P(B|A) conditional probability for B given A


P(A|B) conditional probability for A given B

P(A, B) joint probability for A and B B

P(A,B) = P(B|A) * P(A)

P(A|B) = P (B|A) * P(A) / P(B) Theorem

Intuitive meaning: Infering a conditional probability


for a cause given a symptom.
Core components of this P(SPRINKLER | RAIN) P ( RAIN)

Representation
Nodes represent variables in the Bayesian sense:
observable quantities, hidden variables or hypotheses.
Edges represent conditional dependencies.

Each node is associated with a probability function that


takes, as input, a particular set of probabilities for values
for the node's parent variables, and outputs the probability
of the values of the variable represented by the node.

Prior Probabilities: e.g. P(RAIN) P(GRASS WET | RAIN, SPRINKLER)


Conditional Probabilities: e.g. P(SPRINKLER | RAIN)
Joint Probability Function: P(GRASS WET, SPRINKLER, RAIN) =
P(GRASS WET | RAIN, SPRINKLER) * P(SPRINKLER | RAIN) * P ( RAIN)

Typically the probability functions are described in table


form.
A related example
Conditional Independence

Nodes that are not connected by any path represent variables that are conditionally
independent of each other.

e.g. Sprinkler and Rain in the related example are conditionally independent
P(Sprinkler|Cloudy, Rain) = P(Sprinkler|Cloudy)

This joint probabilities for the network can be calculated as.

P(Wetgrass, Sprinkler, Cloudy, Rain)=

P(Wetgrass | Sprinkler, Rain) * P(Sprinkler | Cloudy) * P(Rain | Cloudy) * P (Cloudy)


Structural aspects of this Representation

In contrast to toy examples, realistic networks are large


Structural aspects of this Representation

A valid DAG NOT a DAG because of the cycle ABEC


Three kinds of substructures in Bayesian networks

A
Sequence
B

C
A B
Convergence

A
Divergence

B C
Problemsolving for this Representation
A Bayesian network is a complete probabilistic model of the variables and their H H
relationships describing effects in terms of causes.

Inferences typically aims to update our beliefs concerning causes in the light of new
evidence. P(E|H) P(H|E)

Backward inferences in a Bayesian network can be viewed as the answering of queries


about the state of a subset of variables (the hypothesis variables) when other variables E E
(the evidence variables) are observed.

Bayesian inference derives the posterior probability of the subset of hypothesis variables
through systematic use of Bayes theorem : P(H|E) = P (E|H) * P(E) / P(H).

Inferences of this kind can be recursively applied throughout the whole structure.
Learning for this Representation

PARAMETER learning
• Learning parameters (conditional probabilities)
given a fixed variable structure

STRUCTURE learning
• Learning the variable structure
• Learning hidden variables (non observables)
NPTEL

Video Course on Machine Learning

Professor Carl Gustaf Jansson, KTH

Thanks for your attention!

The next lecture 3.4 will be on the topic:

Neural Networks

You might also like