Scribd
Upload
16 views

Bayesian Belief Networks

Bayesian Belief Networks (BBNs) are probabilistic graphical models that represent dependencies among attributes, overcoming limitations of Naïve Bayesian classifiers which assume conditional independence. BBNs consist of a directed acyclic graph and conditional probability tables, allowing for the computation of joint probability distributions. Training involves understanding network topology and variables, utilizing methods like gradient descent to optimize conditional probability table entries.

Uploaded by

umakanthn
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
16 views

Bayesian Belief Networks

Bayesian Belief Networks (BBNs) are probabilistic graphical models that represent dependencies among attributes, overcoming limitations of Naïve Bayesian classifiers which assume conditional independence. BBNs consist of a directed acyclic graph and conditional probability tables, allowing for the computation of joint probability distributions. Training involves understanding network topology and variables, utilizing methods like gradient descent to optimize conditional probability table entries.

Uploaded by

umakanthn
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/ 9

Bayesian Belief

Networks

By,
Umakanth N
Session Outcomes

 Train & Use Bayesian belief networks

February 10, 2025 Bayesian Belief Networks 2


Bayesian Belief
Networks
 Probabilistic graphical models, which allows the
representation of dependencies among subset of
attributes.
 Limitation of Naïve Bayesian classifier:
 As it assumes that the attributes are conditionally
independent of each other. If this condition holds
true, Naïve Bayesian classifier is the best &
accurate algorithm among other classifiers.
 Practically, dependencies will exists between the
attributes.
 Bayesian belief network is also known as Belief
networks, Bayesian networks, or Probabilistic
networks
February 10, 2025 Bayesian Belief Networks 3
Bayesian Belief Networks
(Contd.,)
 It is defined by two components:
 Directed acyclic graph
 Each node is the random variable, which can be
discrete or continuous-valued
 Set of conditional probability tables

February 10, 2025 Bayesian Belief Networks 4


Bayesian Belief Networks
(Contd.,)

 Some rules probability values inferred from the


conditional probability table are,
 P(LungCancer = yes | FamilyHistory = yes, Smoker
= yes) = 0.8
 P(LungCancer = no | FamilyHistory = no, Smoker =
no) = 0.9
February 10, 2025 Bayesian Belief Networks 5
Bayesian Belief Networks
(Contd.,)

 Thus, Bayesian belief network can provide joint


probability distribution by,

February 10, 2025 Bayesian Belief Networks 6


Training Bayesian Belief
Networks
 Keys to train Bayesian Belief network,
 Network topology,
 Network variables (Observable / Hidden)
 If network topology is known and variable is
observable, then training the network can be
done by,
 Computing Conditional Probability Table (CPT)
entries by Naïve Bayesian classification.
 If topology is given and some variables are
hidden/missing,
 Gradient descent method (learning the values of
CPT entries)
 It is used to search for the CPT values Wijk that best
February 10, 2025 Bayesian Belief Networks 7
Gradient descent
method
 Performs Greedy-Hill climbing approach.
 For each iteration, the algorithm moves towards the best
solution, without backtracking
 Steps:
 Weights Wijk are initialized with random probability values
 Compute the Gradients

 Take small step in the direction of gradients

 Renormalize the weights


 This learning form is called Adaptive Probabilistic
networks
February 10, 2025 Bayesian Belief Networks 8
? ?
. . ?
ie s
e r
Q u

You might also like