Scribd
Upload
43 views24 pages

Learning Bayesian Network Structure Based On Ant Colony Optimization and Differential Evolution

This document proposes a new algorithm called cooperative ant colony optimization (coACO) for learning the structure of Bayesian networks. The coACO algorithm divides the ant colony into groups and uses differential evolution as the cooperation strategy between groups. It aims to improve the performance of the basic ant colony optimization algorithm in terms of convergence and accuracy when learning Bayesian network structures. The algorithm represents potential advantages such as being effective, accurate, efficient and easy to implement.

Uploaded by

Anjith
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
43 views24 pages

Learning Bayesian Network Structure Based On Ant Colony Optimization and Differential Evolution

This document proposes a new algorithm called cooperative ant colony optimization (coACO) for learning the structure of Bayesian networks. The coACO algorithm divides the ant colony into groups and uses differential evolution as the cooperation strategy between groups. It aims to improve the performance of the basic ant colony optimization algorithm in terms of convergence and accuracy when learning Bayesian network structures. The algorithm represents potential advantages such as being effective, accurate, efficient and easy to implement.

Uploaded by

Anjith
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/ 24

LEARNING BAYESIAN NETWORK

STRUCTURE
BASED ON ANT COLONY OPTIMIZATION
AND DIFFERENTIAL EVOLUTION

1 Aparna S Anand
S7 CS1
Roll no: 25
2 CONTENTS

 INTRODUCTION
 BAYESIAN NETWORK
 PURPOSE OF BAYESIAN NETWORK
 ANT COLONY OPTIMIZATION
 DIFFERENTIAL EVOLUTION
 PROPOSED ALGORITHM
 CONCLUSION
 REFERENCES
3 INTRODUCTION

 Bayesian networks (BNs) can efficiently manage the dependence or


independence relationships between a set of random variables in a
probabilistic way.

 Here, coevolution ACO is used to improve the performance of the


basic ACO in terms of convergence and accuracy in solving the
problem of learning BN structure.
4

 The entire ant colony is divided into several small groups, and the
coevolution mechanism is performed to accomplish the
collaboration of various groups.

 The coevolution mechanism among ant groups is based on the


differential evolution (DE) algorithm.
5 BAYESIAN NETWORK

 A Bayesian network is a probabilistic graphical model that


represents a set of variables and their conditional dependencies via a
directed acyclic graph(DAG).

 Definition : G = (V,E) be a directed acyclic graph (or DAG), and


let X = (Xv), v∈ V be a set of random variablesindexed by V.
6
7
8
9
10 Purpose of BN

Making future predictions


Explaining observations
11 ANT COLONY OPTIMIZATION
 Ant colony optimization algorithm (ACO) is a probabilistic
technique for solving computational problems which can be
reduced to finding good paths through graphs.
12
13
DIFFERENTIAL EVOLUTION

 DE generates new candidates by adding a weighted difference


between two population members to a third member.

 If the resulting candidate is superior to the candidate with which it


was compared, it replaces it; otherwise, the original candidate
remains unchanged.
14 The DE-based coordination operations contain,

(1)Mutation
The mutation operator,

Us= ꞇr1 + F×(ꞇr2 - ꞇr3)

Us(t)={Uijs(t)}, S=1,..,S, is the donor pheromone matrix.

ꞇr1, ꞇr2, ꞇr3 pheromone matrixes of the three randomly


selected ant groups with r1≠r2≠r3≠s.

F is a real positive parameter between [0, 2].


15
(2)Crossover
The crossover operator,

Uijs if rij ≤ cr
Vijs = { ꞇijs otherwise

rij denotes a random value generated for each arc aij in


accordance with a uniform distribution over [0, 1], cr is the
crossover rate between 0 and 1.
16 (3) Selection

ꞇs(t+1) = { vs(t) if f(VS(t))≥f(ꞇs(t))


ꞇs(t) otherwise

ꞇs(t+1) denotes the pheromone matrix of the s-th group


in the (t+1)-th iteration.
17 PROPOSED ALGORITHM

 Step 1. Initialization of parameters : Tmax = maximum number


of iterations; t=0,current iteration counter; Mant=number of ants;
S= number of ant groups.

 Step 2. Initialization of ants : divide the whole ant colony into


different ant groups; ns=number of ants in each group;
ꞇijs(0)=ꞇ0 ;s=1,2,3,..,ns for all arc aij; set G be an empty graph
18
 Step 3. t=t+1;s=1

 Step 4. Perform mutation and crossover operations to the original


pheromone matrix ꞇs(t) of each group and generate trial pheromone
matrix vs(t).

 Step 5. s=s+1; if s<=S,return to Step 4.

 Step 6. s=1;k=1.

 Step 7. The ant k constructs the BN Gk( ꞇs) using pheromone matrix ꞇs.
19
 Step 8. The ant k constructs the BN Gk( vs ) using pheromone matrix vs

 Step 9. k=k+1; if k<= ns, return to Step 7

 Step 10. Calculate the BIC score for the BNs Gk(ꞇs),k=1,2,..,ns, choose
the best BN with the maximum score as the fitness f(ꞇs (t)) for ꞇs

 Step 11. Calculate the BIC score for the BNs Gk(vs),k=1,2,..,ns, choose
the best BN with the maximum score as the fitness f(vs(t)) for vs
20

 Step 12. Compare f(τs(t)) and f(vs(t)), select the better pheromone matrix
according to selection, and select the corresponding better BN for the
ant group.

 Step 13. s=s+1; if s≤S, k=k+1 and return to Step 7.

 Step 14. Select the BN with maximum score from all ant groups, which
is recorded as Gt.

 Step 15. If Gt has the larger BIC score than G+, then G+=Gt.
21

 Step 16. Update the pheromone matrix τs, s=1, 2, …, S, for each ant
group based on the BN G+.

 Step 17. Return to Step 3 until t > Tmax.

 Step 18. Terminate and output the best BN structure.


22 CONCLUSION

 coACO algorithm for learning BNs, which uses DE as the cooperative


strategy and pheromone information as the collaborative factor to guide
the co-evolutionary process of various ant groups.

 coACO based BN structure learning algorithm is,


1. Effective
2. Accurate
3. Efficient
4. Easy to implement.
23 REFERENCES

[1] P.C. Pinto, A Nägele, M Dejori, T.A. Runkler, and J.M.C. Sousa,
“Using a local discovery ant algorithm for Bayesian network structure
learning”, IEEE Transactions on Evolutionary Computation, vol. 13,
no. 4, 2009, pp. 767-779.

[2] L.M. Campos, J.M. Fernández-Luna, J.A. Gámez, and J.M. Puerta,
“Ant colony optimization for learning Bayesian networks”,
International Journal of Approximate Reasoning, vol. 31, no. 3, 2002,
pp. 291-311.
24

THANK YOU

You might also like