190 likes | 430 Views
Linkage Tree Genetic Algorithm. Wei-Ming Chen. Papers. The Linkage Tree Genetic Algorithm, Dirk Thierens , 2010 Pairwise and Problem-Specific Distance Metrics in the Linkage Tree Genetic Algorithm, Martin Pelikan , Mark W. Hauschild , Dirk Thierens , 2011.
E N D
Linkage Tree Genetic Algorithm Wei-Ming Chen
Papers The Linkage Tree Genetic Algorithm, Dirk Thierens, 2010 Pairwise and Problem-Specific Distance Metrics in the Linkage Tree Genetic Algorithm, Martin Pelikan, Mark W. Hauschild, Dirk Thierens, 2011
The Linkage Tree Genetic Algorithm Dirk Thierens GECCO 2010
GA mechanism Evaluation Selection Initialization Until termination Replacement Crossover Mutation
Introduction • Construct the variables to a tree • Hierarchical Clustering • Assign each variable to a single cluster. • Repeat until one cluster left • Join two nearest clusters ciand cj into cij
Clustering Entropy H : Distance D :
Genetic Algorithm Choose a pair of chromosome Crossover mask : apart chromosome into two subsets Replacement : If one of the offspring is better than both of the parents
Algorithm Initial : Create initial population of size N Repeat Build the linkage tree For every pair while the tree is not fully traversed traversed a step and set crossover mask do crossover do replacement if necessary
Result • Test problems • Trap function • NK landscape • Result • The problems are solved in polynomial time • Similar with ECGA and DSMGA
Pairwise and Problem-Specific Distance Metrics in the Linkage Tree Genetic Algorithm Martin Pelikan, Mark W. Hauschild, Dirk Thierens GECCO 2011
local search To improves the quality of the solution In first iteration, do local search before proceeding with the first iteration Based on single-bit neighborhoods choose the step which improves the quality of the solution most Until find the local optimum
Speed up • Original : • Pairwise matrix : • Problem-Specific Metric • decomposable problem composed of m subproblems • prefer decompositions which minimize the sizes of subsets • If two variables in the same subset, the distance of them is 1
Result • Test problems • Trap-5, Trap-6, Trap 7 • NK landscape • 2D spin glass • Result • The problems are solved in polynomial time • Trap functions : almost same • NK landscape : Original< Pairwise < Problem • 2D spin glass : Original < Problem < Pairwise
Conclusion • LTGA : • Small population size • Solve all the problems in low-order polynomial time • Future work : • Problem-specific metrics • Construct all the variables to only one tree ? • Change the minimum mask size ?