410 likes | 1.09k Views
Combining Bacterial Foraging into Genetic Algorithm. Speaker: Pei-Wei Tsai Advisors: Jeng-Shyang Pan Shu-Chuan Chu. Introduction. Concept of Genetic Algorithm (GA) Concept of Bacterial Foraging Our Purpose – Combining BF with GA. Introduction of the Existing Method
E N D
Combining Bacterial Foraging into Genetic Algorithm Speaker: Pei-Wei Tsai Advisors: Jeng-Shyang Pan Shu-Chuan Chu
Introduction • Concept of Genetic Algorithm (GA) • Concept of Bacterial Foraging • Our Purpose – Combining BF with GA. • Introduction of the Existing Method • Our Method • Experiments and Experimental Results • Conclusion
When and Where to Use Evolutionary Algorithms • If you are facing a problem, which has amounts of solutions, and the optimal one can be found via evaluating several conditions or items, then it is called an “optimization problem”. • Actually, optimization problems are everywhere happening in our life.
Genetic Algorithm (GA) • GA is one of the common evolutionary algorithms for optimization problems. • The theory is based on Darwinian Evolutionism. Every species will survive or be eliminated by natural evolution.
Elements of GA • Chromosome: The structure of the solution set. • Population: A class of chromosomes. • Evaluation: The way to find the fitness value of the chromosome. • Selection: Choose the chromosomes, which we want to reserve.
Elements of GA (2) • Crossover: Mix the different chromosomes to create a new one.
Elements of GA (3) • Mutation: Randomly change the value of chromosomes according to a particular probability.
Process of Genetic Algorithm • Randomly create a class of chromosomes. • Evaluate the fitness value of all chromosomes according to the fitness function, and then keep the best one. • Select some specific chromosomes for reserve; others for eliminate. • Process of crossover and mutation. • Termination checking. Repeat step2 to step4 until certain termination is met.
Bacterial Foraging • Moving track of E. Coli:
Bacterial Foraging Algorithm • Bacterium: • Position • Step Sizes for Each Dimension • Health Status
Bacterial Foraging Algorithm(2) • Parameters: • Elimination • Reproduction • Lifetime • Maximum Swim Length • Selection Rate (for Elimination) • Step Size (for Moving) • Attraction and Repellent between cells to cells. (x4)
Bacterial Foraging Algorithm(3) Step1: Initialization. Step2: Evaluation. (fitness function, and then add cell-to-cell affects) Step3: Movement. -Tumbling - Swimming(if Tumbling on a gradient) Step4: Reproduction, go back to Step2 until reaches the reproduction time. Step5: Elimination, go back to Step 2 unless reaches the elimination time.
Our Purpose – Combining GA with BF • We tried to combine GA and BF to produce a new approach of evolutionary algorithm for optimization problems.
Mutated Mutated Mutated The Existing Method • Pseudo-Bacterial GA • T. Furuhashi, Y. Miyata, Y. Uchikawa, “Pseudo-Bacterial Genetic Algorithm and Finding of Fuzzy Rules”, the Second Online Workshop on Evolutionary Computation(1996)(WEC2)
Start Start Initialization Initialization End End Our Purpose – Combining GA with BF(2) • BF: • GA: Evaluation Evaluation Tumble Selection Swim Crossover End of Life? Mutation No Terminate? No Yes Reproduction Yes End of Rep.? No Yes Elimination End of Eli.? No Yes
Our Method -- Idea • Let the bacterium really has its own gene, and then recombine the processes.
Start Initialization Tumble Swim Evaluation Selection (By Swim Length) End Crossover Mutation Our Method – Idea (2) Reproduction End of Rep.? No Yes Elimination End of Life? No End of Eli.? No Yes
Experiments (2) • GA Parameters: • Selection Rate: 0.5 • Mutation Rate: 0.08 • For the first equation: • Population Size: 50 • Iteration: 800 • For the other equations: • Population Size: 160 • Iteration: 2000 / 5000 • BF Parameters: • Elimination: 2 • Reproduction: 4 • Max swim length: 5 • Selection Rate: 0.25 • For the first equation: • Population Size: 50 • Lifetime: 100 • For the other equations: • Population Size: 160 • Lifetime: 250
Experiments (3) • Our Method’s Parameters: • GA Selection Rate: 0.5 • Mutation Rate: 0.08 • Elimination: every 2 iterations • Reproduction: every 4 iterations • BF Selection Rate: 0.25 • Max swim length: 5 • For the first equation: • Population Size: 50 • Iteration: 800 • For the other equations: • Population Size: 160 • Iteration: 2000
Conclusion • To combine Bacterial Foraging with Genetic Algorithm, we applied GA’s chromosome into Bacterium. • This redeems the drawbacks of BF, such as trapped in the local optimum and the slow convergence. • In future experiments, our method still present good performance.