Performance of Multi-Parents Genetic Algorithms (MPGA) for IIR Adaptive System Identification

Open Access
Sun, Guoxin
Graduate Program:
Electrical Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
November 19, 2014
Committee Members:
  • William Kenneth Jenkins, Thesis Advisor
  • Vishal Monga, Thesis Advisor
  • IIR
  • adaptive filter
  • multi-parents GA
  • system identification
Genetic algorithms (GA) are based on the principles of natural selection and natural genetics that originate in biology. The GA has been used for IIR adaptive system identification to deal with its multimodal error surface. However, due to slow convergence rates and high computational complexity, its use for IIR adaptive systems has been limited. This paper proposes a multi-parents genetic algorithm with permutation crossover (PC MPGA) that is a generalization of two-parents genetics used in conventional GA's. Experimental results demonstrate that the proposed PC MPGA does improve the convergence rates in the first several generations but it also increases the mean-square-errors (MSEs) as well. In addition, it requires increased computational complexity. Finally, a modified PC MPGA that decreases the number of parents gradually to 2 when the convergence rate slows down is introduced. Simulation results demonstrate that this modified MPGA can further increase the convergence rate, lower the MSE and reduce the computational complexity, compared with GA.