05 410729 3 22014 iner0a weight vs constric0on

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: w(nt) is the final iner0a weight and w(t) is the iner0a at 0me step t. –  c1=c2=2; Vmax=Xmax. •  idea: large w to explore first and gradually decrease w to exploit the search space when reaching the sweet spots. •  Nonlinear decreasing: shorter explora0on 0me than the linear decreasing method, with more 0me spent on refining solu0ons (exploita0on). •  [Venter and Sobieszczanski ­Sobieski, 2003] •  [Clerc and Kennedy, 2002] developed a Constric0on Coefficient χ for PSO. •  The velocity update equa0on becomes: ω ( t + 1) = α × ω ( t ) where w(0)=1.4, w(n)=0.35, α=0.975 Iner0a weight is adjusted when varia0on of the 20% best € par0cles is < 1.0. χ= 2κ |2 −φ − φ (φ − 4 ) | , φ = c1 + c 2 , φ ≥ 4, κ ∈ [0,1] •  Normally κ=1; c1=c2=2.05; ϕ=4.1;χ=0.729 € 3 2/20/14 Iner0a weight vs. constric0on coefficient •  Both balance explora0on and exp...
View Full Document

This document was uploaded on 02/28/2014 for the course CS 6776 at Memorial University.

Ask a homework question - tutors are online