| OCR Text |
Show experimental results o f this scheme with those of the genetic algorithm alone. The hybrid search scheme is implemented in this research by switching the control between the genetic algorithm and the hill climbing according to simple transition rules. The block diag ram o f the adaptive image segmentation system using the hybrid optimization scheme is shown in Figure 6.1. The image segmentation and evaluation components in Figure 4.1 are grouped in one box. Switching o f the control from the genetic algorithm to the hill climbing takes place when the genetic algorithm finds a new maximum point. The new maximum point is used by the hill climbing as a starting point. The hill climber passes the control over to the genetic algorithm when it reaches a local maximum point, i.e., a point that is higher than all o f its adjacent points. The local maximum point replaces the maximum point in the current population which was the starting point o f the hill climbing, and then the genetic algorithm proceeds using the new population containing the local maximum point. To search through a space o f parameter values using hill climbing, we select a starting point, take a step in each o f the fixed set o f directions, move to the best alternative found, and repeat until we reach a point that is higher than all of its adjacent points [1, 45]. A hill climbing algorithm can be described more formally as follows: (1) Select a point xc at random and evaluate it, producing v(xc). (2) Evaluate all the points x lt ..., xn that are adjacent to xc, producing v ^ ) , ..., v(;tn). Let v(xm) be the maximum of v ( jc,-) for / = 1, ..., n. If viXn) > v(xc) then set xc = xm and repeat this step. Otherwise, stop. A set o f points that are "adjacent" to a certain point can be defined in two ways: First, it can denote the set of points that are a Euclidean distance apart from the given point. It is the set o f all neighboring points in a parameter space. Second, it can denote the set of points that are unit Hamming distance apart from the given point. Each point in this set differs by only one bit value from the given point in binary representation o f points. It defines the set o f points with varying step size from the given point. The set of H amming 175 |
