Boundary mutation replaces a real-valued gene with a random value drawn uniformly between its lower bound (LB) and upper bound (UB). It emphasizes exploration near the boundaries of the search space and can help when optimal solutions lie near variable limits. section 8.2.2
Boundary mutation differs from other real-valued mutation methods:
- Gaussian and Cauchy mutation add random offsets to the existing gene; Cauchy mutation more often produces large changes. section 8.2.2
- Polynomial mutation tends to keep offspring near the parent when its distribution index is high, favoring exploitation. section 8.2.2
Curve-fitting context
A real-valued genetic algorithm can optimize the four coefficients of a third-order polynomial by minimizing mean squared error. Boundary mutation is one available mutation strategy for such bounded coefficient searches. section 8.2.2
Table 8.2 Curve-fitting problem data (view table figure)
| x |
0 |
1.25 |
2.5 |
3.75 |
5 |
| y |
1 |
5.22 |
23.5 |
79.28 |
196 |
Inspect the sample x and y values used in the curve-fitting example. section 8.2.2
