27.06.2013 Views

Evolution and Optimum Seeking

Evolution and Optimum Seeking

Evolution and Optimum Seeking

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

R<strong>and</strong>om Strategies 101<br />

the principles of population, sexual inheritance, recombination, dominance, <strong>and</strong> recessiveness<br />

to improve the convergence behavior yield the hoped for breakthrough. He thus<br />

eventually resigns himself to a largely deterministic strategy. In the linear programming<br />

problem, he chooses from the starting point several r<strong>and</strong>om directions <strong>and</strong> follows these in<br />

turn up to the boundary of the feasible region. The best states on the individual bounding<br />

hyperplanes are used to determine a new starting point by taking the arithmetic mean<br />

of the component parameters. Because of the convexity of the allowed region, the new<br />

starting point isalways within it. The simultaneous choice of several search directions<br />

is supposed to be the analogue of the population principle <strong>and</strong> the construction of the<br />

average the analogue of recombination in sexual propagation. To tackle the problem of<br />

nding the minimum or maximum of an unconstrained, non-linear function, Bremermann<br />

even applies a ve point Lagrangian interpolation to determine relative extrema in the<br />

r<strong>and</strong>om directions.<br />

Rechenberg's evolution strategy changes all the components of the variable vector at<br />

each mutation. In his case, the low mutation rate for many dimensions is expressed by<br />

choosing small values for the step lengths, or the spread in the r<strong>and</strong>om changes. On the<br />

basis of theoretical work with two model functions he nds that the st<strong>and</strong>ard deviations<br />

of the r<strong>and</strong>om components are set optimally when they are inversely proportional to the<br />

number of parameters. His two membered evolution strategy resembles the scheme of<br />

Schumer <strong>and</strong> Steiglitz (1968), which isacknowledged to be particularly good, except that<br />

a(0 2 ) normally distributed r<strong>and</strong>om quantity replaces the xed step length s. He has<br />

also added to it a step length modi cation rule, again derived from theory, whichmakes<br />

this look a very promising search method. It is re ned in Chapter 5, Section 5.1 to meet<br />

the requirements of numerical optimization with digital computers. Amultimembered<br />

strategy is treated in Section 5.2, which follows the same basic concept however, by imitating<br />

the principles of population <strong>and</strong> recombination, it can operate without external<br />

control of the step lengths. Incorporating more than one descendant at a time <strong>and</strong> forgetting<br />

\parental wisdom" at the end of each iteration loop has provoked erce objections<br />

against a more natural evolution strategy.<br />

Box (1957) also considers that his EVOP (evolutionary operation) strategy resembles<br />

the biological mutation-selection process. He regards the vertices of his pattern of<br />

trial points, of which the best becomes the center of the next pattern, as individuals of<br />

a population, of which only the best \survives." The \o spring" are, however, generated<br />

by purely deterministic rules. R<strong>and</strong>om decisions, as used by Satterthwaite (1959a<br />

after Lowe, 1964) in his REVOP (r<strong>and</strong>om evolutionary operation) variant, are actually<br />

explicitly rejected by Box(seeYouden et al., 1959 Satterthwaite, 1959b Budne, 1959<br />

Anscombe, 1959).<br />

From a biological or cybernetic pointofview,Pask (1962, 1971), Schmalhausen (1964),<br />

Berg <strong>and</strong> Timofejew-Ressowski (1964), Dobzhansky (1965), Moran (1967), <strong>and</strong> Kussul <strong>and</strong><br />

Luk (1971) among others have examined the analogy between optimization <strong>and</strong> evolution.<br />

The fact that no practical algorithms have come out of this is no doubt because the<br />

evolutionary processes are described only verbally. Although they sometimes even include<br />

their more subtle e ects, they have so far not produced a really quantitative, predictive<br />

theory. Exceptions, such as the work of Eigen (1971 see also Schuster, 1972), Merzenich

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!