INTERFACE HELP

evolution strategy parameters

• strategy-choice
  Select a strategy for the evolution process. The strategy decides how the parents 'generate' their children and how the selection of parents for the next generation works.
  The table below shows the strategies' characteristics.

  strategy	selection of fittest from	children generated by
  p+c	parents and children	mutation
  p,c	children	mutation
  p/r,c	children	mutation and recombination
  p/r+c	parents and children	mutation and recombination

• parents-scroller
  Select the number of parents. Changes to this also apply to the children scroller. As the number of parents must be greater or equal the number of children, the start-value of the children' scroller's range will be set to the number of parents.
• children-scroller
  Select the number of children. You may not select a value smaller than the number of parents (see: 'parents-scroller').
• recombination-choice
  Evolution strategy basically uses two different ways to 'produce' children. The first one is MUTATION, which randomly changes a value (values) in the indiviual's genotype. The second one is RECOMBINATION, which combinates two(or more) parents to 'produce' one child. This choice applies for the type of recombination used. The types differ in the number of parents used and the way the genotypes of these parents are combined.
  The recombination-choice is set insensitive when a strategy is selected that does not use recombination (see: 'strategy-choice').
  Let Ax, Bx be the x.th value in the genotype of parent A,B and Zx be the x.th value in the genotype of their child Z, then the table below shows the possible types of recombination and their characteristics.

  recombination	discrete	intermediate	global discrete	global intermediate
  number of parents used in recombination	2	2	r >= 2	r >= 2
  result	Zx=Ax or Bx	Zx=1/2*(Ax+Bx)	Zx=Ax or Bx or ...	Zx=1/r(Ax+Bx+...)

• recombinators scroller
  Select the number of recombinators, i.e. the number of parents used in recombination. The scroller is set insensitive when a strategy is selected that does not use recombination (see 'strategy-choice') and when a recombination type is selected that uses exactly two recombinators, i.e. two parents (see 'recombination-type choice').
• alpha
  In evolution strategy the genotype is divided into two parts. The first one, the object parameters, is a vector of real numbers that contains the indiviual's 'blueprint' (its genes). The second one, the strategy parameters, is also a vector of real numbers (having the same length as the first vector) that states if and how a corresponding entry in the object parameters vector may be modified (i.e. mutated). Let O be an object parameter and S be the corresponding strategy parameter. Then if O is mutated, this mutation is given by O+N(0,S) where N(0,S) is the Gaussian Distribution with mean-value 0 and standard derivation S (especially a value of 0 for S doesn't change O).
  The alpha value states how the strategy parameters vector changes during the evolution process. It is therefore called 'strategy parameters adaption value'. If the indiviudal is mutated then each strategy parameter is adapted by multiplying it with a randomly choosen value from alpha and 1/alpha. This means the mutation-'step-size' will be adapted within the evolution process.
  Rechenberg ["Evolution Strategy", 1994, Chapter 14] recommends an alpha-value of 1.3.

end of evolution process

• 'stop at generation' scroller
  Select the maximum generation. The evolution process stops when this generation is reached. It may be continued by increasing the value.
• 'optimum' toggle button
  Stop the evolution process when the optimum is reached. Note, that you may specify a function that returns the optimal fitness value for the phenotypes of a specific proplem. If you do so, this button will be sensitive. However, for most of the problems, you will not know the optimum.

control buttons

• 'quality-graph' button
  Opens a window displaying the graph of quality values. Pressing the button again will close the window.
• 'edit' button
  Opens a window to create/edit a phenotype (however, only if an editor window is provided with the problem implementation). The evolution process can be restarted with the corresponding genotype. You will (should) find more information following a link within the problem's main page.
• 'start' button
  Starts the evolution process.
• 'stop' button
  Stops the evolution process.
• 'continue' button
  Continues a stopped evolution process.

phenotype display

Displays the best indivual of the current generation, together with the generation number and its fitness (smaller fitness values are better, as we always look for the minimum).