selectparents              package:seao              R Documentation

_S_e_l_e_c_t _t_h_e _p_a_r_e_n_t_s _t_o _m_a_t_e

_D_e_s_c_r_i_p_t_i_o_n:

     Select the parents which are the basis for the next (child)
     generation. The individuals (set-ups) with highest fitness have
     the highest chance to be chosen, so some parents may be chosen a
     few times.

_U_s_a_g_e:

       selectparents(struc.ea, gen=NULL,
                     method=list(base="fit", rescale=0))

_A_r_g_u_m_e_n_t_s:

struc.ea: An evolutionary algorithm structure

     gen: The number of the generation (batch) for which the parents
          have to be chosen. The standard value is the last generation.

  method: A list containing the base for selection, and possible
          selection. The base can be either "fit" (standard value) or
          "rank"; rescaling a number, standard value is 0, but 1.5 is a
          value which gives often nice results.

_D_e_t_a_i_l_s:

     The method is a list of 'base' and 'rescale'. 'base' gives which
     basis should be used for calculating the weights of the different
     individuals:

     "_f_i_t_n_e_s_s" The fitness values are the weights given, before
          rescaling

     "_r_a_n_k" The individuals are ranked from 1 to the number of
          individuals in the generation. The higher the number, the
          higher the fitness of the individual. If two individuals have
          exactly the same fitness, the average of the next and
          previous ranking is used, e.g. imagene that two fitnesses
          should be at ranking 7 and 8. However, they have the same
          value, so they are both set to 7.5 (in this case, the sum is
          still 15).

     In many cases, the difference between the lowest and highest
     weights are very high. This would result in choosing the same
     parent(s) every time. In order to decrease this extreme selection,
     one can rescale the weights. Rescaling is done so that the highest
     fitness is 'recale' times the average value of the fitnesses.

_V_a_l_u_e:

     The function returns an evolutionary algorithm object, which is
     actually the same as the input argument 'struc.ea', but with new
     individuals (experimental set-ups) and the selection method added
     in the generation (batch) of selection:  

   genes: a list containing the genome (parameter set) stucture. This
          is the same as the genome (parameter set) structure of in the
          input ('struc.ea$genes').

generations: a list of structures with information of each generation: 

          _f_i_t: a vector containing the fitnesses of all individuals
               (experimental set-ups) of the generation (batch); this
               are just copies of the fitnesses of the selected
               parents.

          _a_l_l_e_l_e: a matrix containing the allele-values of all
               individuals (experimental set-ups) of the generation
               (batch); this are just copies of the alleles (parameter
               values) of the selected parents. 

          _p_a_r_e_n_t_s: a string vector with the numbers of the parent
               generations.

          _s_e_l_e_c_t_i_o_n: a list containing the selection base and
               rescaling.

_N_o_t_e:

     As for all functions in this package, evolutionary terminology is
     used as described in 'seao.terminology'.

_A_u_t_h_o_r(_s):

     Kurt Sys (kurt.sys@rug.ac.be)

_S_e_e _A_l_s_o:

     genomestruc, newgen, chooseparents

_E_x_a_m_p_l_e_s:

