GODLIKE Global optimizer that combines the power
of a Genetic Algorithm, Diffential Evolution,
Particle Swarm Optimization and Adaptive
Simulated Annealing algorithms.
Usage:
(Single-objective optimization)
================================
sol = GODLIKE(obj_fun, popsize, lb, ub)
sol = GODLIKE(..., ub, which_ones)
sol = GODLIKE(..., which_ones, options)
sol = GODLIKE(..., which_ones, 'option', value, ...)
[sol, fval] = GODLIKE(...)
[sol, fval, exitflag] = GODLIKE(...)
[sol, fval, exitflag, output] = GODLIKE(...)
(Multi-objective optimization)
==============================
sol = GODLIKE(obj_fun12..., popsize, lb, ub)
sol = GODLIKE({obj_fun1, obj_fun2,...}, popsize, lb, ub)
sol = GODLIKE(..., ub, which_ones, options)
sol = GODLIKE(..., which_ones, 'option', value, ...)
[sol, fval] = GODLIKE(...)
[..., fval, Pareto_front] = GODLIKE(...)
[..., Pareto_front, Pareto_Fvals] = GODLIKE(...)
[..., Pareto_Fvals, exitflag] = GODLIKE(...)
[..., exitflag, output] = GODLIKE(...)
INPUT ARGUMENTS:
================
obj_fun The objective function of which the global minimum
will be determined (function_handle). For multi-
objective optimization, several objective functions
may be provided as a cell array of function handles,
or alternatively, in a single function that returns
the different function values along the second
dimension.
Objective functions must accept either a [popsize x
dimensions] matrix argument, or a [1 x dimensions]
vector argument, and return a [popsize x number of
objectives] matrix or [1 x number of objective]
vector of associated function values (number of
objectives may be 1). With the first format, the
function is evaluated vectorized, in the second
case CELLFUN() is used, which is a bit slower in
general.
popsize positive integer. Indicates the TOTAL population
size, that is, the number of individuals of all
populations combined.
lb, ub The lower and upper bounds of the problem's search
space, for each dimension. May be scalar in case all
bounds in all dimensions are equal. Note that at
least ONE of these must have a size of [1 x
dimensions], since the problem's dimensionality is
derived from it.
which_ones The algorithms to be used in the optimizations. May
be a single string, e.g., 'DE', in which case the
optimization is equal to just running a single
Differential Evolution optimization. May also be a
cell array of strings, e.g., {'DE'; 'GA'; 'ASA'},
which uses all the indicated algorithms. When
omitted or left empty, defaults to {'DE';'GA';'PSO';
'ASA'} (all algorithms once).
options/ Sets the options to be used by GODLIKE. Options may
'option', be a structure created by set_options, or given as
value individual ['option', value] pairs. See set_options
for a list of all the available options and their
defaults.
OUTPUT ARGUMENTS:
=================
sol The solution that minizes the problem globally,
of size [1 x dimensions]. For multi-objective
optimization, this indicates the point with the
smallest distance to the (shifted) origin.
fval The globally minimal function value
exitflag Additional information to facilitate fully automated
optimization. Negative is `bad', positive `good'. A
value of '0' indicates GODLIKE did not perform any
operations and exited prematurely. A value of '1'
indicates normal exit conditions. A value of '-1'
indicates a premature exit due to exceeding the preset
maximum number of function evaluations. A value of
'-2' indicates that the amount of maximum GODLIKE
iterations has been exceeded, and a value of '-3'
indicates no optimum has been found (only for single-
objective optimization).
output structure, containing much additional information
about the optimization as a whole; see the manual
for a more detailed description.
(For multi-objective optimization only)
Pareto_front, Pareto_Fvals
The full set of non-dominated solutions, and their
associated function values.
See also pop_single, pop_multi, set_options.