poll.m

matlab7 gads工具箱

function  [msg,nextIterate,FunEval,Successdir] = poll(ObjFunc,Xin,pollmethod,Completepoll,pollorder ... 
    ,Iterate,Successdir,MeshSize,scale,ToL,A,LB,UB,IndIneqcstr,IndEqcstr,Iter,FunEval,maxFun,FLAG, ... 
    NotVectorized,Cache,cachetol,cachelimit,varargin) 
%POLL: Performs the poll step in GPS. This step gaurantees the local convergence. 
%  
%  OBJFUNC: The objective function on which POLL step is implemented. 
%  
%  POLLMETHOD: Search directions in POLL step is obtained according to 
%  different pollmethod. 
%  
%  COMPLETEPOLL: If 'off' indicates that POLL can be called off as soon  
%  as a better point is found i.e. no sufficient decrease condion is imposed;  
%  If 'on' then ALL the points are evaluated and point with least function value  
%  is returned. Default is 'off'. If function is expensive, make this 'off'. 
%  
%  POLLORDER: Ordering of poll directions. 
%  
%  ITERATE: Incumbent point around which polling will be done. Iterate Stores  
%  the current point 'x' and function value 'f' at this point. 
%  
%  SUCCESSDIR: Last successful POLL/SEARCH direction. This information can be used  
%  by the POLL step in ordering the search direction (last successful 
%  direction is polled first). 
%  
%  MESHSIZE: Current mesh size used in POLL step.  
%  
%  SCALE: Scale factor used to scale the design points. 
%  
%  TOL: Tolerance used for determining whether constraints are active or not. 
%  
%  A,LB,UB: Defines the feasible region in case of linear/bound constraints 
%  as LB<=A*X<=UB. 
%  
%  IndIneqcstr: Logical indices of inequality constraints. A(IndIneqcstr), LB(IndIneqcstr) 
%  UB(IndIneqcstr) represents inequality constraints. 
%  
%  IndEqcstr: Logical indices of equality constraints. A(IndEqcstr), LB(IndEqcstr) 
%  UB(IndEqcstr) represents equality constraints. 
%  
%  FUNEVAL: Counter for number of function evaluations. FunEval is always less than  
%  'MAXFUN', which is maximum number of function evaluation. 
%  
%  MAXFUN: Limit on number of function evaluation. 
%  
%  FLAG: This flag is passed to the SEARCH routines, indicating that the 
%  problem is 'unconstrained', 'boundconstraints', 'linearconstraints'. 
%  
%  NotVectorized: A flag indicating ObjFunc is not evaluated as vectorized 
%  
%  CACHE: A flag for using CACHE. If 'off', no cache is used. 
%  
%  CACHETOL: Tolerance used in cache in order to determine whether two points  
%  are same or not. 
%  
%  CACHELIMIT: Limit the cache size to 'cachelimit'.  
%  
%  NEXTITERATE: Successful iterate after polling is done. If POLL is NOT 
%  successful, NEXTITERATE is same as ITERATE. 
%  
%  MSG: A binary flag indicating, POLL is successful or not. 

%   Copyright 2004 The MathWorks, Inc. 
%   $Revision: 1.15 $  $Date: 2004/01/16 16:49:50 $ 
%   Rakesh Kumar 

%Get the directons which forms the positive basis(minimal or maximal basis) 
if    strcmpi(FLAG, 'unconstrained') 
    Dirvector = uncondirections(pollmethod,Iterate.x); 
elseif strcmpi(FLAG,'boundconstraints')    % Only Box constraints 
    Dirvector = boxdirections(pollmethod,Iterate.x,A,LB,UB,ToL);
    %If the point is on the constraint boundary, we use scale = 1
    if any(activecstr(Iterate.x,A,LB,UB,ToL))
        scale = 1;
    end
elseif strcmpi(FLAG, 'linearconstraints')  %Linear constraints 
    Dirvector = lcondirections(pollmethod,Iterate.x,A,LB,UB,ToL,IndEqcstr); 
    %If the point is on the constraint boundary, we use scale = 1
    if any(activecstr(Iterate.x,A,LB,UB,ToL))
        scale = 1;
    end
end 

%make sure that diretcions does not have any empty entries (safegaurd) 
Dirvector(:,(~any(Dirvector))) = []; 
span = size(Dirvector,2); 
%Get order of direction vector based on polling order choice. (Mark Abramson) 
if strcmpi(pollorder,'consecutive') 
    OrderVec = 1:span; 
elseif strcmpi(pollorder, 'success') 
    Successdir = min(Successdir,span); 
    OrderVec = 1:span; 
    OrderVec = [Successdir OrderVec(1:Successdir-1) OrderVec(Successdir+1:end)]; 
elseif strcmpi(pollorder, 'random') 
    OrderVec = randperm(span); 
else 
    msg = 'Invalid choice of polling order\nSee PSOPTIMSET for valid choice choosing consecutive order'; 
    warning('gads:POLL:invalidPollOrder',msg); 
    OrderVec = 1:span; 
end    

%Get the trial points along the direction vectors;
sites = struct('x',cell(span,1,1),'f',cell(span,1,1)); 
for k = 1:span 
    sites(k).x = Iterate.x + MeshSize*scale.*Dirvector(:,OrderVec(k)); 
end 
%Find an iterate with lower objective  
[msg,nextIterate,direction,FunEval] = nextfeasible(ObjFunc,Xin,sites,Iterate,A,LB,UB,FunEval, ... 
    maxFun,Completepoll,ToL,IndEqcstr,NotVectorized,Cache,cachetol,cachelimit,varargin{:}); 
%Set the successful direction, if a better iterate is found in 'direction' 
if (msg) 
    Successdir = direction; 
else
    Successdir = 1;
end