Passing values to PSO options

Hi there,

How may I pass swarmSize variable value to opts in the PSO? I have included the particular code piece below. When I do as below, it seems 'swarmSize' doesn't get passed to opts. Many thanks in advance.

swarmSize = 10; %doesn't pass to opts??
opts = optimoptions(@particleswarm,... 
            'Display','iter',...
            'PlotFcn','pswplotbestf',...
            'SwarmSize', swarmSize,...                
            'MaxIterations',20,...
            'UseVectorized',true,...                
            'UseParallel',false);

% Cost {function-handle}
cost = @(particlePosition) costfn(iStep,nn,swarmSize,nvars,particlePosition);
% PSO Call 
[particlePosition,particleCost,exitflag,output] = particleswarm(cost,nvars-2,particlelb,particleub,opts);

Answer 1

[Alan Weiss]

It works for me. Here is a little test script:

fun = @(x)x(1)*exp(-norm(x)^2);
lb = [-5,-5];
ub = -lb;
opts = optimoptions('particleswarm',"SwarmSize",20,"Display","iter");
[sol,fv,ef,output] = particleswarm(fun,2,lb,ub,opts)

                                 Best            Mean     Stall
Iteration     f-count            f(x)            f(x)    Iterations
    0              20        -0.02168        0.008792        0
    1              40         -0.4151        -0.02415        0
    2              60         -0.4151      -2.352e-05        1
    3              80         -0.4151       -0.002245        2
    4             100         -0.4151       -0.000726        3
    5             120         -0.4151           0.015        4
    6             140         -0.4151        0.004306        5
    7             160         -0.4151          0.0146        6
    8             180         -0.4151        -0.04039        7
    9             200         -0.4151        -0.02363        8
   10             220         -0.4151         -0.1728        9
   11             240         -0.4177         -0.3032        0
   12             260         -0.4177         -0.3321        1
   13             280         -0.4206         -0.3799        0
   14             300         -0.4288         -0.4074        0
   15             320         -0.4288         -0.4167        1
   16             340         -0.4288         -0.4254        2
   17             360         -0.4289         -0.4278        0
   18             380         -0.4289         -0.4283        1
   19             400         -0.4289         -0.4286        0
   20             420         -0.4289         -0.4288        1
   21             440         -0.4289         -0.4289        2
   22             460         -0.4289         -0.4289        3
   23             480         -0.4289         -0.4289        0
   24             500         -0.4289         -0.4289        0
   25             520         -0.4289         -0.4289        1
   26             540         -0.4289         -0.4289        0
   27             560         -0.4289         -0.4289        0
   28             580         -0.4289         -0.4289        0
   29             600         -0.4289         -0.4289        1
   30             620         -0.4289         -0.4289        2

                                 Best            Mean     Stall
Iteration     f-count            f(x)            f(x)    Iterations
   31             640         -0.4289         -0.4289        0
   32             660         -0.4289         -0.4289        0
   33             680         -0.4289         -0.4289        0
   34             700         -0.4289         -0.4289        0
   35             720         -0.4289         -0.4289        1
   36             740         -0.4289         -0.4289        0
   37             760         -0.4289         -0.4289        0
   38             780         -0.4289         -0.4289        0
Optimization ended: relative change in the objective value 
over the last OPTIONS.MaxStallIterations iterations is less than OPTIONS.FunctionTolerance.

sol = 1×2

-0.7071 0.0000

fv = -0.4289

ef = 1

output = struct with fields:
      rngstate: [1×1 struct]
    iterations: 38
     funccount: 780
       message: 'Optimization ended: relative change in the objective value ↵over the last OPTIONS.MaxStallIterations iterations is less than OPTIONS.FunctionTolerance.'
    hybridflag: []

You see that there were 38 iterations and 780 function evaluations. This corresponds to a swarm size of 20: 20*38 = 760, plus 20 initial evaluations = 780 function evaluations.

Alan Weiss

MATLAB mathematical toolbox documentation

[Buddhi Wimarshana]

Many thanks Alan, but I am rying to pass this 'swarmSize' from outside the options declare. May be it wasn't clear in my question, so I added a simplified version of the code below:

swarmSize = 10; %want to declare the value outside the opts
opts = optimoptions(@particleswarm,... 
            'Display','iter',...
            'PlotFcn','pswplotbestf',...
            'SwarmSize', swarmSize,... %then pass it to here

[Alan Weiss]

Try the code your way:

rng default
fun = @(x)x(1)*exp(-norm(x)^2);
lb = [-5,-5];
ub = -lb;
swarmSize = 20

swarmSize = 20

opts = optimoptions('particleswarm',"SwarmSize",swarmSize,"Display","iter");
[sol,fv,ef,output] = particleswarm(fun,2,lb,ub,opts)

                                 Best            Mean     Stall
Iteration     f-count            f(x)            f(x)    Iterations
    0              20        -0.00953        0.005882        0
    1              40        -0.02456       -0.001228        0
    2              60        -0.05993        0.001832        0
    3              80        -0.05993      -0.0008519        1
    4             100        -0.08234       -0.005554        0
    5             120         -0.2747         0.02441        0
    6             140         -0.2747         0.02433        1
    7             160         -0.3332        -0.01632        0
    8             180         -0.3332         0.01548        1
    9             200         -0.3332        -0.00039        2
   10             220         -0.3332         0.00387        3
   11             240         -0.3332         0.01008        4
   12             260         -0.3332        -0.01568        5
   13             280         -0.3332        0.005135        6
   14             300         -0.3332      -0.0003027        7
   15             320         -0.3332       -0.006723        8
   16             340         -0.3797         -0.0914        0
   17             360         -0.4253         -0.1851        0
   18             380         -0.4263         -0.2863        0
   19             400         -0.4272         -0.3619        0
   20             420         -0.4285          -0.417        0
   21             440         -0.4287         -0.4215        0
   22             460         -0.4288         -0.4272        0
   23             480         -0.4289         -0.4277        0
   24             500         -0.4289         -0.4282        1
   25             520         -0.4289         -0.4281        0
   26             540         -0.4289         -0.4276        0
   27             560         -0.4289         -0.4279        1
   28             580         -0.4289         -0.4275        2
   29             600         -0.4289         -0.4265        3
   30             620         -0.4289          -0.427        4

                                 Best            Mean     Stall
Iteration     f-count            f(x)            f(x)    Iterations
   31             640         -0.4289         -0.4237        5
   32             660         -0.4289         -0.4249        6
   33             680         -0.4289         -0.4257        7
   34             700         -0.4289         -0.4283        8
   35             720         -0.4289         -0.4286        0
   36             740         -0.4289         -0.4284        0
   37             760         -0.4289         -0.4288        0
   38             780         -0.4289         -0.4289        1
   39             800         -0.4289         -0.4289        0
   40             820         -0.4289         -0.4289        1
   41             840         -0.4289         -0.4289        0
   42             860         -0.4289         -0.4289        0
   43             880         -0.4289         -0.4289        0
   44             900         -0.4289         -0.4289        0
   45             920         -0.4289         -0.4289        0
Optimization ended: relative change in the objective value 
over the last OPTIONS.MaxStallIterations iterations is less than OPTIONS.FunctionTolerance.

sol = 1×2

-0.7071 0.0000

fv = -0.4289

ef = 1

output = struct with fields:
      rngstate: [1×1 struct]
    iterations: 45
     funccount: 920
       message: 'Optimization ended: relative change in the objective value ↵over the last OPTIONS.MaxStallIterations iterations is less than OPTIONS.FunctionTolerance.'
    hybridflag: []

This time (with rng default for reproducibility) I get 45 iterations times 20 particles = 900 fevals, +20 initial fevals = 920 fevals, as shown. The options get passed as you wanted.

Alan Weiss

MATLAB mathematical toolbox documentation

[Buddhi Wimarshana]

Oh silly me!!! have done a silly mistake, it works great...thanks Alan so much...really shouldn't have asked here, but didn't notice the mistake for hours, but I do now