Skip to content

Commit

Permalink
Update
Browse files Browse the repository at this point in the history
  • Loading branch information
@Fantasy-Dzx
Fantasy-Dzx committed Jul 10, 2024
1 parent d022d07 commit 1274e25
Show file tree
Hide file tree
Showing 9 changed files with 431 additions and 0 deletions.
49 changes: 49 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/AGSEA.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
classdef AGSEA < ALGORITHM
% <multi> <real/integer/binary> <large/none> <constrained/none> <sparse>
% Automated guiding vector selection-based evolutionary algorithm

%------------------------------- Reference --------------------------------
% S. Shao, Y. Tian, and X. Zhang, Deep reinforcement learning assisted
% automated guiding vector selection for large-scale sparse multi-objective
% optimization, Swarm and Evolutionary Computation, 2024.
%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------

methods
function main(Algorithm,Problem)
%% Population initialization
% Calculate the fitness of each decision variable
[Fitness1,TDec,TMask,TempPop] = FitnessCal(Problem,5);
[Population,Dec,Mask,FitnessSpea2] = EnvironmentalSelection(TempPop,TDec,TMask,Problem.N);
num_feature = 14;
max_act = 3;
inputn = zeros(num_feature,1);
outputn = zeros(1,1);
net = newff(inputn,outputn,[10 10 10],{'tansig','purelin'},'trainlm');
Memory = [];
action = 1;
Fitness3 = zeros(1,Problem.D);
Memory = UpdateMemory(Problem,Memory,action,Population,Mask,Population,Mask);
clear TempPop TDec TMask;

%% Optimization
while Algorithm.NotTerminated(Population)
MatingPool = TournamentSelection(2,2*Problem.N,FitnessSpea2);
LastPopulation = Population;
LastMask = Mask;
[action, Fitness,Fitness3] = UsingNet(Problem,Fitness1,net,Memory,num_feature,max_act,Mask,action,Fitness3);
[OffDec,OffMask] = Operator(Problem,Dec(MatingPool,:),Mask(MatingPool,:),Fitness,Mask,num_feature,Memory);
Offspring = Problem.Evaluation(OffDec.*OffMask);
[Population,Dec,Mask,FitnessSpea2] = EnvironmentalSelection([Population,Offspring],[Dec;OffDec],[Mask;OffMask],Problem.N);
Memory = UpdateMemory(Problem,Memory,action,LastPopulation,LastMask,Population,Mask);
net = TrainNet(Problem,net,Memory,num_feature,max_act);
end
end
end
end
45 changes: 45 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/CalFitness.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,45 @@
function Fitness = CalFitness(PopObj)
% Calculate the fitness of each solution

%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------

N = size(PopObj,1);

%% Detect the dominance relation between each two solutions
Dominate = false(N);
for i = 1 : N-1
for j = i+1 : N
k = any(PopObj(i,:)<PopObj(j,:)) - any(PopObj(i,:)>PopObj(j,:));
if k == 1
Dominate(i,j) = true;
elseif k == -1
Dominate(j,i) = true;
end
end
end

%% Calculate S(i)
S = sum(Dominate,2);

%% Calculate R(i)
R = zeros(1,N);
for i = 1 : N
R(i) = sum(S(Dominate(:,i)));
end

%% Calculate D(i)
Distance = pdist2(PopObj,PopObj);
Distance(logical(eye(length(Distance)))) = inf;
Distance = sort(Distance,2);
D = 1./(Distance(:,floor(sqrt(N)))+2);

%% Calculate the fitnesses
Fitness = R + D';
end
55 changes: 55 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/EnvironmentalSelection.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,55 @@
function [Population,Dec,Mask,FitnessSpea2] = EnvironmentalSelection(Population,Dec,Mask,N)
% The environmental selection of SPEA2

%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------


%% Delete duplicated solutions
[~,uni] = unique(Population.objs,'rows');
Population = Population(uni);
Dec = Dec(uni,:);
Mask = Mask(uni,:);
N = min(N,length(Population));

%% Calculate the fitness of each solution
Fitness = CalFitness(Population.objs);

%% Environmental selection
Next = Fitness < 1;
if sum(Next) < N
[~,Rank] = sort(Fitness);
Next(Rank(1:N)) = true;
elseif sum(Next) > N
Del = Truncation(Population(Next).objs,sum(Next)-N);
Temp = find(Next);
Next(Temp(Del)) = false;
end
% Population for next generation
Population = Population(Next);
Fitness = Fitness(Next);
Dec = Dec(Next,:);
Mask = Mask(Next,:);
FitnessSpea2 = Fitness;
end

function Del = Truncation(PopObj,K)
% Select part of the solutions by truncation

%% Truncation
Distance = pdist2(PopObj,PopObj);
Distance(logical(eye(length(Distance)))) = inf;
Del = false(1,size(PopObj,1));
while sum(Del) < K
Remain = find(~Del);
Temp = sort(Distance(Remain,Remain),2);
[~,Rank] = sortrows(Temp);
Del(Remain(Rank(1))) = true;
end
end
69 changes: 69 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/FitnessCal.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,69 @@
function [Fitness,TDec,TMask,TempPop] = FitnessCal(Problem,SampleNum)

%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------

REAL = any(Problem.encoding==1);
TDec = [];
TMask = [];
TempPop = [];
if REAL
% Latin hypercube sampling
Fitness = zeros(1,Problem.D);
DecMat = repmat((Problem.upper - Problem.lower),SampleNum,1).*lhsdesign(SampleNum,Problem.D) - repmat(Problem.lower,SampleNum,1);
for i = 1 : SampleNum
Dec = repmat(DecMat(i,:),Problem.D,1);
Mask = eye(Problem.D);
Population = Problem.Evaluation(Dec.*Mask);
TDec = [TDec;Dec];
TMask = [TMask;Mask];
TempPop = [TempPop,Population];
Fitness = Fitness + sum(Population.objs,2)';
end
% To reduce computation and support parallelism
if Problem.D > 0
AllSample = randperm(length(TempPop));
FinalSample = AllSample(1:Problem.D);
TempPop = TempPop(FinalSample);
TDec = TDec(FinalSample,:);
TMask = TMask(FinalSample,:);
end
Dec = unifrnd(repmat(Problem.lower,Problem.N,1),repmat(Problem.upper,Problem.N,1));
Dec(:,Problem.encoding==4) = 1;
Mask = false(Problem.N,Problem.D);
for i = 1 : Problem.N
Mask(i,TournamentSelection(2,ceil(rand*Problem.D),Fitness)) = 1;
end
Population = Problem.Evaluation(Dec.*Mask);
TDec = [TDec;Dec];
TMask = [TMask;Mask];
TempPop = [TempPop,Population];
else
Fitness = zeros(1,Problem.D);
for i = 1
Dec = ones(Problem.D,Problem.D);
Mask = eye(Problem.D);
Population = Problem.Evaluation(Dec.*Mask);
TDec = [TDec;Dec];
TMask = [TMask;Mask];
TempPop = [TempPop,Population];
Fitness = Fitness + sum(Population.objs,2)';
end
Dec = unifrnd(repmat(Problem.lower,Problem.N,1),repmat(Problem.upper,Problem.N,1));
Dec(:,Problem.encoding==4) = 1;
Mask = false(Problem.N,Problem.D);
for i = 1 : Problem.N
Mask(i,TournamentSelection(2,ceil(rand*Problem.D),Fitness)) = 1;
end
Population = Problem.Evaluation(Dec.*Mask);
TDec = [TDec;Dec];
TMask = [TMask;Mask];
TempPop = [TempPop,Population];
end
end
60 changes: 60 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/GLP_OperatorGAhalf.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,60 @@
function [Offspring,outIndexList,chosengroups] = GLP_OperatorGAhalf(Problem,Parent1,Parent2,numberOfGroups)
% Parent1 and Parent2 are the matrix of decision variables, not solutions

%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------

%% Parameter setting
[proC,disC,~,disM] = deal(1,20,1,20);
[N,D] = size(Parent1);

%% Genetic operators for real encoding
beta = zeros(N,D);
mu = rand(N,D);
beta(mu<=0.5) = (2*mu(mu<=0.5)).^(1/(disC+1));
beta(mu>0.5) = (2-2*mu(mu>0.5)).^(-1/(disC+1));
beta = beta.*(-1).^randi([0,1],N,D);
beta(rand(N,D)<0.5) = 1;
beta(repmat(rand(N,1)>proC,1,D)) = 1;
Offspring = (Parent1+Parent2)/2+beta.*(Parent1-Parent2)/2;
Lower = repmat(Problem.lower,N,1);
Upper = repmat(Problem.upper,N,1);
[outIndexList,~] = CreateGroups(numberOfGroups,Offspring,D);
chosengroups = randi(numberOfGroups,size(outIndexList,1),1);
Site = outIndexList == chosengroups;
mu = rand(N,1);
mu = repmat(mu,1,D);
temp = Site & mu<=0.5;
Offspring = min(max(Offspring,Lower),Upper);
Offspring(temp) = Offspring(temp)+(Upper(temp)-Lower(temp)).*((2.*mu(temp)+(1-2.*mu(temp)).*...
(1-(Offspring(temp)-Lower(temp))./(Upper(temp)-Lower(temp))).^(disM+1)).^(1/(disM+1))-1);
temp = Site & mu>0.5;
Offspring(temp) = Offspring(temp)+(Upper(temp)-Lower(temp)).*(1-(2.*(1-mu(temp))+2.*(mu(temp)-0.5).*...
(1-(Upper(temp)-Offspring(temp))./(Upper(temp)-Lower(temp))).^(disM+1)).^(1/(disM+1)));
end

function [outIndexArray,numberOfGroupsArray] = CreateGroups(numberOfGroups, xPrime, numberOfVariables)
% Creat groups by ordered grouping

outIndexArray = [];
numberOfGroupsArray = [];
noOfSolutions = size(xPrime,1);
for sol = 1 : noOfSolutions
varsPerGroup = floor(numberOfVariables/numberOfGroups);
vars = xPrime(sol,:);
[~,I] = sort(vars);
outIndexList = ones(1,numberOfVariables);
for i = 1 : numberOfGroups-1
outIndexList(I(((i-1)*varsPerGroup)+1:i*varsPerGroup)) = i;
end
outIndexList(I(((numberOfGroups-1)*varsPerGroup)+1:end)) = numberOfGroups;
outIndexArray = [outIndexArray;outIndexList];
numberOfGroupsArray = [numberOfGroupsArray;numberOfGroups];
end
end
53 changes: 53 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/Operator.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,53 @@
function [OffDec,OffMask] = Operator(Problem,ParentDec,ParentMask,Fitness,Mask,num_feature,Memory)
% The operator of AGSEA

%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------

%% Parameter setting
[N,~] = size(ParentDec);
Parent1Dec = ParentDec(1:floor(end/2),:);
Parent2Dec = ParentDec(floor(end/2)+1:floor(end/2)*2,:);
Parent1Mask = ParentMask(1:floor(end/2),:);
Parent2Mask = ParentMask(floor(end/2)+1:floor(end/2)*2,:);

VaryGroup = kmeans(Fitness',2)';
MaxGroup = max(VaryGroup);

%% Crossover and mutation for dec
if any(Problem.encoding~=4)
[OffDec,~,~] = GLP_OperatorGAhalf(Problem,Parent1Dec,Parent2Dec,4); % 4 -- numberofgroups
OffDec(:,Problem.encoding==4) = 1;
else
OffDec = ones(size(Parent1Dec));
end

%% Crossover for mask
OffMask = Parent1Mask;
for i = 1 : N/2
SelectedGroup = randi(MaxGroup,1);
index = xor(Parent1Mask(i,:),Parent2Mask(i,:));
if rand < 0.5
index = (SelectedGroup == VaryGroup) & index & rand(1,Problem.D) < 1;
OffMask(i,index) = 0;
else
index = (SelectedGroup == VaryGroup) & index & rand(1,Problem.D) < 1;
OffMask(i,index) = 1;
end
end
for i = 1 : N/2
if rand < 0.5
index = rand(1,Problem.D) < (100*mean(Mask,'all'))/(Problem.D*length(unique(Memory(max(end-10,1):end,num_feature)))^2);
OffMask(i,index) = 0;
else
index = rand(1,Problem.D) < (100*mean(Mask,'all'))/(Problem.D*length(unique(Memory(max(end-10,1):end,num_feature)))^2);
OffMask(i,index) = 1;
end
end
end
37 changes: 37 additions & 0 deletions PlatEMO/Algorithms/Multi-objective optimization/AGSEA/TrainNet.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,37 @@
function net = TrainNet(Problem,net,Memory,num_feature,max_act)

%------------------------------- Copyright --------------------------------
% Copyright (c) 2024 BIMK Group. You are free to use the PlatEMO for
% research purposes. All publications which use this platform or any code
% in the platform should acknowledge the use of "PlatEMO" and reference "Ye
% Tian, Ran Cheng, Xingyi Zhang, and Yaochu Jin, PlatEMO: A MATLAB platform
% for evolutionary multi-objective optimization [educational forum], IEEE
% Computational Intelligence Magazine, 2017, 12(4): 73-87".
%--------------------------------------------------------------------------

inputn = Memory(:,1:num_feature)';
outputn = Memory(:,num_feature+1)';
net.trainParam.epochs = 100;
net.trainParam.lr = 0.001;
net.trainParam.goal = 0.001;
net.trainParam.showWindow = 0;
output = zeros(1,max_act);
alpha = 0.1;

if size(Memory,1) == ceil(0.25*Problem.maxFE/100)
net = newff(inputn,outputn,[10 10 10],{'tansig','purelin'},'trainlm');
end
if size(Memory,1) > ceil(0.25*Problem.maxFE/100) && mod(size(Memory,1),10) == 0
MTrain = Memory(end-9:end,:);
for i = 1 : size(MTrain,1)
for j = 1 : max_act
input = [MTrain(i,(num_feature+2):end) j]';
output(j) = sim(net,input);
end
MTrain(i,num_feature+1) = MTrain(i,num_feature+1) + alpha*max(output);
end
inputn = MTrain(:,1:num_feature)';
outputn = MTrain(:,num_feature+1)';
net = train(net,inputn,outputn);
end
end
Loading

0 comments on commit 1274e25

Please sign in to comment.