diff --git a/PlatEMO/Algorithms/MOEA-D-DE/MOEADDE.m b/PlatEMO/Algorithms/MOEA-D-DE/MOEADDE.m
index 7626a0815..096bb15cc 100644
--- a/PlatEMO/Algorithms/MOEA-D-DE/MOEADDE.m
+++ b/PlatEMO/Algorithms/MOEA-D-DE/MOEADDE.m
@@ -39,7 +39,7 @@ function MOEADDE(Global)
         for i = 1 : Global.N
             % Choose the parents
             if rand < delta
-                P = B(i,randperm(size(B,2)));
+                P = B(i,randperm(end));
             else
                 P = randperm(Global.N);
             end
diff --git a/PlatEMO/Algorithms/MOEA-D-DRA/MOEADDRA.m b/PlatEMO/Algorithms/MOEA-D-DRA/MOEADDRA.m
index 835891fa8..a92d30bfa 100644
--- a/PlatEMO/Algorithms/MOEA-D-DRA/MOEADDRA.m
+++ b/PlatEMO/Algorithms/MOEA-D-DRA/MOEADDRA.m
@@ -46,7 +46,7 @@ function MOEADDRA(Global)
             for i = I
                 % Choose the parents
                 if rand < 0.9
-                    P = B(i,randperm(size(B,2)));
+                    P = B(i,randperm(end));
                 else
                     P = randperm(Global.N);
                 end
diff --git a/PlatEMO/Algorithms/SparseEA/EnvironmentalSelection.m b/PlatEMO/Algorithms/SparseEA/EnvironmentalSelection.m
new file mode 100644
index 000000000..f42b8b2d8
--- /dev/null
+++ b/PlatEMO/Algorithms/SparseEA/EnvironmentalSelection.m
@@ -0,0 +1,38 @@
+function [Population,Dec,Mask,FrontNo,CrowdDis] = EnvironmentalSelection(Population,Dec,Mask,N)
+% The environmental selection of SparseEA
+
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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));
+
+    %% Non-dominated sorting
+    [FrontNo,MaxFNo] = NDSort(Population.objs,Population.cons,N);
+    Next = FrontNo < MaxFNo;
+    
+    %% Calculate the crowding distance of each solution
+    CrowdDis = CrowdingDistance(Population.objs,FrontNo);
+    
+    %% Select the solutions in the last front based on their crowding distances
+    Last     = find(FrontNo==MaxFNo);
+    [~,Rank] = sort(CrowdDis(Last),'descend');
+    Next(Last(Rank(1:N-sum(Next)))) = true;
+    
+    %% Population for next generation
+    Population = Population(Next);
+    FrontNo    = FrontNo(Next);
+    CrowdDis   = CrowdDis(Next);
+    Dec        = Dec(Next,:);
+    Mask       = Mask(Next,:);
+end
\ No newline at end of file
diff --git a/PlatEMO/Algorithms/SparseEA/Operator.m b/PlatEMO/Algorithms/SparseEA/Operator.m
new file mode 100644
index 000000000..20ffcf9c3
--- /dev/null
+++ b/PlatEMO/Algorithms/SparseEA/Operator.m
@@ -0,0 +1,61 @@
+function [OffDec,OffMask] = Operator(ParentDec,ParentMask,Fitness,REAL)
+% The operator of SparseEA
+
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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,D]       = size(ParentDec);
+    Parent1Mask = ParentMask(1:N/2,:);
+    Parent2Mask = ParentMask(N/2+1:end,:);
+    
+    %% Crossover for mask
+    OffMask = Parent1Mask;
+    for i = 1 : N/2
+        if rand < 0.5
+            index = find(Parent1Mask(i,:)&~Parent2Mask(i,:));
+            index = index(TS(-Fitness(index)));
+            OffMask(i,index) = 0;
+        else
+            index = find(~Parent1Mask(i,:)&Parent2Mask(i,:));
+            index = index(TS(Fitness(index)));
+            OffMask(i,index) = Parent2Mask(i,index);
+        end
+    end
+    
+    %% Mutation for mask
+    for i = 1 : N/2
+        if rand < 0.5
+            index = find(OffMask(i,:));
+            index = index(TS(-Fitness(index)));
+            OffMask(i,index) = 0;
+        else
+            index = find(~OffMask(i,:));
+            index = index(TS(Fitness(index)));
+            OffMask(i,index) = 1;
+        end
+    end
+    
+    %% Crossover and mutation for dec
+    if REAL
+        OffDec = GAhalf(ParentDec);
+    else
+        OffDec = ones(N/2,D);
+    end
+end
+
+function index = TS(Fitness)
+% Binary tournament selection
+
+    if isempty(Fitness)
+        index = [];
+    else
+        index = TournamentSelection(2,1,Fitness);
+    end
+end
\ No newline at end of file
diff --git a/PlatEMO/Algorithms/SparseEA/SparseEA.m b/PlatEMO/Algorithms/SparseEA/SparseEA.m
new file mode 100644
index 000000000..3d91d9683
--- /dev/null
+++ b/PlatEMO/Algorithms/SparseEA/SparseEA.m
@@ -0,0 +1,58 @@
+function SparseEA(Global)
+% <algorithm> <S>
+% Evolutionary algorithm for sparse multi-objective optimization problems
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    %% Population initialization
+    % Calculate the fitness of each decision variable
+    TDec    = [];
+    TMask   = [];
+    TempPop = [];
+    Fitness = zeros(1,Global.D);
+    REAL    = ~strcmp(Global.encoding,'binary');
+    for i = 1 : 1+4*REAL
+        if REAL
+            Dec = unifrnd(repmat(Global.lower,Global.D,1),repmat(Global.upper,Global.D,1));
+        else
+            Dec = ones(Global.D,Global.D);
+        end
+        Mask       = eye(Global.D);
+        Population = INDIVIDUAL(Dec.*Mask);
+        TDec       = [TDec;Dec];
+        TMask      = [TMask;Mask];
+        TempPop    = [TempPop,Population];
+        Fitness    = Fitness + NDSort([Population.objs,Population.cons],inf);
+    end
+    % Generate initial population
+    if REAL
+        Dec = unifrnd(repmat(Global.lower,Global.N,1),repmat(Global.upper,Global.N,1));
+    else
+        Dec = ones(Global.N,Global.D);
+    end
+    Mask = zeros(Global.N,Global.D);
+    for i = 1 : Global.N
+        Mask(i,TournamentSelection(2,ceil(rand*Global.D),Fitness)) = 1;
+    end
+    Population = INDIVIDUAL(Dec.*Mask);
+    [Population,Dec,Mask,FrontNo,CrowdDis] = EnvironmentalSelection([Population,TempPop],[Dec;TDec],[Mask;TMask],Global.N);
+
+    %% Optimization
+    while Global.NotTermination(Population)
+        MatingPool       = TournamentSelection(2,2*Global.N,FrontNo,-CrowdDis);
+        [OffDec,OffMask] = Operator(Dec(MatingPool,:),Mask(MatingPool,:),Fitness,REAL);
+        Offspring        = INDIVIDUAL(OffDec.*OffMask);
+        [Population,Dec,Mask,FrontNo,CrowdDis] = EnvironmentalSelection([Population,Offspring],[Dec;OffDec],[Mask;OffMask],Global.N);
+    end
+end
\ No newline at end of file
diff --git a/PlatEMO/GUI/Controls/newAxes.m b/PlatEMO/GUI/Controls/newAxes.m
index fa06ad918..c8048c377 100644
--- a/PlatEMO/GUI/Controls/newAxes.m
+++ b/PlatEMO/GUI/Controls/newAxes.m
@@ -72,10 +72,10 @@ function draw(obj,Data,varargin)
             if ~obj.viewFix
                 if M == 2
                     view(0,90);
-                    xlabel('\itf\rm_1'); ylabel('\itf\rm_2');
+                    xlabel('\it f\rm_1'); ylabel('\it f\rm_2');
                 elseif M == 3
                     view(135,30);
-                    xlabel('\itf\rm_1'); ylabel('\itf\rm_2'); zlabel('\itf\rm_3');
+                    xlabel('\it f\rm_1'); ylabel('\it f\rm_2'); zlabel('\it f\rm_3');
                 elseif M > 3
                     view(0,90);
                     xlabel('Dimension No.'); ylabel('Value');
diff --git a/PlatEMO/GUI/Controls/newButtonT.m b/PlatEMO/GUI/Controls/newButtonT.m
index e6371a5b9..0a3ec7cfd 100644
--- a/PlatEMO/GUI/Controls/newButtonT.m
+++ b/PlatEMO/GUI/Controls/newButtonT.m
@@ -37,7 +37,7 @@
             end
             obj@newButton(parent,pos,str,varargin{:});
             obj.icon = icon;
-            tiplabel = uicontrol('Parent',obj.figure.handle,'Style','text','Units','pixels','Position',[0 0 1 1],'FontSize',10,'String',tip,'BackgroundColor',[1 1 .9]);
+            tiplabel = uicontrol('Parent',obj.figure.handle,'Style','text','Units','pixels','Position',[0 0 1 1],'FontSize',9,'String',tip,'BackgroundColor',[1 1 .9]);
             obj.tip  = uipanel('Parent',obj.figure.handle,'Units','pixels','Position',[tiplabel.Extent(1:3),16],'HighlightColor',[.5 .5 .5],'BorderType','line','Visible','off');
             set(tiplabel,'Parent',obj.tip,'Position',[0 3 tiplabel.Extent(3) 13]);
             obj.updateCData();
diff --git a/PlatEMO/GUI/Modules/GUI.m b/PlatEMO/GUI/Modules/GUI.m
index 981dddf83..3d0c6cdce 100644
--- a/PlatEMO/GUI/Modules/GUI.m
+++ b/PlatEMO/GUI/Modules/GUI.m
@@ -28,7 +28,7 @@
             % Read the function lists
             obj.readList();
             % Create the figure
-            obj.figure = newFigure([0 0 1200 650],'PlatEMO v2.0');
+            obj.figure = newFigure([0 0 1200 650],'PlatEMO v2.1');
             % Create the menu
             obj.menu = newTab(obj.figure,[0 551 1202 100,1 1 0 1],{'Modules','Help'});
             obj.figure.busy = false;
diff --git a/PlatEMO/GUI/Modules/module_experiment.m b/PlatEMO/GUI/Modules/module_experiment.m
index aed6cd38b..de20e942f 100644
--- a/PlatEMO/GUI/Modules/module_experiment.m
+++ b/PlatEMO/GUI/Modules/module_experiment.m
@@ -67,7 +67,7 @@
             obj.control.showButton     = newPopmenu3(obj.control.tableToolBar,[681 1 105 25,0 1 1 0],[{obj.GUI.icons.runTimes,[],obj.GUI.icons.runtime,[]},cell(1,length(obj.GUI.metList{1,2})+length(obj.GUI.metList{2,2}))],...
                                                      [{'Run times','','runtime',''},obj.GUI.metList{2,2}',obj.GUI.metList{1,2}'],'callback',@(~,~)obj.flushTable);
             obj.control.table          = newTable(panel,[24 90 788 390,1 1 1 1],'Fontname','Times New Roman','Fontsize',9,'ButtonDownFcn',@obj.cb_table);
-            obj.control.tableMenu      = newMenu(panel,[1 1 245 20]);
+            obj.control.tableMenu      = newMenu(panel,[1 1 250 20]);
             obj.control.tableMenu.add(obj.GUI.icons.showPF,'PF with median metric value','callback',@(~,~)obj.cb_tableMenu(1));
             obj.control.tableMenu.add([],'PF with best metric value','callback',@(~,~)obj.cb_tableMenu(2));
             obj.control.tableMenu.add([],'');
diff --git a/PlatEMO/GUI/Modules/module_test.m b/PlatEMO/GUI/Modules/module_test.m
index e10f1f479..cd9b322c1 100644
--- a/PlatEMO/GUI/Modules/module_test.m
+++ b/PlatEMO/GUI/Modules/module_test.m
@@ -50,7 +50,7 @@
             obj.control.figureMenu = newMenu(panel,[1 1 155 20]);
             obj.control.figureMenu.add([],'Show in new figure','callback',@obj.cb_figureMenu);
             newTip(obj.control.axesToolBar,[1 1 12 25,1 0 1 0],x,'callback',@obj.cb_figureMenuButton);
-            obj.control.showButton  = newPopmenu3(obj.control.axesToolBar,[453 1 105 25,0 1 1 0],[{obj.GUI.icons.showPF,obj.GUI.icons.showPS,[],obj.GUI.icons.showTPF,[]},cell(1,length(obj.GUI.metList{1,2})+length(obj.GUI.metList{2,2}))],...
+            obj.control.showButton  = newPopmenu3(obj.control.axesToolBar,[443 1 115 25,0 1 1 0],[{obj.GUI.icons.showPF,obj.GUI.icons.showPS,[],obj.GUI.icons.showTPF,[]},cell(1,length(obj.GUI.metList{1,2})+length(obj.GUI.metList{2,2}))],...
                                                   [{'Pareto Front','Pareto Set','','True PF',''},obj.GUI.metList{2,2}',obj.GUI.metList{1,2}'],'callback',@obj.cb_showbutton);
             obj.control.axes        = newAxes(panel,[82 108 452 360,1 1 1 1]);
             obj.control.playbar     = newPlay(panel,[20 32 570 27,1 1 1 0],'callback',@obj.cb_playbar);
diff --git a/PlatEMO/Metrics/CPF.m b/PlatEMO/Metrics/CPF.m
index 02929acdb..deb4f9ee0 100644
--- a/PlatEMO/Metrics/CPF.m
+++ b/PlatEMO/Metrics/CPF.m
@@ -1,6 +1,6 @@
 function Score = CPF(PopObj,PF)
 % <metric> <max>
-% Coverage of Pareto front
+% Coverage over Pareto front
 
 %------------------------------- Reference --------------------------------
 % Y. Tian, R. Cheng, X. Zhang, M. Li, and Y. Jin, Diversity assessment of
diff --git a/PlatEMO/Problems/Sparse MOPs/Dataset_CN.mat b/PlatEMO/Problems/Sparse MOPs/Dataset_CN.mat
new file mode 100644
index 000000000..e2d4210a8
Binary files /dev/null and b/PlatEMO/Problems/Sparse MOPs/Dataset_CN.mat differ
diff --git a/PlatEMO/Problems/Sparse MOPs/Dataset_FS_NN.mat b/PlatEMO/Problems/Sparse MOPs/Dataset_FS_NN.mat
new file mode 100644
index 000000000..c67a0e4d7
Binary files /dev/null and b/PlatEMO/Problems/Sparse MOPs/Dataset_FS_NN.mat differ
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP1.m b/PlatEMO/Problems/Sparse MOPs/SMOP1.m
new file mode 100644
index 000000000..48e451928
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP1.m	
@@ -0,0 +1,57 @@
+classdef SMOP1 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP1()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sum(g1(X(:,M:M+K-1),pi/3),2) + sum(g2(X(:,M+K:end),0),2);
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),X(:,1:M-1)],2)).*[ones(N,1),1-X(:,M-1:-1:1)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            P = UniformPoint(N,obj.Global.M);
+        end
+    end
+end
+
+function g = g1(x,t)
+    g = (x-t).^2;
+end
+
+function g = g2(x,t)
+    g = 2*(x-t).^2 + sin(2*pi*(x-t)).^2;
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP2.m b/PlatEMO/Problems/Sparse MOPs/SMOP2.m
new file mode 100644
index 000000000..8dc8640c4
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP2.m	
@@ -0,0 +1,57 @@
+classdef SMOP2 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP2()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sum(g2(X(:,M:M+K-1),pi/3),2) + sum(g3(X(:,M+K:end),0),2);
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),X(:,1:M-1)],2)).*[ones(N,1),1-X(:,M-1:-1:1)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            P = UniformPoint(N,obj.Global.M);
+        end
+    end
+end
+
+function g = g2(x,t)
+    g = 2*(x-t).^2 + sin(2*pi*(x-t)).^2;
+end
+
+function g = g3(x,t)
+    g = 4-(x-t)-4./exp(100*(x-t).^2);
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP3.m b/PlatEMO/Problems/Sparse MOPs/SMOP3.m
new file mode 100644
index 000000000..0c2b1c0db
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP3.m	
@@ -0,0 +1,57 @@
+classdef SMOP3 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP3()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sum(g1(X(:,M:M+K-1),pi/3),2);
+            for i = 1 : ceil((D-M-K+1)/10)
+                temp = 50 - sum(g1(X(:,M+K+(i-1)*10:min(M+K+i*10-1,end)),0),2);
+                g(temp<50) = g(temp<50) + temp(temp<50);
+            end
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),X(:,1:M-1)],2)).*[ones(N,1),1-X(:,M-1:-1:1)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            P = UniformPoint(N,obj.Global.M);
+        end
+    end
+end
+
+function g = g1(x,t)
+    g = (x-t).^2;
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP4.m b/PlatEMO/Problems/Sparse MOPs/SMOP4.m
new file mode 100644
index 000000000..86dcf5160
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP4.m	
@@ -0,0 +1,64 @@
+classdef SMOP4 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP4()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sort(g3(X(:,M:end),0),2);
+            g = sum(g(:,1:D-M-K+1),2);
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),1-cos(X(:,1:M-1)*pi/2)],2)).*[ones(N,1),1-sin(X(:,M-1:-1:1)*pi/2)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            M = obj.Global.M;
+            P = UniformPoint(N,M);
+            c = ones(size(P,1),M);
+            for i = 1 : size(P,1) 
+                for j = 2 : M
+                    temp = P(i,j)/P(i,1)*prod(1-c(i,M-j+2:M-1));
+                    c(i,M-j+1) = (temp^2-temp+sqrt(2*temp))/(temp^2+1);
+                end
+            end
+            x = acos(c)*2/pi;
+            P = fliplr(cumprod([ones(size(x,1),1),1-cos(x(:,1:end-1)*pi/2)],2)).*[ones(size(x,1),1),1-sin(x(:,end-1:-1:1)*pi/2)];
+        end
+    end
+end
+
+function g = g3(x,t)
+    g = 4-(x-t)-4./exp(100*(x-t).^2);
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP5.m b/PlatEMO/Problems/Sparse MOPs/SMOP5.m
new file mode 100644
index 000000000..d237bfb3b
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP5.m	
@@ -0,0 +1,67 @@
+classdef SMOP5 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP5()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sum(g1(X(:,M:end),pi/3).*g2(X(:,M:end),0),2) + abs(K-sum(X(:,M:end)~=0,2));
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),1-cos(X(:,1:M-1)*pi/2)],2)).*[ones(N,1),1-sin(X(:,M-1:-1:1)*pi/2)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            M = obj.Global.M;
+            P = UniformPoint(N,M);
+            c = ones(size(P,1),M);
+            for i = 1 : size(P,1) 
+                for j = 2 : M
+                    temp = P(i,j)/P(i,1)*prod(1-c(i,M-j+2:M-1));
+                    c(i,M-j+1) = (temp^2-temp+sqrt(2*temp))/(temp^2+1);
+                end
+            end
+            x = acos(c)*2/pi;
+            P = fliplr(cumprod([ones(size(x,1),1),1-cos(x(:,1:end-1)*pi/2)],2)).*[ones(size(x,1),1),1-sin(x(:,end-1:-1:1)*pi/2)];
+        end
+    end
+end
+
+function g = g1(x,t)
+    g = (x-t).^2;
+end
+
+function g = g2(x,t)
+    g = 2*(x-t).^2 + sin(2*pi*(x-t)).^2;
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP6.m b/PlatEMO/Problems/Sparse MOPs/SMOP6.m
new file mode 100644
index 000000000..64dd9080a
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP6.m	
@@ -0,0 +1,71 @@
+classdef SMOP6 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP6()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = g4(X(:,M:end),repmat(linspace(0,1,D-M+1),N,1));
+            [g,rank] = sort(g,2);
+            temp = false(size(rank));
+            for i = 1 : size(rank,1)
+                temp(i,X(i,M-1+rank(i,:))==0) = true;
+            end
+            temp(:,1:K) = false;
+            g(temp) = 0;
+            g = sum(g,2);
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),1-cos(X(:,1:M-1)*pi/2)],2)).*[ones(N,1),1-sin(X(:,M-1:-1:1)*pi/2)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            M = obj.Global.M;
+            P = UniformPoint(N,M);
+            c = ones(size(P,1),M);
+            for i = 1 : size(P,1) 
+                for j = 2 : M
+                    temp = P(i,j)/P(i,1)*prod(1-c(i,M-j+2:M-1));
+                    c(i,M-j+1) = (temp^2-temp+sqrt(2*temp))/(temp^2+1);
+                end
+            end
+            x = acos(c)*2/pi;
+            P = fliplr(cumprod([ones(size(x,1),1),1-cos(x(:,1:end-1)*pi/2)],2)).*[ones(size(x,1),1),1-sin(x(:,end-1:-1:1)*pi/2)];
+        end
+    end
+end
+
+function g = g4(x,t)
+    g = (x-pi/3).^2 + t.*sin(6*pi*(x-pi/3)).^2;
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP7.m b/PlatEMO/Problems/Sparse MOPs/SMOP7.m
new file mode 100644
index 000000000..f987b1920
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP7.m	
@@ -0,0 +1,54 @@
+classdef SMOP7 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP7()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sum(g2(X(:,M:M+K-1),pi/3),2) + sum(g2(X(:,M+K:end),X(:,[M+K+1:end,M+K])*0.9),2);
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),cos(X(:,1:M-1)*pi/2)],2)).*[ones(N,1),sin(X(:,M-1:-1:1)*pi/2)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            P = UniformPoint(N,obj.Global.M);
+            P = P./repmat(sqrt(sum(P.^2,2)),1,obj.Global.M);
+        end
+    end
+end
+
+function g = g2(x,t)
+    g = 2*(x-t).^2 + sin(2*pi*(x-t)).^2;
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/SMOP8.m b/PlatEMO/Problems/Sparse MOPs/SMOP8.m
new file mode 100644
index 000000000..cacc1f221
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/SMOP8.m	
@@ -0,0 +1,54 @@
+classdef SMOP8 < PROBLEM
+% <problem> <Sparse MOP>
+% Benchmark MOP with sparse Pareto optimal solutions
+% theta --- 0.1 --- Sparsity of the Pareto set
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        theta = 0.1;    % Sparsity of the Pareto set
+    end
+    methods
+        %% Initialization
+        function obj = SMOP8()
+            obj.theta = obj.Global.ParameterSet(0.1);
+            if isempty(obj.Global.M)
+                obj.Global.M = 2;
+            end
+            if isempty(obj.Global.D)
+                obj.Global.D = 100;
+            end
+            obj.Global.lower    = [zeros(1,obj.Global.M-1)+0,zeros(1,obj.Global.D-obj.Global.M+1)-1];
+            obj.Global.upper    = [zeros(1,obj.Global.M-1)+1,zeros(1,obj.Global.D-obj.Global.M+1)+2];
+            obj.Global.encoding = 'real';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,X)
+            [N,D] = size(X);
+            M = obj.Global.M;
+            K = ceil(obj.theta*(D-M+1));
+            g = sum(g3(X(:,M:M+K-1),mod(X(:,M+1:M+K)+pi,2)),2) + sum(g3(X(:,M+K:end-1),X(:,M+K+1:end)*0.9),2);
+            PopObj = repmat(1+g/(D-M+1),1,M).*fliplr(cumprod([ones(N,1),cos(X(:,1:M-1)*pi/2)],2)).*[ones(N,1),sin(X(:,M-1:-1:1)*pi/2)];
+        end
+        %% Sample reference points on Pareto front
+        function P = PF(obj,N)
+            P = UniformPoint(N,obj.Global.M);
+            P = P./repmat(sqrt(sum(P.^2,2)),1,obj.Global.M);
+        end
+    end
+end
+
+function g = g3(x,t)
+    g = 4-(x-t)-4./exp(100*(x-t).^2);
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/Sparse_CN.m b/PlatEMO/Problems/Sparse MOPs/Sparse_CN.m
new file mode 100644
index 000000000..e261bfe5d
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/Sparse_CN.m	
@@ -0,0 +1,74 @@
+classdef Sparse_CN < PROBLEM
+% <problem> <Sparse MOP>
+% The critical node detection problem
+% dataNo --- 1 --- Number of dataset
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+% The datasets are taken from the Pajek datasets in
+% http://vlado.fmf.uni-lj.si/pub/networks/data/default.htm
+% No.   Name            Nodes   Edges
+% 1     Movies           102     192
+% 2     GD99             234     154
+% 3     GD01             311     640
+% 4     GD97             452     460
+
+    properties(Access = private)
+        G;  % Adjacency matrix of the graph
+    end
+    methods
+        %% Initialization
+        function obj = Sparse_CN()
+            % Load data
+            dataNo = obj.Global.ParameterSet(1);
+            str    = {'Movies','GD99','GD01','GD97'};
+            CallStack = dbstack('-completenames');
+            load(fullfile(fileparts(CallStack(1).file),'Dataset_CN.mat'),'Dataset');
+            obj.G = Dataset.(str{dataNo});
+            % Parameter setting
+            obj.Global.M        = 2;
+            obj.Global.D        = length(obj.G);
+            obj.Global.encoding = 'binary';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,PopDec)
+            PopDec = logical(PopDec);
+            PopObj = zeros(size(PopDec,1),2);
+            for i = 1 : size(PopDec,1)
+                PopObj(i,1) = mean(PopDec(i,:));
+                PopObj(i,2) = PairwiseConnectivity(obj.G(~PopDec(i,:),~PopDec(i,:)));
+            end
+            PopObj(:,2) = PopObj(:,2)/PairwiseConnectivity(obj.G);
+        end
+    end
+end
+
+function f = PairwiseConnectivity(G)
+    f = 0;
+    remain = true(1,length(G));
+    while any(remain)
+        c = false(1,length(G));
+        c(find(remain,1)) = true;
+        while true
+            c1 = any(G(c,:),1) & remain;
+            if all(c==c1)
+                break;
+            else
+                c = c1;
+            end
+        end
+        remain = remain & ~c;
+        f = f + sum(c)*(sum(c)-1)/2;
+    end
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/Sparse_FS.m b/PlatEMO/Problems/Sparse MOPs/Sparse_FS.m
new file mode 100644
index 000000000..481c9b5d5
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/Sparse_FS.m	
@@ -0,0 +1,81 @@
+classdef Sparse_FS < PROBLEM
+% <problem> <Sparse MOP>
+% The feature selection problem
+% dataNo --- 1 --- Number of dataset
+
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+% The datasets are taken from the UCI machine learning repository in
+% http://archive.ics.uci.edu/ml/index.php
+% No.   Name                              Samples Features Classes
+% 1     Wine                                178      13       3
+% 2     Statlog_Australian                  690      14       2
+% 3     Climate                             540      18       2
+% 4     Parkinsons                          195      22       2
+% 5     Statlog_German                     1000      24       2
+% 6     Breast_cancer_Wisconsin_Diagnostic  569      30       2
+% 7     Ionosphere                          351      34       2
+% 8     SPECTF_Heart                        267      44       2
+% 9     Lung_cancer                          32      56       3
+% 10    Connectionist_bench_Sonar           208      60       2
+% 11    Libras_movement                     360      90      15
+% 12    Hill_Valley                         606     100       2
+% 13    MUSK1                               476     166       2
+% 14    Semeion_handwritten_digit          1593     256      10
+% 15    LSVT_voice_rehabilitation           126     310       2
+% 16    Madelon                            2600     500       2
+% 17    ISOLET                             1557     617      26
+% 18    Multiple_features                  2000     649      10
+% 19    CNAE9                              1080     857       9
+
+    properties(Access = private)
+        TrainIn;    % Input of training set
+        TrainOut;   % Output of training set
+        ValidIn;    % Input of validation set
+        ValidOut;   % Output of validation set
+    end
+    methods
+        %% Initialization
+        function obj = Sparse_FS()
+            % Load data
+            dataNo = obj.Global.ParameterSet(1);
+            str    = {'Wine','Statlog_Australian','Climate','Parkinsons','Statlog_German','Breast_cancer_Wisconsin_Diagnostic',...
+                      'Ionosphere','SPECTF_Heart','Lung_cancer','Connectionist_bench_Sonar','Libras_movement','Hill_Valley',...
+                      'MUSK1','Semeion_handwritten_digit','LSVT_voice_rehabilitation','Madelon','ISOLET','Multiple_features','CNAE9'};
+            CallStack = dbstack('-completenames');
+            load(fullfile(fileparts(CallStack(1).file),'Dataset_FS_NN.mat'),'Dataset');
+            Data = Dataset.(str{dataNo});
+            Fmin = min(Data(:,1:end-1),[],1);
+            Fmax = max(Data(:,1:end-1),[],1);
+            Data(:,1:end-1) = (Data(:,1:end-1)-repmat(Fmin,size(Data,1),1))./repmat(Fmax-Fmin,size(Data,1),1);
+            obj.TrainIn     = Data(1:ceil(end*0.8),1:end-1);
+            obj.TrainOut    = Data(1:ceil(end*0.8),end);
+            obj.ValidIn     = Data(ceil(end*0.8)+1:end,1:end-1);
+            obj.ValidOut    = Data(ceil(end*0.8)+1:end,end);
+            % Parameter setting
+            obj.Global.M        = 2;
+            obj.Global.D        = size(obj.TrainIn,2);
+            obj.Global.encoding = 'binary';
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,PopDec)
+            PopDec   = logical(PopDec);
+            PopObj   = zeros(size(PopDec,1),2);
+            Category = unique(obj.TrainOut);
+            for i = 1 : size(PopObj,1)
+                [~,Rank] = sort(pdist2(obj.ValidIn(:,PopDec(i,:)),obj.TrainIn(:,PopDec(i,:))),2);
+                [~,Out]  = max(hist(obj.TrainOut(Rank(:,1:3))',Category),[],1);
+                Out      = Category(Out);
+            	PopObj(i,1) = mean(PopDec(i,:));
+                PopObj(i,2) = mean(Out~=obj.ValidOut);
+            end
+        end
+    end
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/Sparse_NN.m b/PlatEMO/Problems/Sparse MOPs/Sparse_NN.m
new file mode 100644
index 000000000..ebbfed61c
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/Sparse_NN.m	
@@ -0,0 +1,134 @@
+classdef Sparse_NN < PROBLEM
+% <problem> <Sparse MOP>
+% The neural network training problem (only for bi-category classification)
+% dataNo  ---  1 --- Number of dataset
+% nHidden --- 20 --- Size of hidden layer
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+% The datasets are taken from the UCI machine learning repository in
+% http://archive.ics.uci.edu/ml/index.php
+% No.   Name                              Samples Features Classes
+% 1     Wine                                178      13       3
+% 2     Statlog_Australian                  690      14       2
+% 3     Climate                             540      18       2
+% 4     Parkinsons                          195      22       2
+% 5     Statlog_German                     1000      24       2
+% 6     Breast_cancer_Wisconsin_Diagnostic  569      30       2
+% 7     Ionosphere                          351      34       2
+% 8     SPECTF_Heart                        267      44       2
+% 9     Lung_cancer                          32      56       3
+% 10    Connectionist_bench_Sonar           208      60       2
+% 11    Libras_movement                     360      90      15
+% 12    Hill_Valley                         606     100       2
+% 13    MUSK1                               476     166       2
+% 14    Semeion_handwritten_digit          1593     256      10
+% 15    LSVT_voice_rehabilitation           126     310       2
+% 16    Madelon                            2600     500       2
+% 17    ISOLET                             1557     617      26
+% 18    Multiple_features                  2000     649      10
+% 19    CNAE9                              1080     857       9
+
+    properties(Access = private)
+        TrainIn;    % Input of training set
+        TrainOut;   % Output of training set
+        TestIn;   	% Input of validation set
+        TestOut;  	% Output of validation set
+        nHidden;    % Size of hidden layer
+    end
+    methods
+        function obj = Sparse_NN()
+            % Load data
+            [dataNo,obj.nHidden] = obj.Global.ParameterSet(1,20);
+            str    = {'Wine','Statlog_Australian','Climate','Parkinsons','Statlog_German','Breast_cancer_Wisconsin_Diagnostic',...
+                      'Ionosphere','SPECTF_Heart','Lung_cancer','Connectionist_bench_Sonar','Libras_movement','Hill_Valley',...
+                      'MUSK1','Semeion_handwritten_digit','LSVT_voice_rehabilitation','Madelon','ISOLET','Multiple_features','CNAE9'};
+            CallStack = dbstack('-completenames');
+            load(fullfile(fileparts(CallStack(1).file),'Dataset_FS_NN.mat'),'Dataset');
+            Data = Dataset.(str{dataNo});
+            Mean = mean(Data(:,1:end-1),1);
+            Std  = std(Data(:,1:end-1),[],1);
+            Data(:,1:end-1) = (Data(:,1:end-1)-repmat(Mean,size(Data,1),1))./repmat(Std,size(Data,1),1);
+            Data(:,end)     = Data(:,end) == Data(1,end);	% Only for bi-category classification
+            obj.TrainIn     = Data(1:ceil(end*0.8),1:end-1);
+            obj.TrainOut    = Data(1:ceil(end*0.8),end);
+            obj.TestIn      = Data(ceil(end*0.8)+1:end,1:end-1);
+            obj.TestOut     = Data(ceil(end*0.8)+1:end,end);
+            % Parameter setting
+            obj.Global.M        = 2;
+            obj.Global.D        = (size(obj.TrainIn,2)+1)*obj.nHidden + (obj.nHidden+1)*size(obj.TrainOut,2);
+            obj.Global.lower    = zeros(1,obj.Global.D) - 1;
+            obj.Global.upper    = zeros(1,obj.Global.D) + 1;
+            obj.Global.encoding = 'real';
+        end
+        %% Generate initial population
+        function PopDec = Init(obj,N)
+            PopDec = (rand(N,obj.Global.D)-0.5)*2.*randi([0 1],N,obj.Global.D);
+        end
+        %% Fine-tune solutions
+        function PopDec = CalDec(obj,PopDec)
+            for i = 1 : size(PopDec,1)
+                W1          = reshape(PopDec(i,1:(size(obj.TrainIn,2)+1)*obj.nHidden),size(obj.TrainIn,2)+1,obj.nHidden);
+                W2          = reshape(PopDec(i,(size(obj.TrainIn,2)+1)*obj.nHidden+1:end),obj.nHidden+1,size(obj.TrainOut,2));
+                [W1,W2]     = Train(obj.TrainIn,obj.TrainOut,W1,W2,1);
+                PopDec(i,:) = [W1(:)',W2(:)'];
+            end
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,PopDec)
+            PopObj = zeros(size(PopDec,1),2);
+            for i = 1 : size(PopDec,1)
+                W1 = reshape(PopDec(i,1:(size(obj.TrainIn,2)+1)*obj.nHidden),size(obj.TrainIn,2)+1,obj.nHidden);
+                W2 = reshape(PopDec(i,(size(obj.TrainIn,2)+1)*obj.nHidden+1:end),obj.nHidden+1,size(obj.TrainOut,2));
+                Z  = Predict(obj.TrainIn,W1,W2);
+                PopObj(i,1) = mean(PopDec(i,:)~=0);
+                PopObj(i,2) = mean(round(Z)~=obj.TrainOut);
+            end
+        end
+        %% Draw special figure
+        function Draw(obj,PopDec)
+            PopObj = zeros(size(PopDec,1),2);
+            for i = 1 : size(PopDec,1)
+                W1 = reshape(PopDec(i,1:(size(obj.TestIn,2)+1)*obj.nHidden),size(obj.TestIn,2)+1,obj.nHidden);
+                W2 = reshape(PopDec(i,(size(obj.TestIn,2)+1)*obj.nHidden+1:end),obj.nHidden+1,size(obj.TestOut,2));
+                Z  = Predict(obj.TestIn,W1,W2);
+                PopObj(i,1) = mean(PopDec(i,:)~=0);
+                PopObj(i,2) = mean(round(Z)~=obj.TestOut);
+            end
+            cla; Draw(PopObj);
+            xlabel('Complexity'); ylabel('Test error');
+        end
+    end
+end
+
+function [W1,W2] = Train(X,T,W1,W2,nEpoch)
+    for epoch = 1 : nEpoch
+        [Z,Y] = Predict(X,W1,W2);
+        P     = (Z-T).*Z.*(1-Z);
+        Q     = P*W2(2:end,:)'.*(1-Y.^2);
+        D1    = 0;
+        D2    = 0;
+        for i = 1 : size(X,1)
+            D2 = D2 + [0,Y(i,:)]'*P(i,:);
+            D1 = D1 + [0,X(i,:)]'*Q(i,:);
+        end
+        W1 = W1 - D1/size(X,1);
+        W2 = W2 - D2/size(X,1);
+    end
+end
+
+function [Z,Y] = Predict(X,W1,W2)
+    Y = 1 - 2./(1+exp(2*[zeros(size(X,1),1),X]*W1));
+    Z = 1./(1+exp(-[zeros(size(Y,1),1),Y]*W2));
+end
\ No newline at end of file
diff --git a/PlatEMO/Problems/Sparse MOPs/Sparse_PM.m b/PlatEMO/Problems/Sparse MOPs/Sparse_PM.m
new file mode 100644
index 000000000..7304efac6
--- /dev/null
+++ b/PlatEMO/Problems/Sparse MOPs/Sparse_PM.m	
@@ -0,0 +1,85 @@
+classdef Sparse_PM < PROBLEM
+% <problem> <Sparse MOP>
+% The pattern mining problem
+% numTran --- 10000 --- Number of transactions
+% lenTran ---    50 --- Average length of transactions
+% numPa   ---   100 --- Number of patterns
+% lenPa   ---     5 --- Average length of patterns
+% numItem ---   100 --- Number of items (decision variables)
+
+%------------------------------- Reference --------------------------------
+% Y. Tian, X. Zhang, C. Wang, and Y. Jin, An evolutionary algorithm for
+% large-scale sparse multi-objective optimization problems, IEEE
+% Transactions on Evolutionary Computation, 2019.
+%------------------------------- Copyright --------------------------------
+% Copyright (c) 2018-2019 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".
+%--------------------------------------------------------------------------
+
+    properties(Access = private)
+        Data;   % Transaction dataset
+    end
+    methods
+        %% Initialization
+        function obj = Sparse_PM()
+            % Parameter setting
+            [numTran,lenTran,numPa,lenPa,numItem] = obj.Global.ParameterSet(10000,50,100,5,100);
+            obj.Global.M        = 2;
+            obj.Global.D        = numItem;
+            obj.Global.encoding = 'binary';
+            % Randomly generate transaction dataset
+            fileName = fullfile(fileparts(mfilename('fullpath')),sprintf('Dataset_PM-D%d-T%d-L%d-I%d-N%d.mat',numTran,lenTran,numPa,lenPa,numItem));
+            if exist(fileName,'file') == 2
+                load(fileName,'Data');
+            else
+                % Generate patterns
+                PaSet = cell(1,numPa);
+                RAND1 = min(numItem,max(1,random('Poisson',lenPa,1,numPa)));    % Length of each pattern
+                RAND2 = min(1,max(0,random('Exponential',0.5,1,numPa)));	% Fraction of items in each pattern chosen from previous pattern
+                PaSet{1} = randperm(numItem,RAND1(1));
+                for i = 2 : numPa
+                    PaSet{i} = [PaSet{i-1}(randperm(end,min(end,round(RAND2(i)*RAND1(i))))),...
+                                randperm(numItem,round((1-RAND2(i))*RAND1(i)))];
+                end
+                % Generate transactions
+                RAND3 = max(0,random('Exponential',1,1,numPa));	% Probability that each pattern will be picked
+                RAND3 = cumsum(RAND3)./sum(RAND3);
+                RAND4 = random('Normal',0.5,0.1,1,numPa);	% Corruption level of each pattern
+                Data  = false(numTran,numItem);
+                RAND5 = min(min(numItem,length(unique(cat(2,PaSet{:})))),max(1,random('Poisson',lenTran,1,numTran)));	% Length of each transaction
+                Data = false(numTran,numItem);
+                for i = 1 : numTran
+                    while sum(Data(i,:)) < RAND5(i)
+                        current = find(rand<=RAND3,1);
+                        Data(i,PaSet{current}(rand(1,end)>RAND4(current))) = true;
+                    end
+                end
+                save(fileName,'Data');
+            end
+            obj.Data = Data;
+        end
+        %% Generate initial population
+        function PopDec = Init(obj,N)
+            PopDec = rand(N,obj.Global.D) < 20/obj.Global.D;
+        end
+        %% Calculate objective values
+        function PopObj = CalObj(obj,PopDec)
+            PopDec = logical(PopDec);
+            PopObj = zeros(size(PopDec,1),2);
+            for i = 1 : size(PopDec,1)
+                x  = PopDec(i,:);
+                Tx = all(obj.Data(:,x),2);
+                if ~any(Tx)
+                    PopObj(i,:) = 1;
+                else
+                    PopObj(i,1) = 1 - mean(Tx);
+                    PopObj(i,2) = 1 - mean(sum(x)./sum(obj.Data(Tx,:),2));
+                end
+            end
+        end
+    end
+end
\ No newline at end of file
diff --git a/PlatEMO/Public/Draw.m b/PlatEMO/Public/Draw.m
index 6334b1667..f67104bdb 100644
--- a/PlatEMO/Public/Draw.m
+++ b/PlatEMO/Public/Draw.m
@@ -50,12 +50,12 @@ function Draw(Data,varargin)
     set(gca,'NextPlot','add','Box','on','Fontname','Times New Roman','FontSize',Size(4));
     if M == 2
         plot(Data(:,1),Data(:,2),varargin{:});
-        xlabel('\itf\rm_1'); ylabel('\itf\rm_2');
+        xlabel('\it f\rm_1'); ylabel('\it f\rm_2');
         set(gca,'XTickMode','auto','YTickMode','auto','View',[0 90]);
         axis tight;
     elseif M == 3
         plot3(Data(:,1),Data(:,2),Data(:,3),varargin{:});
-        xlabel('\itf\rm_1'); ylabel('\itf\rm_2'); zlabel('\itf\rm_3');
+        xlabel('\it f\rm_1'); ylabel('\it f\rm_2'); zlabel('\it f\rm_3');
         set(gca,'XTickMode','auto','YTickMode','auto','ZTickMode','auto','View',[135 30]);
         axis tight;
     elseif M > 3