Merge branch 'release/v2.2'

README.md

@@ -110,10 +110,12 @@ If you use this toolbox in your research please cite ["A Comparative Study of Ke
 
     @inproceedings{vanvaerenbergh2013comparative,
       author = {Van Vaerenbergh, Steven and Santamar{\'i}a, Ignacio},
-      booktitle = {2013 IEEE Digital Signal Processing (DSP) Workshop and IEEE Signal Processing Education (SPE)},
+      booktitle = {2013 IEEE Digital Signal Processing and Signal Processing Education Meeting (DSP/SPE)},
       title = {A Comparative Study of Kernel Adaptive Filtering Algorithms},
       doi = {10.1109/DSP-SPE.2013.6642587},
+      pages = {181--186},
       year = {2013},
+      month = aug,
       note = {Software available at \url{https://github.com/steven2358/kafbox/}}
     }
 

demo/demo_parameter_estimation_lorenz.m

@@ -37,7 +37,7 @@
 kaf = krlst(struct('lambda',lambda_est,'M',100,'sn2',reg_est,...
     'kerneltype','gauss','kernelpar',sigma_est));
 
-for i=1:N,
+for i=1:N
     if ~mod(i,floor(N/10)), fprintf('.'); end % progress indicator, 10 dots
 
     Y_est(i) = kaf.evaluate(X(i,:)); % predict the next output

demo/demo_prediction.m

@@ -12,7 +12,7 @@
 %% RUN ALGORITHM
 N = size(X,1);
 Y_est = zeros(N,1);
-for i=1:N,
+for i=1:N
     if ~mod(i,floor(N/10)), fprintf('.'); end % progress indicator, 10 dots
     Y_est(i) = kaf.evaluate(X(i,:)); % predict the next output
     kaf.train(X(i,:),Y(i)); % train with one input-output pair

demo/demo_prediction_kstep.m

@@ -18,7 +18,7 @@
 %% RUN ALGORITHM
 N = size(X,1);
 Y_est = zeros(N,1);
-for i=k:N,
+for i=k:N
     if ~mod(i,floor(N/10)), fprintf('.'); end % progress indicator, 10 dots
     Y_est(i) = kaf.evaluate(X(i,:)); % predict the next output
 

demo/demo_prediction_kstep_split.m

@@ -34,7 +34,7 @@
 Y_test = Y(N_train+1:N_train+N_test);
 
 Y_test_MSE = zeros(N_train,1);
-for i=1:N_train,
+for i=1:N_train
     % train on data set 1
     if ~mod(i,floor(N_train/10)), fprintf('.'); end % progress indicator, 10 dots
     kaf.train(X(i,:),Y(i)); % train with one input-output pair

demo/demo_prediction_mackey_glass.m

@@ -40,11 +40,11 @@
 MSE = zeros(n_train,n_algos);
 y_est_final = zeros(n_test,n_algos);
 titles = cell(n_algos,1);
-for j=1:n_algos,
+for j=1:n_algos
     kaf = algos{j};
     titles{j} = upper(class(kaf)); % store algorithm name
     fprintf('Training %s',titles{j})
-    for i=1:n_train,
+    for i=1:n_train
         if ~mod(i,floor(n_train/10)), fprintf('.'); end % progress indicator
 
         kaf.train(X_train(i,:),y_train(i)); % train with one input-output pair
@@ -69,7 +69,7 @@
 plot(y_test,'LineWidth',2);
 titles2 = {'original'};
 line_styles = {'--','-.'};
-for j=1:n_algos,
+for j=1:n_algos
     plot(y_est_final(:,j),'LineWidth',2,'LineStyle',line_styles{j});
 end
 titles2(2:n_algos+1) = titles;

demo/demo_profiler_prediction_lorenz.m

@@ -68,7 +68,7 @@
 %     output_dir = output_dir_default;
 %     fprintf('Using default folder "%s" for storing results.\n', output_dir_default);
 % end
-output_dir = fullfile(mfilename('fullpath'),'..','results');
+output_dir = fullfile(fileparts(mfilename('fullpath')),'results');
 
 t1 = tic;
 [data,algorithms,results] = kafbox_profiler(data,sim_opts,algorithms,output_dir);

demo/demo_reconverge_all.m

@@ -19,7 +19,7 @@
 %% PROGRAM
 
 w = who;
-for i=1:length(w), % copy all parameters to option structure
+for i=1:length(w) % copy all parameters to option structure
     eval(sprintf('opt.%s = %s;',w{i},w{i}))
 end
 
@@ -30,8 +30,8 @@
 fdir = fileparts(which('kafbox_template.m'));
 files = dir(fullfile(fdir,'*.m'));
 [~,algorithms] = cellfun(@fileparts, {files.name}, 'UniformOutput',0);
-for i=length(algorithms):-1:1,
-    if ~exist(algorithms{i},'class'),
+for i=length(algorithms):-1:1
+    if ~exist(algorithms{i},'class')
         algorithms(i) = []; % remove files that do not represent classes
     end
 end
@@ -49,7 +49,7 @@
     titles{algo_ind} = strrep(upper(algorithm),'_','\_');
 
     kaf = feval(algorithm);
-    for i=1:N,
+    for i=1:N
         if ~mod(i,floor(N/10)), fprintf('.'); end
 
         y_est = kaf.evaluate(X_test);
@@ -81,7 +81,7 @@
     num_in_league = (rm==0)*5 + rm;
     % plot the results for the num_in_league worst results
     league_inds = num_alg-remaining+num_in_league:-1:num_alg-remaining+1;
-    for i=league_inds,
+    for i=league_inds
         plot(10*log10(MSE(:,ind(i))),'LineWidth',1)
     end
     title(sprintf('League %d',fig_ind))

demo/demo_regression_2d.m

@@ -13,7 +13,7 @@
 
 %% PROGRAM
 
-kaf = feval(algorithm, opts);
+kaf = feval(algorithm, opts); %#ok<FVAL>
 
 % generate some data
 c = 5;
@@ -22,7 +22,7 @@
 y = sin(3*x(:,1)).*cos(x(:,1)+x(:,2));
 
 fprintf('Training')
-for i=1:N,
+for i=1:N
     if ~mod(i,floor(N/10)), fprintf('.'); end
     kaf.train(x(i,:),y(i));
     % y_test = kaf.evaluate(x(i+1,:));

demo/demo_sample_run.m

@@ -47,7 +47,7 @@
 
 fprintf('Running KRLS-T\n');
 t1 = tic;
-for i=1:n,
+for i=1:n
     if ~mod(i,floor(n/10)), fprintf('.'); end
     Qold = kaf.Q;
     kaf.train(x(i),y(i));
@@ -108,7 +108,7 @@
         end
 
         if savefig
-            if ~exist('fig','file')
+            if ~exist('fig','file') %#ok<UNRCH>
                 mkdir('fig');
             end
             % write individual png

demo/demo_sinc.m

@@ -25,9 +25,9 @@
 %% RUN ALGORITHM
 fprintf('%s: ',upper(algorithm));
 Y_est = zeros(N_test,1);
-kaf = feval(algorithm);
+kaf = feval(algorithm); %#ok<FVAL>
 t1 = tic;
-for i=1:N,
+for i=1:N
     if ~mod(i,floor(N/10)), fprintf('.'); end
     kaf.train(x(i),y(i));
 end

demo/demo_sinc_all.m

@@ -27,8 +27,8 @@
 fdir = fileparts(which('kafbox_template.m'));
 files = dir(fullfile(fdir,'*.m'));
 [~,algorithms] = cellfun(@fileparts, {files.name}, 'UniformOutput',0);
-for i=length(algorithms):-1:1,
-    if ~exist(algorithms{i},'class'),
+for i=length(algorithms):-1:1
+    if ~exist(algorithms{i},'class')
         algorithms(i) = []; % remove files that do not represent classes
     end
 end
@@ -46,7 +46,7 @@
     titles{algo_ind} = strrep(upper(algorithm),'_','\_');
 
     kaf = feval(algorithm);
-    for i=1:N,
+    for i=1:N
         if ~mod(i,floor(N/10)), fprintf('.'); end
         kaf.train(x(i),y(i));
     end
@@ -73,7 +73,7 @@
     num_in_league = (rm==0)*5 + rm;
     % plot the results for the num_in_league worst results
     league_inds = num_alg-remaining+num_in_league:-1:num_alg-remaining+1;
-    for i=league_inds,
+    for i=league_inds
         plot(x_test,Y_est(:,ind(i)),'LineWidth',2)
     end
     title(sprintf('League %d',fig_ind))

demo/demo_template_prediction.m

@@ -69,7 +69,7 @@
 
 fprintf('Running prediction algorithm %s',upper(class(kaf)))
 Y_est = zeros(N,1);
-for i=1:N,
+for i=1:N
     if ~mod(i,floor(N/10)), fprintf('.'); end
 
     Y_est(i) = kaf.evaluate(X(i,:)); % make prediction

demo/demo_template_system_identification.m

@@ -26,7 +26,7 @@
 
 fprintf('Running system identification algorithm')
 Y_est = zeros(N,1);
-for i=1:N,
+for i=1:N
     if ~mod(i,floor(N/10)), fprintf('.'); end
 
     Y_est(i) = kaf.evaluate(X(i,:)); % make prediction

demo/krlst_split.m

@@ -39,8 +39,8 @@
 
         function kaf = krlst_split(parameters) % constructor
             if (nargin > 0) % copy valid parameters
-                for fn = fieldnames(parameters)',
-                    if ismember(fn,fieldnames(kaf)),
+                for fn = fieldnames(parameters)'
+                    if ismember(fn,fieldnames(kaf))
                         kaf.(fn{1}) = parameters.(fn{1});
                     end
                 end
@@ -124,7 +124,7 @@ function train(kaf,x,y) % train the algorithm
         function kaf = prune(kaf,Qold,gamma2)
             m = size(kaf.Sigma,1);
 
-            if gamma2<kaf.jitter, % to avoid roundoff error
+            if gamma2<kaf.jitter % to avoid roundoff error
                 if gamma2<kaf.jitter/10
                     warning('Numerical roundoff error too high, you should increase jitter noise') %#ok<WNTAG>
                 end
@@ -136,7 +136,7 @@ function train(kaf,x,y) % train the algorithm
             [~, r] = min(criterion); % remove element r, which incurs in the minimum error
             smaller = 1:m; smaller(r) = [];
 
-            if r == m, % if we must remove the element we just added, perform reduced update instead
+            if r == m % if we must remove the element we just added, perform reduced update instead
                 kaf.Q = Qold;
                 kaf.reduced = true;
             else

demo/literature/liu2010kernel/fig2_12.m

@@ -20,7 +20,7 @@
 N_tr = 1000; % number of training data
 N_te = 50; % number of testing data
 
-setups{1} = lms(struct('mu',.01));
+setups{1} = lms(struct('mu',.01)); %#ok<CMLMS>
 setups{2} = klms(struct('eta',0.2,'M',N_tr,'kerneltype','gauss','kernelpar',1));
 
 %% PREPARE DATA
@@ -36,7 +36,7 @@
 y = z - 0.9*z.^2 + ns; % output of nonlinear channel
 
 X_all = zeros(N_tr+N_te,embedding); % time-embedding
-for k=1:embedding,
+for k=1:embedding
     X_all(:,k) = y(k:N_tr+N_te+k-1);
 end
 
@@ -52,7 +52,7 @@
 
 MSE = zeros(N_tr,num_setup);
 
-for setup_ind = 1:num_setup,
+for setup_ind = 1:num_setup
     kaf = setups{setup_ind};
 
     for n=1:N_tr

demo/literature/richard2009online/fig1.m

@@ -35,13 +35,13 @@
 MSE = zeros(N,num_setup);
 titles = cell(num_setup,1);
 
-for sim_ind = 1:numsim,
+for sim_ind = 1:numsim
     fprintf('SIM %d:\n',sim_ind)
 
     % Generate the data
     [X,y,yref] = generate_doddbench(N);
 
-    for setup_ind = 1:num_setup,
+    for setup_ind = 1:num_setup
         kaf = setups{setup_ind};
         titles{setup_ind} = upper(class(kaf));
 

demo/literature/richard2009online/fig2.m

@@ -37,13 +37,13 @@
 MSE = zeros(N,num_setup);
 titles = cell(num_setup,1);
 
-for sim_ind = 1:numsim,
+for sim_ind = 1:numsim
     fprintf('SIM %d:\n',sim_ind)
 
     % Generate the data
     [X,y,yref] = generate_richardbench(N);
 
-    for setup_ind = 1:num_setup,
+    for setup_ind = 1:num_setup
         kaf = setups{setup_ind};
         titles{setup_ind} = upper(class(kaf));
 

demo/literature/richard2009online/generate_doddbench.m

@@ -20,7 +20,7 @@
 dref = zeros(1,N+2);
 dref(1:2)=[0.1 0.1];
 
-for t=3:N+2,
+for t=3:N+2
     dref(t) = (0.8-0.5*exp(-dref(t-1)^2))*dref(t-1) - ...
         (0.3+0.9*exp(-dref(t-1)^2))*dref(t-2)+0.1*sin(pi*dref(t-1));
 end

demo/literature/richard2009online/generate_richardbench.m

@@ -21,7 +21,7 @@
 u = 0.25*randn(N+1,1);
 dref = zeros(1,N+1);
 
-for t=2:N+1,
+for t=2:N+1
     v(t) = 1.1*exp(-abs(v(t-1)))+u(t);
     dref(t) = v(t)^2;
 end

demo/literature/vanvaerenbergh2006sliding/fig2.m

@@ -41,7 +41,7 @@
 num_alg = length(setups);
 MSE = zeros(N,num_alg);
 MSE_final = zeros(1,num_alg);
-for sim_ind = 1:num_sim,
+for sim_ind = 1:num_sim
     % generate data
     [X,y,y_ref,X_test,y_test] = generate_channel_switch(opt);
 
@@ -54,7 +54,7 @@
         fprintf('%9s M=%3d: ',upper(class(kaf)),kaf.M);
 
         mse = zeros(N,1);
-        for i=1:N,
+        for i=1:N
             if ~mod(i,floor(N/10)), fprintf('.'); end
 
             y_est = kaf.evaluate(X_test);

demo/literature/vanvaerenbergh2012kernel/fig3.m

@@ -41,7 +41,7 @@
 % generate Gaussian input data s
 s = rand(N+N_test,1);
 s_mem = zeros(N+N_test,embedding);
-for i = 1:embedding,
+for i = 1:embedding
     s_mem(i:N+N_test,i) = s(1:N+N_test-i+1);	% time-embedding
 end
 s = s_mem(1:N+N_test,:);	% input data, stored in columns
@@ -78,9 +78,9 @@
 
     titles{setup_ind} = upper(class(kaf));
     fprintf('%s\t',titles{setup_ind});
-    for n=1:N,
+    for n=1:N
         if ~mod(n,round(N/10)), fprintf('.'); end
-        if n<=Nswitch,
+        if n<=Nswitch
             Y_test = Y_test1;
         else
             Y_test = Y_test2;

demo/literature/yukawa2012multikernel/fig1a.m

@@ -28,7 +28,7 @@
 % mk_kernelpars = [.2 .5 1 2 5 10 20 50]; % alternative values
 
 % remark 1 from paper determines mu0
-setups{1} = lms(struct('mu',.01));
+setups{1} = lms(struct('mu',.01)); %#ok<CMLMS>
 setups{2} = knlms(struct('mu0',mk_thresh^(.2/min(mk_kernelpars)),'eta',0.2,'eps',3E-2,'kerneltype','gauss','kernelpar',1/sqrt(.2)));
 setups{3} = knlms(struct('mu0',mk_thresh^(.5/min(mk_kernelpars)),'eta',0.2,'eps',3E-2,'kerneltype','gauss','kernelpar',1/sqrt(.5)));
 setups{4} = mknlms_cs(struct('delta',mk_thresh,'eta',0.2,'rho',6E-2,'kerneltype','gauss','kernelpars',1./sqrt(mk_kernelpars)));
@@ -47,7 +47,7 @@
 y = z - 0.9*z.^2 + ns; % output of nonlinear channel
 
 X_all = zeros(N_tr+N_te,embedding); % time-embedding
-for k=1:embedding,
+for k=1:embedding
     X_all(:,k) = y(k:N_tr+N_te+k-1);
 end
 
@@ -63,7 +63,7 @@
 num_setup = length(setups);
 MSE = zeros(N_tr,num_setup);
 
-for setup_ind = 1:num_setup,
+for setup_ind = 1:num_setup
     kaf = setups{setup_ind};
 
     for n=1:N_tr

demo/run_all_demos.m

@@ -17,7 +17,7 @@
 
 % get list of literature scripts
 folders_literature = dir(fullfile(fdir,'literature','*2*'));
-for i=1:length(folders_literature);
+for i=1:length(folders_literature)
     folder_i = fullfile(fdir,'literature',folders_literature(i).name);
     files_i = dir(fullfile(folder_i,'fig*.m'));
     folders_i = repmat({folder_i},length(files_i),1);