From a78542352f1dbf9d4f4e6cc5316f09e79d527b75 Mon Sep 17 00:00:00 2001
From: mattrosenblatt7 <52167418+mattrosenblatt7@users.noreply.github.com>
Date: Sun, 7 Feb 2021 21:57:06 -0500
Subject: [PATCH] update cpm_cv.m

adjusted cross-validation indices to avoid completely leaving some subject out
---
 matlab/func/cpm_cv.m | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)

diff --git a/matlab/func/cpm_cv.m b/matlab/func/cpm_cv.m
index 3bb91d8..bd8d7fe 100644
--- a/matlab/func/cpm_cv.m
+++ b/matlab/func/cpm_cv.m
@@ -7,10 +7,16 @@
 % y_test       y data used for testing
 % y_predict    Predictions of y data used for testing
 
-% Split data
+% cross-validation indices
 nsubs=size(x,2);
-randinds=randperm(nsubs);
-ksample=floor(nsubs/kfolds);
+all_ind=[];
+fold_size=floor(nsubs/kfolds);
+for idx=1:kfolds
+    all_ind=[all_ind; idx*ones(fold_size, 1)];
+end
+leftover=mod(nsubs, kfolds);
+indices=[all_ind; randperm(kfolds, leftover)'];
+indices=indices(randperm(length(indices)));
 
 y_predict = zeros(nsubs, 1);
 % Run CPM over all folds
@@ -18,16 +24,9 @@
 for leftout = 1:kfolds
     fprintf('%1.0f ',leftout);
     
-    if kfolds == nsubs % doing leave-one-out
-        testinds=randinds(leftout);
-        traininds=setdiff(randinds,testinds);
-    else
-        si=1+((leftout-1)*ksample);
-        fi=si+ksample-1;
-        
-        testinds=randinds(si:fi);
-        traininds=setdiff(randinds,testinds);
-    end
+    testinds=(indices==leftout);
+    traininds=(indices~=leftout);
+    
     
     % Assign x and y data to train and test groups 
     x_train = x(:,traininds);