diff --git a/matlab/meet/evaluate/evalcv.m b/matlab/meet/evaluate/evalcv.m
deleted file mode 100644
index c6ffff2..0000000
--- a/matlab/meet/evaluate/evalcv.m
+++ /dev/null
@@ -1,37 +0,0 @@
-function cvstat = evalcv(stat)
-nfold = length(stat);
-train = zeros(2, nfold);
-validate = zeros(2, nfold);
-if isfield(stat, 'test')
-  test = zeros(2, 1);
-end
-
-for i = 1 : nfold
-  train(:, i) =  stat{i}.train.error;
-  validate(:, i) = stat{i}.validate.error;
-  if isfield(stat, 'test')
-    test(:, i) = stat{i}.test.error;
-  end
-end
-
-cvstat.train.meanerror = mean(train, 2);
-disp('training error mean:');
-disp(cvstat.train.meanerror);
-
-cvstat.train.stderror = std(train, 0, 2);
-disp('training error std:');
-disp(cvstat.train.stderror);
-
-cvstat.validate.meanerror = mean(validate, 2);
-disp('validation error mean:');
-disp(cvstat.validate.meanerror);
-
-cvstat.validate.stderror = std(validate, 0, 2);
-disp('validation error std:');
-disp(cvstat.validate.stderror);
-
-if isfield(stat, 'test')
-  cvstat.test.meanerror = mean(test, 2);
-  cvstat.test.stderror = std(test, 0, 2);
-end
-end
diff --git a/matlab/meet/learn/ahmm/checkahmm.m b/matlab/meet/learn/ahmm/checkahmm.m
new file mode 100644
index 0000000..23720b7
--- /dev/null
+++ b/matlab/meet/learn/ahmm/checkahmm.m
@@ -0,0 +1,4 @@
+function checkahmm(ahmm)
+learned_GCPT = CPD_to_CPT(ahmm.CPD{5});
+learned_GCPT(:, 1, :);
+end
\ No newline at end of file
diff --git a/matlab/meet/learn/ahmm/inferenceahmm.m b/matlab/meet/learn/ahmm/inferenceahmm.m
index f9217e9..46bf01e 100644
--- a/matlab/meet/learn/ahmm/inferenceahmm.m
+++ b/matlab/meet/learn/ahmm/inferenceahmm.m
@@ -7,6 +7,8 @@
 for i = 1 : nseq
   evidence = data{i};
   engine = enter_evidence(engine, evidence);
-  R{i} = mapest(engine, predictNode, length(evidence));
+  %R{i} = mapest(engine, predictNode, length(evidence));
+  mpe = find_mpe(engine, evidence);
+  R{i} = mpe(predictNode, :);
 end
 end
diff --git a/matlab/meet/learn/ahmm/learnahmm.m b/matlab/meet/learn/ahmm/learnahmm.m
index cee7b94..624f41d 100644
--- a/matlab/meet/learn/ahmm/learnahmm.m
+++ b/matlab/meet/learn/ahmm/learnahmm.m
@@ -21,6 +21,7 @@
 predictNode = [param.G1 param.F1];
 param.onodes = [param.X1];
 finalAhmm = sethiddenbit(finalAhmm, param.onodes);
+checkahmm(finalAhmm);
 
 trainData = createInputData(Y.train, X.train, param);
 R.train = inferenceahmm(finalAhmm, trainData, predictNode);
diff --git a/matlab/meet/learn/ahmm/mapest.m b/matlab/meet/learn/ahmm/mapest.m
index fcea80a..8894e53 100644
--- a/matlab/meet/learn/ahmm/mapest.m
+++ b/matlab/meet/learn/ahmm/mapest.m
@@ -1,4 +1,6 @@
 function mapEst = mapest(engine, hnode, T)
+% Args:
+% - hnode: hidden node that we want to estimate the state.
 % Return
 % - mapEst: a cell array.
 nhnode = length(hnode);
diff --git a/matlab/test/TestAHMM.m b/matlab/test/TestAHMM.m
index 1a55f77..54a4449 100644
--- a/matlab/test/TestAHMM.m
+++ b/matlab/test/TestAHMM.m
@@ -150,8 +150,8 @@ function testInference(self)
   function testLearning(self)
     T = 20;
     max_iter = 10;
-    true_params = self.deterministicParams;
-    ahmm = createahmm(true_params);
+    trueParam = self.deterministicParams;
+    ahmm = createahmm(trueParam);
     ev = sample_dbn(ahmm, 'length', T);
     ss = length(ahmm.intra);
     evidence = cell(1, 1);
@@ -173,19 +173,19 @@ function testLearning(self)
     assertTrue(all(cell2mat(mapS(:)) == trueS(:)));
     
     learned_Gstartprob = CPD_to_CPT(final_ahmm.CPD{1});
-    assertTrue(all(learned_Gstartprob == true_params.Gstartprob(:)));
+    assertTrue(all(learned_Gstartprob == trueParam.Gstartprob(:)));
     learned_Sstartprob = CPD_to_CPT(final_ahmm.CPD{2});
     assertTrue(learned_Sstartprob(1, 1) == 1);
     learned_Stermprob = CPD_to_CPT(final_ahmm.CPD{3});
     assertTrue(all(learned_Stermprob(:, 1) == 0));
     learned_hand = struct(final_ahmm.CPD{4}).hand;
-    assertTrue(all(learned_hand(:) == true_params.hand(:)));
+    assertTrue(all(learned_hand(:) == trueParam.hand(:)));
     learned_GCPT = CPD_to_CPT(final_ahmm.CPD{5});
     learned_GCPT = learned_GCPT(:, 1, :);
     expected = eye(new_params.nG, new_params.nG);
     assertTrue(all(learned_GCPT(:) == expected(:)));
     learned_Gtransprob = struct(final_ahmm.CPD{5}).transprob;
-    assertTrue(all(learned_Gtransprob(:) == true_params.Gtransprob(:)));
+    assertTrue(all(learned_Gtransprob(:) == trueParam.Gtransprob(:)));
     learned_Stransprob = CPD_to_CPT(final_ahmm.CPD{6});
     assertTrue(learned_Stransprob(1, 2, 2) == 1);
     assertTrue(learned_Stransprob(2, 3, 3) == 1);
diff --git a/matlab/test/TestEvalClassification.m b/matlab/test/TestEvalClassification.m
index 3a991c7..7cd64e6 100644
--- a/matlab/test/TestEvalClassification.m
+++ b/matlab/test/TestEvalClassification.m
@@ -11,9 +11,11 @@ function testEvalOneFold(self) %#ok<MANU>
     R.train = {{1 2; 3 4}};
     R.validate = {{5 6; 7 8}};
     stat = evalclassification(Y, R, @zerooneloss);
-    assertTrue(length(stat.train.error) == 2);
-    assertTrue(all(stat.train.error == 0));
-    assertTrue(all(stat.validate.error == 0));
+    train = stat('train');
+    validate = stat('validate');
+    assertTrue(length(train('error')) == 2);
+    assertTrue(all(train('error') == 0));
+    assertTrue(all(validate('error') == 0));
   end
 end
 end
\ No newline at end of file