updated niche/functional

fastlib/u/niche/functional/test_genes_postica.m

@@ -7,48 +7,13 @@ load gene_results;
 g1_indices = find(phases == g1_phase);
 nong1_indices = find(phases ~=g1_phase & phases ~= unknown_phase);
 
-used_scores = pc_scores;
+used_scores = ic_scores;
 
 used_scores = used_scores([1],:);
 
 svm_data = [used_scores(:,g1_indices) used_scores(:,nong1_indices)]';
 svm_labels = [1 * ones(length(g1_indices),1);
 	      -1 * ones(length(nong1_indices),1)];
-%{
-correct = 0;
-same_p = 0;
-
-for i = 1:size(svm_data, 1)
-  train_data = svm_data([1:(i-1) (i+1):end]);
-  train_labels = svm_labels([1:(i-1) (i+1):end]);
-  
-  g1_mu = mean(train_data(find(train_labels == 1)));
-  g1_sigma = std(train_data(find(train_labels == 1)));
-  nong1_mu = mean(train_data(find(train_labels == -1)));
-  nong1_sigma = std(train_data(find(train_labels == -1)));
-  
-  p_g1 = normpdf(svm_data(i), g1_mu, g1_sigma);
-  p_nong1 = normpdf(svm_data(i), nong1_mu, nong1_sigma);
-  
-  if svm_labels(i) == 1
-    if p_g1 > p_nong1
-      correct = correct + 1;
-    end
-  else
-    if p_nong1 > p_g1
-      correct = correct + 1;
-    end
-  end
-  
-  if p_g1 == p_nong1
-    same_p = same_p + 1;
-  end
-  
-end
-
-correct / size(svm_data,1)
-%}
-
 
 
 
@@ -58,27 +23,22 @@ svmlwrite('funknet', svm_data, svm_labels);
 % set initial svm options %
 svm_options = ...
     svmlopt('Kernel', 2, 'KernelParam', 3, 'C', .2, 'ComputeLOO', 1, ...
-	    'ExecPath','/home/niche/matlab/toolboxes/svml');
+	    'ExecPath','../../../../matlab/toolboxes/svml');
 
 
 % let's do a retarded random sampling of sigma and C on a grid!
 
 
-num_sigma_epochs = 75;
-%100;
-sigma_init = 1e-3;
-%1e-4;
+num_sigma_epochs = 100;
+sigma_init = 1e-4;
 sigma_grow = 1.1;
 %num_order_epochs = 4;
 %order_init = 1;
 %order_inc = 1;
 
-num_C_epochs = 3;
-%81;
-C_init = 30;
-%1e-2;
-C_grow = 1 + 1/3;
-%1.1;
+num_C_epochs = 81;
+C_init = 1e-2;
+C_grow = 1.1;
 
 
 
@@ -142,3 +102,55 @@ end
 
 % find a way to do fPCA/fICA using multiple sets of curves
 
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+%{
+correct = 0;
+same_p = 0;
+
+for i = 1:size(svm_data, 1)
+  train_data = svm_data([1:(i-1) (i+1):end]);
+  train_labels = svm_labels([1:(i-1) (i+1):end]);
+  
+  g1_mu = mean(train_data(find(train_labels == 1)));
+  g1_sigma = std(train_data(find(train_labels == 1)));
+  nong1_mu = mean(train_data(find(train_labels == -1)));
+  nong1_sigma = std(train_data(find(train_labels == -1)));
+  
+  p_g1 = normpdf(svm_data(i), g1_mu, g1_sigma);
+  p_nong1 = normpdf(svm_data(i), nong1_mu, nong1_sigma);
+  
+  if svm_labels(i) == 1
+    if p_g1 > p_nong1
+      correct = correct + 1;
+    end
+  else
+    if p_nong1 > p_g1
+      correct = correct + 1;
+    end
+  end
+  
+  if p_g1 == p_nong1
+    same_p = same_p + 1;
+  end
+  
+end
+
+correct / size(svm_data,1)
+%}