New private method for factoring out some common code.

src/main/java/org/apache/commons/math4/ml/neuralnet/sofm/KohonenUpdateAction.java

@@ -145,6 +145,26 @@ public class KohonenUpdateAction implements UpdateAction {
         return numberOfCalls.get();
     }
 
+    /**
+     * Tries to update a neuron.
+     *
+     * @param n Neuron to be updated.
+     * @param features Training data.
+     * @param learningRate Learning factor.
+     * @return {@code true} if the update succeeded, {@code true} if a
+     * concurrent update has been detected.
+     */
+    private boolean attemptNeuronUpdate(Neuron n,
+                                        double[] features,
+                                        double learningRate) {
+        final double[] expect = n.getFeatures();
+        final double[] update = computeFeatures(expect,
+                                                features,
+                                                learningRate);
+
+        return n.compareAndSetFeatures(expect, update);
+    }
+
     /**
      * Atomically updates the given neuron.
      *
@@ -156,11 +176,7 @@ public class KohonenUpdateAction implements UpdateAction {
                                           double[] features,
                                           double learningRate) {
         while (true) {
-            final double[] expect = n.getFeatures();
+            if (attemptNeuronUpdate(n, features, learningRate)) {
-            final double[] update = computeFeatures(expect,
-                                                    features,
-                                                    learningRate);
-            if (n.compareAndSetFeatures(expect, update)) {
                 break;
             }
         }
@@ -181,11 +197,7 @@ public class KohonenUpdateAction implements UpdateAction {
         while (true) {
             final Neuron best = MapUtils.findBest(features, net, distance);
 
-            final double[] expect = best.getFeatures();
+            if (attemptNeuronUpdate(best, features, learningRate)) {
-            final double[] update = computeFeatures(expect,
-                                                    features,
-                                                    learningRate);
-            if (best.compareAndSetFeatures(expect, update)) {
                 return best;
             }