From 78763d1966bc6bb7702906b73aeb6b154577418e Mon Sep 17 00:00:00 2001
From: "trevor.cohn" <trevor.cohn@ec762483-ff6d-05da-a07a-a48fb63a330f>
Date: Fri, 9 Jul 2010 19:39:33 +0000
Subject: Prettyfied output regarding optimization failure.

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@210 ec762483-ff6d-05da-a07a-a48fb63a330f
---
 .../prjava/src/phrase/PhraseCluster.java              | 19 +++++++++++++++++--
 .../prjava/src/phrase/PhraseObjective.java            | 18 +++++++++++-------
 2 files changed, 28 insertions(+), 9 deletions(-)

(limited to 'gi/posterior-regularisation')

diff --git a/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java b/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java
index 7d7c46dd..b9b1b98c 100644
--- a/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java
+++ b/gi/posterior-regularisation/prjava/src/phrase/PhraseCluster.java
@@ -9,6 +9,7 @@ import java.util.List;
 import java.util.concurrent.ExecutorService;
 import java.util.concurrent.Executors;
 import java.util.concurrent.LinkedBlockingQueue;
+import java.util.concurrent.atomic.AtomicInteger;
 
 import phrase.Corpus.Edge;
 
@@ -110,10 +111,13 @@ public class PhraseCluster {
 		double [][]exp_pi=new double[n_phrases][K];
 		
 		double loglikelihood=0, kl=0, l1lmax=0, primal=0;
+		int failures=0, iterations=0;
 		//E
 		for(int phrase=0; phrase<n_phrases; phrase++){
 			PhraseObjective po=new PhraseObjective(this,phrase);
-			po.optimizeWithProjectedGradientDescent();
+			boolean ok = po.optimizeWithProjectedGradientDescent();
+			if (!ok) ++failures;
+			iterations += po.iterations;
 			double [][] q=po.posterior();
 			loglikelihood += po.loglikelihood();
 			kl += po.KL_divergence();
@@ -136,6 +140,9 @@ public class PhraseCluster {
 			}
 		}
 		
+		if (failures > 0)
+			System.out.println("WARNING: failed to converge in " + failures + "/" + n_phrases + " cases");
+		System.out.println("\tmean iters: 	  " + iterations/(double)n_phrases);
 		System.out.println("\tllh:            " + loglikelihood);
 		System.out.println("\tKL:             " + kl);
 		System.out.println("\tphrase l1lmax:  " + l1lmax);
@@ -170,6 +177,8 @@ public class PhraseCluster {
 		double [][]exp_pi=new double[n_phrases][K];
 		
 		double loglikelihood=0, kl=0, l1lmax=0, primal=0;
+		final AtomicInteger failures = new AtomicInteger(0);
+		int iterations=0;
 
 		//E
 		for(int phrase=0;phrase<n_phrases;phrase++){
@@ -179,7 +188,8 @@ public class PhraseCluster {
 					try {
 						//System.out.println("" + Thread.currentThread().getId() + " optimising lambda for " + p);
 						PhraseObjective po = new PhraseObjective(PhraseCluster.this, p);
-						po.optimizeWithProjectedGradientDescent();
+						boolean ok = po.optimizeWithProjectedGradientDescent();
+						if (!ok) failures.incrementAndGet();
 						//System.out.println("" + Thread.currentThread().getId() + " done optimising lambda for " + p);
 						expectations.put(po);
 						//System.out.println("" + Thread.currentThread().getId() + " added to queue " + p);
@@ -206,6 +216,8 @@ public class PhraseCluster {
 				kl += po.KL_divergence();
 				l1lmax += po.l1lmax();
 				primal += po.primal();
+				iterations += po.iterations;
+
 				
 				List<Edge> edges = c.getEdgesForPhrase(phrase);
 				for(int edge=0;edge<q.length;edge++){
@@ -227,6 +239,9 @@ public class PhraseCluster {
 			}
 		}
 		
+		if (failures.get() > 0)
+			System.out.println("WARNING: failed to converge in " + failures.get() + "/" + n_phrases + " cases");
+		System.out.println("\tmean iters: 	  " + iterations/(double)n_phrases);
 		System.out.println("\tllh:            " + loglikelihood);
 		System.out.println("\tKL:             " + kl);
 		System.out.println("\tphrase l1lmax:  " + l1lmax);
diff --git a/gi/posterior-regularisation/prjava/src/phrase/PhraseObjective.java b/gi/posterior-regularisation/prjava/src/phrase/PhraseObjective.java
index 3314f74a..f24b903d 100644
--- a/gi/posterior-regularisation/prjava/src/phrase/PhraseObjective.java
+++ b/gi/posterior-regularisation/prjava/src/phrase/PhraseObjective.java
@@ -22,7 +22,7 @@ public class PhraseObjective extends ProjectedObjective
 {
 	static final double GRAD_DIFF = 0.00002;
 	static double INIT_STEP_SIZE = 300;
-	static double VAL_DIFF = 1e-4; // FIXME needs to be tuned - and this might be too weak
+	static double VAL_DIFF = 1e-6; // FIXME needs to be tuned - and this might be too weak
 	static int ITERATIONS = 100;
 	//private double c1=0.0001; // wolf stuff
 	//private double c2=0.9;
@@ -164,7 +164,9 @@ public class PhraseObjective extends ProjectedObjective
 		return q;
 	}
 	
-	public void optimizeWithProjectedGradientDescent(){
+	public int iterations = 0;
+	
+	public boolean optimizeWithProjectedGradientDescent(){
 		LineSearchMethod ls =
 			new ArmijoLineSearchMinimizationAlongProjectionArc
 				(new InterpolationPickFirstStep(INIT_STEP_SIZE));
@@ -181,13 +183,14 @@ public class PhraseObjective extends ProjectedObjective
 		compositeStop.add(stopValue);
 		optimizer.setMaxIterations(ITERATIONS);
 		updateFunction();
-		boolean succed = optimizer.optimize(this,stats,compositeStop);
+		boolean success = optimizer.optimize(this,stats,compositeStop);
+		iterations += optimizer.getCurrentIteration();
 //		System.out.println("Ended optimzation Projected Gradient Descent\n" + stats.prettyPrint(1));
-		if(succed){
+		//if(succed){
 			//System.out.println("Ended optimization in " + optimizer.getCurrentIteration());
-		}else{
-			System.out.println("Failed to optimize");
-		}
+		//}else{
+//			System.out.println("Failed to optimize");
+		//}
 		lambda[phrase]=parameters;
 		//	ps.println(Arrays.toString(parameters));
 		
@@ -195,6 +198,7 @@ public class PhraseObjective extends ProjectedObjective
 		//	ps.println(Arrays.toString(q[edge]));
 		//	}
 		
+		return success;
 	}
 	
 	public double KL_divergence()
-- 
cgit v1.2.3