[MINOR] Fix proper seed handling in sparsity estimator LayeredGraph

In order to fix spurious test failures for specific seeds, we now
fix the seeds in the respective tests, and implement proper seed
handling in the sparsity estimator LayeredGraph.

Author: mboehm7

src/main/java/org/apache/sysds/hops/estim/EstimatorLayeredGraph.java

@@ -32,6 +32,7 @@
 import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.List;
+import java.util.Random;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
 
@@ -44,15 +45,21 @@
  */
 public class EstimatorLayeredGraph extends SparsityEstimator {
 
-	private static final int ROUNDS = 512;
+	public static final int ROUNDS = 512;
 	private final int _rounds;
+	private final Random _seeds;
 
 	public EstimatorLayeredGraph() {
 		this(ROUNDS);
 	}
 
 	public EstimatorLayeredGraph(int rounds) {
+		this(rounds, (int)System.currentTimeMillis());
+	}
+	
+	public EstimatorLayeredGraph(int rounds, int seed) {
 		_rounds = rounds;
+		_seeds = new Random(seed);
 	}
 
 	@Override
@@ -73,9 +80,9 @@ public LayeredGraph traverse(MMNode node) {
 		LayeredGraph ret, left, right;
 
 		left = (node.getLeft().getData() == null)
-			? retL : new LayeredGraph(node.getLeft().getData(), _rounds);
+			? retL : new LayeredGraph(node.getLeft().getData(), _rounds, _seeds.nextInt());
 		right = (node.getRight().getData() == null)
-			? retR : new LayeredGraph(node.getRight().getData(), _rounds);
+			? retR : new LayeredGraph(node.getRight().getData(), _rounds, _seeds.nextInt());
 
 		ret = estimInternal(left, right, node.getOp());
 
@@ -86,24 +93,24 @@ public LayeredGraph traverse(MMNode node) {
 	public double estim(MatrixBlock m1, MatrixBlock m2, OpCode op) {
 		if( op == OpCode.MM )
 			return estim(m1, m2);
-		LayeredGraph lg1 = new LayeredGraph(m1, _rounds);
-		LayeredGraph lg2 = new LayeredGraph(m2, _rounds);
+		LayeredGraph lg1 = new LayeredGraph(m1, _rounds, _seeds.nextInt());
+		LayeredGraph lg2 = new LayeredGraph(m2, _rounds, _seeds.nextInt());
 		LayeredGraph output = estimInternal(lg1, lg2, op);
 		return OptimizerUtils.getSparsity(
 			output._nodes.get(0).length, output._nodes.get(output._nodes.size() - 1).length, output.estimateNnz());
 	}
 
 	@Override
 	public double estim(MatrixBlock m, OpCode op) {
-		LayeredGraph lg1 = new LayeredGraph(m, _rounds);
+		LayeredGraph lg1 = new LayeredGraph(m, _rounds, _seeds.nextInt());
 		LayeredGraph output = estimInternal(lg1, null, op);
 		return OptimizerUtils.getSparsity(
 			output._nodes.get(0).length, output._nodes.get(output._nodes.size() - 1).length, output.estimateNnz());
 	}
 
 	@Override
 	public double estim(MatrixBlock m1, MatrixBlock m2) {
-		LayeredGraph graph = new LayeredGraph(Arrays.asList(m1,m2), _rounds);
+		LayeredGraph graph = new LayeredGraph(Arrays.asList(m1,m2), _rounds, _seeds.nextInt());
 		return OptimizerUtils.getSparsity(
 			m1.getNumRows(), m2.getNumColumns(), graph.estimateNnz());
 	}
@@ -153,16 +160,21 @@ private List<OpCode> getOps(MMNode node, List<OpCode> ops) {
 	public static class LayeredGraph {
 		private final List<Node[]> _nodes; //nodes partitioned by graph level
 		private final int _rounds;         //length of propagated r-vectors 
+		private final Random _seeds;
 
-		public LayeredGraph(List<MatrixBlock> chain, int r) {
+		public LayeredGraph(int r, int seed) {
 			_nodes = new ArrayList<>();
 			_rounds = r;
+			_seeds = new Random(seed);
+		}
+		
+		public LayeredGraph(List<MatrixBlock> chain, int r, int seed) {
+			this(r, seed);
 			chain.forEach(i -> buildNext(i));
 		}
 
-		public LayeredGraph(MatrixBlock m, int r) {
-			_nodes = new ArrayList<>();
-			_rounds = r;
+		public LayeredGraph(MatrixBlock m, int r, int seed) {
+			this(r, seed);
 			buildNext(m);
 		}
 
@@ -215,7 +227,8 @@ public void buildNext(MatrixBlock mb) {
 		public long estimateNnz() {
 			//step 1: assign random vectors ~exp(lambda=1) to all leaf nodes
 			//(lambda is not the mean, if lambda is 2 mean is 1/2)
-			ExponentialDistribution random = new ExponentialDistribution(new Well1024a(), 1);
+			ExponentialDistribution random = new ExponentialDistribution(
+				new Well1024a(_seeds.nextInt()), 1);
 			for( Node n : _nodes.get(0) ) {
 				double[] rvect = new double[_rounds];
 				for (int g = 0; g < _rounds; g++)
@@ -234,7 +247,7 @@ private static double calcNNZ(double[] inpvec, int rounds) {
 		}
 
 		public LayeredGraph rbind(LayeredGraph lg) {
-			LayeredGraph ret = new LayeredGraph(List.of(), _rounds);
+			LayeredGraph ret = new LayeredGraph(List.of(), _rounds, _seeds.nextInt());
 
 			Node[] rows = new Node[_nodes.get(0).length + lg._nodes.get(0).length];
 			Node[] columns = _nodes.get(1).clone();
@@ -258,7 +271,7 @@ public LayeredGraph rbind(LayeredGraph lg) {
 		}
 
 		public LayeredGraph cbind(LayeredGraph lg) {
-			LayeredGraph ret = new LayeredGraph(List.of(), _rounds);
+			LayeredGraph ret = new LayeredGraph(List.of(), _rounds, _seeds.nextInt());
 			int colLength = _nodes.get(1).length + lg._nodes.get(1).length;
 
 			Node[] rows = _nodes.get(0).clone();
@@ -286,11 +299,11 @@ public LayeredGraph matMult(LayeredGraph lg) {
 			List<MatrixBlock> m = Stream.concat(
 				this.toMatrixBlockList().stream(), lg.toMatrixBlockList().stream())
 				.collect(Collectors.toList());
-			return new LayeredGraph(m, _rounds);
+			return new LayeredGraph(m, _rounds, _seeds.nextInt());
 		}
 
 		public LayeredGraph or(LayeredGraph lg) {
-			LayeredGraph ret = new LayeredGraph(List.of(), _rounds);
+			LayeredGraph ret = new LayeredGraph(List.of(), _rounds, _seeds.nextInt());
 			Node[] rows = new Node[_nodes.get(0).length];
 			for (int i = 0; i < _nodes.get(0).length; i++)
 				rows[i] = new Node();
@@ -319,7 +332,7 @@ public LayeredGraph or(LayeredGraph lg) {
 		}
 
 		public LayeredGraph and(LayeredGraph lg) {
-			LayeredGraph ret = new LayeredGraph(List.of(), _rounds);
+			LayeredGraph ret = new LayeredGraph(List.of(), _rounds, _seeds.nextInt());
 			Node[] rows = new Node[_nodes.get(0).length];
 			for (int i = 0; i < _nodes.get(0).length; i++)
 				rows[i] = new Node();
@@ -348,7 +361,7 @@ public LayeredGraph and(LayeredGraph lg) {
 		}
 
 		public LayeredGraph transpose() {
-			LayeredGraph ret = new LayeredGraph(List.of(), _rounds);
+			LayeredGraph ret = new LayeredGraph(List.of(), _rounds, _seeds.nextInt());
 			Node[] rows = new Node[_nodes.get(_nodes.size() - 1).length];
 			for (int i = 0; i < rows.length; i++)
 				rows[i] = new Node();
@@ -377,7 +390,7 @@ public LayeredGraph transpose() {
 		}
 
 		public LayeredGraph diag() {
-			LayeredGraph ret = new LayeredGraph(List.of(), _rounds);
+			LayeredGraph ret = new LayeredGraph(List.of(), _rounds, _seeds.nextInt());
 			Node[] rowsOld = _nodes.get(0);
 			Node[] columnsOld = _nodes.get(1);
 

src/test/java/org/apache/sysds/test/component/estim/OpBindChainTest.java

@@ -116,12 +116,16 @@ public void testBitsetCasecbind() {
 	//Layered Graph
 	@Test
 	public void testLGCaserbind() {
-		runSparsityEstimateTest(new EstimatorLayeredGraph(), m, k, n, sparsity, rbind);
+		runSparsityEstimateTest(
+			new EstimatorLayeredGraph(EstimatorLayeredGraph.ROUNDS, 7), 
+			m, k, n, sparsity, rbind);
 	}
 
 	@Test
 	public void testLGCasecbind() {
-		runSparsityEstimateTest(new EstimatorLayeredGraph(), m, k, n, sparsity, cbind);
+		runSparsityEstimateTest(
+			new EstimatorLayeredGraph(EstimatorLayeredGraph.ROUNDS, 3), 
+			m, k, n, sparsity, cbind);
 	}