[MINOR] Refactoring input and output parameters of dbscanApply and db…

scripts/builtin/dbscan.dml

@@ -39,11 +39,15 @@
 # ----------------------------------------------------------------------------------------------------------------------
 
 m_dbscan = function (Matrix[Double] X, Double eps = 0.5, Integer minPts = 5)
-    return (Matrix[Double] clusterMembers, Matrix[Double] clusterModel)
+    return (Matrix[Double] X, Matrix[Double] clusterModel, Double eps)
 {
   #check input parameter assertions
   if(minPts < 0) { stop("DBSCAN: Stopping due to invalid inputs: minPts should be greater than 0"); }
-  if(eps < 0) { stop("DBSCAN: Stopping due to invalid inputs: Epsilon (eps) should be greater than 0"); }
+  if(eps < 0) 
+  { 
+    print("DBSCAN: Epsilon (eps) should be greater than 0. Setting eps = 0.5");
+    eps = 0.5
+  }
 
   UNASSIGNED = 0;
 
@@ -76,6 +80,7 @@ m_dbscan = function (Matrix[Double] X, Double eps = 0.5, Integer minPts = 5)
     clusterMembers = components(G=adjacency, verbose=FALSE);
     # noise to 0
     clusterMembers = clusterMembers * (rowSums(adjacency) > 0);
-    clusterModel = removeEmpty(target=X, margin="rows", select = (clusterMembers > 0))     
+    clusterModel = removeEmpty(target=X, margin="rows", select = (clusterMembers > 0))
+    X = clusterMembers
   }
 }

scripts/builtin/dbscanApply.dml

@@ -25,7 +25,7 @@
 # ----------------------------------------------------------------------------
 # NAME              TYPE             DEFAULT    MEANING
 # ----------------------------------------------------------------------------
-# Xtest             Matrix[Double]   ---        The input Matrix to do outlier detection on.
+# X             Matrix[Double]   ---        The input Matrix to do outlier detection on.
 # clusterModel      Matrix[Double]   ---        Model of clusters to predict outliers against.
 # eps               Double           0.5        Maximum distance between two points for one to be considered reachable for the other.
 
@@ -36,23 +36,24 @@
 # outlierPoints     Matrix[Double]   ---        Predicted outliers
 
 
-m_dbscanApply = function (Matrix[Double] Xtest, Matrix[Double] clusterModel, Double eps = 0.5)
-    return (Matrix[double] outlierPoints)
+m_dbscanApply = function (Matrix[Double] X, Matrix[Double] clusterModel, Double eps)
+  return (Matrix[Double] cluster, Matrix[Double] outlierPoints)
 {
-  num_features_Xtest = ncol(Xtest);
-  num_rows_Xtest = nrow(Xtest);
+  num_features_Xtest = ncol(X);
+  num_rows_Xtest = nrow(X);
   num_features_model = ncol(clusterModel);
   num_rows_model = nrow(clusterModel);
 
   if(num_features_Xtest != num_features_model) {stop("DBSCAN Outlier: Stopping due to invalid inputs: features need to match");}
   if(eps < 0) { stop("DBSCAN Outlier: Stopping due to invalid inputs: Epsilon (eps) should be greater than 0"); }
   if(num_rows_model <= 0) { stop("DBSCAN Outlier: Stopping due to invalid inputs: Model is empty"); }
 
-  X = rbind(clusterModel, Xtest);
-  neighbors = dist(X);
+  Xall = rbind(clusterModel, X);
+  neighbors = dist(Xall);
   neighbors = replace(target = neighbors, pattern = 0, replacement = 2.225e-307);
   neighbors = neighbors - diag(diag(neighbors));
-  Xtest_dists = neighbors[(num_rows_model+1):nrow(X), 1:num_rows_model];
+  Xtest_dists = neighbors[(num_rows_model+1):nrow(Xall), 1:num_rows_model];
   withinEps = ((Xtest_dists <= eps) * (0 < Xtest_dists));
   outlierPoints = rowSums(withinEps) >= 1;
+  cluster = removeEmpty(target=outlierPoints, margin="rows", select=outlierPoints)
 }

src/test/java/org/apache/sysds/test/functions/builtin/part1/BuiltinDbscanApplyTest.java

@@ -86,7 +86,7 @@ private void runOutlierByDBSCAN(boolean defaultProb, int seedA, int seedB, doubl
 			String HOME = SCRIPT_DIR + TEST_DIR;
 
 			fullDMLScriptName = HOME + TEST_NAME + ".dml";
-			programArgs = new String[]{"-explain","-nvargs",
+			programArgs = new String[]{"-nvargs",
 				"X=" + input("A"), "Y=" + input("B"),"Z=" + output("C"), "eps=" + epsDB, "minPts=" + minPts};
 			fullRScriptName = HOME + TEST_NAME + ".R";
 			rCmd = getRCmd(inputDir(), inputDir(), Double.toString(epsDB), Integer.toString(minPts), expectedDir());

src/test/scripts/functions/builtin/dbscan.dml

@@ -22,5 +22,5 @@
 X = read($X);
 eps = as.double($eps);
 minPts = as.integer($minPts);
-[Y, model] = dbscan(X, eps, minPts);
+[Y, model, eps] = dbscan(X, eps, minPts);
 write(Y, $Y);

src/test/scripts/functions/builtin/dbscanApply.dml

@@ -24,6 +24,6 @@ Y = read($Y)
 eps = as.double($eps);
 minPts = as.integer($minPts);
 
-[indices, clusterModel] = dbscan(X = X, eps = eps, minPts = minPts);
-Z = dbscanApply(Xtest=Y, clusterModel = clusterModel, eps = eps);
+[indices, clusterModel, eps] = dbscan(X = X, eps = eps, minPts = minPts);
+[C, Z] = dbscanApply(X=Y, clusterModel = clusterModel, eps = eps);
 write(Z, $Z);