Add overlap parameter

README.md

@@ -55,6 +55,10 @@ Minimum size (#edges) of a pattern. Default is 1.
 
 Number of best patterns to report at end. Default is 3.
 
+`--overlap <overlap_type>`
+
+Controls how instances of a pattern may overlap. Possible overlap_type values are: none, vertex, edge. Overlap of "none" means no part of two instances of a pattern can overlap. Overlap of "vertex" means that any number of vertices can overlap, but no edges. Overlap of "edge" means that edges and vertices can overlap, but the two instances cannot be identical. Default is "none".
+
 `--prune`
 
 If enabled, Subdue removes any pattern whose value is worse than its parent pattern. Disabled by default.

src/Parameters.py

@@ -18,6 +18,7 @@ def __init__(self):
         self.maxSize = 0              # Maximum size (#edges) of a pattern; default (0) is |E|/2.
         self.minSize = 1              # Minimum size (#edges) of a pattern; default is 1.
         self.numBest = 3              # Number of best patterns to report at end; default is 3.
+        self.overlap = "none"         # Extent that pattern instances can overlap (none, vertex, edge)
         self.prune = False            # Remove any patterns that are worse than their parent.
         self.valueBased = False       # Retain all patterns with the top beam best values.
         self.writeCompressed = False  # Write compressed graph after iteration i to file outputFileName-compressed-i.json
@@ -55,6 +56,11 @@ def set_parameters (self, args):
             if optionName == "--numbest":
                 index += 1
                 self.numBest = int(args[index])
+            if optionName == "--overlap":
+                index += 1
+                overlap_type = args[index]
+                if overlap_type in ["none", "vertex", "edge"]:
+                    self.overlap = overlap_type
             if optionName == "--prune":
                 self.prune = True
             if optionName == "--valuebased":
@@ -79,6 +85,7 @@ def print(self):
         print("  Max Size: " + str(self.maxSize))
         print("  Min Size: " + str(self.minSize))
         print("  Num Best: " + str(self.numBest))
+        print("  Overlap: " + self.overlap)
         print("  Prune: " + str(self.prune))
         print("  Value Based: " + str(self.valueBased))
         print("  Write Compressed: " + str(self.writeCompressed))

src/Pattern.py

@@ -98,7 +98,7 @@ def CreatePatternFromInstances(definition, instances):
 
 # ----- Pattern Extension
 
-def ExtendPattern (pattern, temporal = False):
+def ExtendPattern (parameters, pattern):
     """Return list of patterns created by extending each instance of the given pattern by one edge in all possible ways,
        and then collecting matching extended instances together into new patterns."""
     extendedInstances = []
@@ -110,15 +110,15 @@ def ExtendPattern (pattern, temporal = False):
     while extendedInstances:
         newInstance = extendedInstances.pop(0)
         newInstanceGraph = Graph.CreateGraphFromInstance(newInstance)
-        if temporal:
+        if parameters.temporal:
             newInstanceGraph.TemporalOrder()
         matchingInstances = [newInstance]
         nonmatchingInstances = []
         for extendedInstance in extendedInstances:
             extendedInstanceGraph = Graph.CreateGraphFromInstance(extendedInstance)
-            if temporal:
+            if parameters.temporal:
                 extendedInstanceGraph.TemporalOrder()
-            if Graph.GraphMatch(newInstanceGraph,extendedInstanceGraph) and (not InstancesOverlap(matchingInstances,extendedInstance)):
+            if Graph.GraphMatch(newInstanceGraph,extendedInstanceGraph) and (not InstancesOverlap(parameters.overlap, matchingInstances, extendedInstance)):
                 matchingInstances.append(extendedInstance)
             else:
                 nonmatchingInstances.append(extendedInstance)
@@ -163,16 +163,26 @@ def InstanceMatch(instance1,instance2):
     else:
         return False
 
-def InstancesOverlap(instanceList,instance):
-    """Returns True if instance contains a vertex that is contained in an instance of the given instanceList."""
+def InstancesOverlap(overlap, instanceList, instance):
+    """Returns True if instance overlaps with an instance in the given instanceList
+    according to the overlap parameter, which indicates what type of overlap ignored.
+    Overlap="none" means no overlap ignored. Overlap="vertex" means vertex overlap
+    ignored. Overlap="edge" means vertex and edge overlap ignored, but the instances
+    cannot be identical."""
     for instance2 in instanceList:
-        if InstanceOverlap(instance,instance2):
+        if InstanceOverlap(overlap, instance, instance2):
             return True
     return False
 
-def InstanceOverlap(instance1,instance2):
-    """Returns True if given instances share a vertex."""
-    return instance1.vertices.intersect(instance2.vertices)
+def InstanceOverlap(overlap, instance1, instance2):
+    """Returns True if given instances overlap according to given overlap parameter.
+    See InstancesOverlap for explanation."""
+    if overlap == "edge":
+        return InstanceMatch(instance1, instance2)
+    elif overlap == "vertex":
+        return instance1.edges.intersect(instance2.edges)
+    else: # overlap == "none"
+        return instance1.vertices.intersect(instance2.vertices)
 
 
 # ----- Pattern List Operations

src/Subdue.py

@@ -28,7 +28,7 @@ def DiscoverPatterns(parameters, graph):
     """The main discovery loop. Finds and returns best patterns in given graph."""
     patternCount = 0
     # get initial one-edge patterns
-    parentPatternList = GetInitialPatterns(graph, parameters.temporal)
+    parentPatternList = GetInitialPatterns(parameters, graph)
     if DEBUGFLAG:
         print("Initial patterns (" + str(len(parentPatternList)) + "):")
         for pattern in parentPatternList:
@@ -42,7 +42,7 @@ def DiscoverPatterns(parameters, graph):
             parentPattern = parentPatternList.pop(0)
             if ((len(parentPattern.instances) > 1) and (patternCount < parameters.limit)):
                 patternCount += 1
-                extendedPatternList = Pattern.ExtendPattern(parentPattern, parameters.temporal)
+                extendedPatternList = Pattern.ExtendPattern(parameters, parentPattern)
                 while (extendedPatternList):
                     extendedPattern = extendedPatternList.pop(0)
                     if DEBUGFLAG:
@@ -66,15 +66,14 @@ def DiscoverPatterns(parameters, graph):
             Pattern.PatternListInsert(parentPattern, discoveredPatternList, parameters.numBest, False) # valueBased = False
     return discoveredPatternList
 
-def GetInitialPatterns(graph, temporal = False, overlap = True):
-    """Returns list of single-edge, evaluated patterns in given graph with more than one instance.
-    If overlap=False, then instances of a single-edge pattern cannot share vertices."""
+def GetInitialPatterns(parameters, graph):
+    """Returns list of single-edge, evaluated patterns in given graph with more than one instance."""
     initialPatternList = []
     # Create a graph and an instance for each edge
     edgeGraphInstancePairs = []
     for edge in graph.edges.values():
         graph1 = Graph.CreateGraphFromEdge(edge)
-        if temporal:
+        if parameters.temporal:
             graph1.TemporalOrder()
         instance1 = Pattern.CreateInstanceFromEdge(edge)
         edgeGraphInstancePairs.append((graph1,instance1))
@@ -89,7 +88,7 @@ def GetInitialPatterns(graph, temporal = False, overlap = True):
         for edgePair2 in edgeGraphInstancePairs:
             graph2 = edgePair2[0]
             instance2 = edgePair2[1]
-            if Graph.GraphMatch(graph1,graph2) and (overlap or (not Pattern.InstancesOverlap(pattern.instances,instance2))):
+            if Graph.GraphMatch(graph1,graph2) and (not Pattern.InstancesOverlap(parameters.overlap, pattern.instances, instance2)):
                 pattern.instances.append(instance2)
             else:
                 nonmatchingEdgePairs.append(edgePair2)
@@ -171,6 +170,7 @@ def nx_subdue(
     :param maxSize: (Default: 0)              -- Maximum size (#edges) of a pattern; default (0) is |E|/2.
     :param minSize: (Default: 1)              -- Minimum size (#edges) of a pattern; default is 1.
     :param numBest: (Default: 3)              -- Number of best patterns to report at end; default is 3.
+    :param overlap: (Defaul: none)            -- Extent that pattern instances can overlap (none, vertex, edge)
     :param prune: (Default: False)            -- Remove any patterns that are worse than their parent.
     :param valueBased: (Default: False)       -- Retain all patterns with the top beam best values.
     :param temporal: (Default: False)         -- Discover static (False) or temporal (True) patterns
@@ -223,7 +223,7 @@ def unwrap_output(iterations):
     return out
 
 def main():
-    print("SUBDUE v1.3 (python)\n")
+    print("SUBDUE v1.4 (python)\n")
     parameters = Parameters.Parameters()
     parameters.set_parameters(sys.argv)
     graph = ReadGraph(parameters.inputFileName)

testing/output.txt

@@ -1,148 +0,0 @@
-SUBDUE v1.3 (python)
-
-Parameters:
-  Input File Name: inputgraph.json
-  Output File Name: inputgraph
-  Beam Width: 4
-  Iterations: 1
-  Limit: 7
-  Max Size: 7
-  Min Size: 1
-  Num Best: 3
-  Prune: False
-  Value Based: False
-  Write Compressed: False
-  Write Pattern: False
-  Write Instances: False
-  Temporal: False
-
-Graph: 15 vertices, 15 edges
-7 patterns left
-4 patterns left
-1 patterns left
-No more patterns to consider
-
-Best 3 patterns:
-
-  Pattern (value=0.8, instances=5):
-    Graph:
-      vertex "1": timestamp=1, label=v1
-      vertex "2": timestamp=1, label=v2
-      vertex "3": timestamp=1, label=v3
-      edge "1" (1--2): timestamp=1, label=e12
-      edge "2" (1--3): timestamp=1, label=e13
-      edge "3" (2--3): timestamp=7, label=e23
-    Instance 1:
-      vertex "1": timestamp=1, label=v1
-      vertex "2": timestamp=1, label=v2
-      vertex "3": timestamp=1, label=v3
-      edge "1" (1--2): timestamp=1, label=e12
-      edge "2" (1--3): timestamp=1, label=e13
-      edge "3" (2--3): timestamp=7, label=e23
-    Instance 2:
-      vertex "4": timestamp=2, label=v1
-      vertex "5": timestamp=2, label=v2
-      vertex "6": timestamp=2, label=v3
-      edge "4" (4--5): timestamp=2, label=e12
-      edge "5" (4--6): timestamp=2, label=e13
-      edge "6" (5--6): timestamp=8, label=e23
-    Instance 3:
-      vertex "7": timestamp=4, label=v1
-      vertex "8": timestamp=4, label=v2
-      vertex "9": timestamp=4, label=v3
-      edge "9" (7--8): timestamp=4, label=e12
-      edge "10" (7--9): timestamp=4, label=e13
-      edge "11" (8--9): timestamp=11, label=e23
-    Instance 4:
-      vertex "10": timestamp=5, label=v1
-      vertex "11": timestamp=5, label=v2
-      vertex "12": timestamp=5, label=v3
-      edge "13" (10--11): timestamp=5, label=e12
-      edge "14" (10--12): timestamp=5, label=e13
-      edge "15" (11--12): timestamp=12, label=e23
-    Instance 5:
-      vertex "13": timestamp=9, label=v1
-      vertex "14": timestamp=9, label=v2
-      vertex "15": timestamp=9, label=v3
-      edge "20" (13--14): timestamp=9, label=e12
-      edge "21" (13--15): timestamp=9, label=e13
-      edge "22" (14--15): timestamp=16, label=e23
-
-  Pattern (value=0.5333333333333333, instances=5):
-    Graph:
-      vertex "1": timestamp=1, label=v1
-      vertex "2": timestamp=1, label=v2
-      vertex "3": timestamp=1, label=v3
-      edge "1" (1--2): timestamp=1, label=e12
-      edge "2" (1--3): timestamp=1, label=e13
-    Instance 1:
-      vertex "1": timestamp=1, label=v1
-      vertex "2": timestamp=1, label=v2
-      vertex "3": timestamp=1, label=v3
-      edge "1" (1--2): timestamp=1, label=e12
-      edge "2" (1--3): timestamp=1, label=e13
-    Instance 2:
-      vertex "4": timestamp=2, label=v1
-      vertex "5": timestamp=2, label=v2
-      vertex "6": timestamp=2, label=v3
-      edge "4" (4--5): timestamp=2, label=e12
-      edge "5" (4--6): timestamp=2, label=e13
-    Instance 3:
-      vertex "7": timestamp=4, label=v1
-      vertex "8": timestamp=4, label=v2
-      vertex "9": timestamp=4, label=v3
-      edge "9" (7--8): timestamp=4, label=e12
-      edge "10" (7--9): timestamp=4, label=e13
-    Instance 4:
-      vertex "10": timestamp=5, label=v1
-      vertex "11": timestamp=5, label=v2
-      vertex "12": timestamp=5, label=v3
-      edge "13" (10--11): timestamp=5, label=e12
-      edge "14" (10--12): timestamp=5, label=e13
-    Instance 5:
-      vertex "13": timestamp=9, label=v1
-      vertex "14": timestamp=9, label=v2
-      vertex "15": timestamp=9, label=v3
-      edge "20" (13--14): timestamp=9, label=e12
-      edge "21" (13--15): timestamp=9, label=e13
-
-  Pattern (value=0.5333333333333333, instances=5):
-    Graph:
-      vertex "1": timestamp=1, label=v1
-      vertex "2": timestamp=1, label=v2
-      vertex "3": timestamp=1, label=v3
-      edge "1" (1--2): timestamp=1, label=e12
-      edge "2" (2--3): timestamp=7, label=e23
-    Instance 1:
-      vertex "1": timestamp=1, label=v1
-      vertex "2": timestamp=1, label=v2
-      vertex "3": timestamp=1, label=v3
-      edge "1" (1--2): timestamp=1, label=e12
-      edge "3" (2--3): timestamp=7, label=e23
-    Instance 2:
-      vertex "4": timestamp=2, label=v1
-      vertex "5": timestamp=2, label=v2
-      vertex "6": timestamp=2, label=v3
-      edge "4" (4--5): timestamp=2, label=e12
-      edge "6" (5--6): timestamp=8, label=e23
-    Instance 3:
-      vertex "7": timestamp=4, label=v1
-      vertex "8": timestamp=4, label=v2
-      vertex "9": timestamp=4, label=v3
-      edge "9" (7--8): timestamp=4, label=e12
-      edge "11" (8--9): timestamp=11, label=e23
-    Instance 4:
-      vertex "10": timestamp=5, label=v1
-      vertex "11": timestamp=5, label=v2
-      vertex "12": timestamp=5, label=v3
-      edge "13" (10--11): timestamp=5, label=e12
-      edge "15" (11--12): timestamp=12, label=e23
-    Instance 5:
-      vertex "13": timestamp=9, label=v1
-      vertex "14": timestamp=9, label=v2
-      vertex "15": timestamp=9, label=v3
-      edge "20" (13--14): timestamp=9, label=e12
-      edge "22" (14--15): timestamp=16, label=e23
-
-SUBDUE done. Elapsed time = 0.0024871826171875 seconds
-