Multimodal Networks

Multimodal networks in SNAP are represented by TMMNet, which consist of modes, which are of class TModeNet, and the links between them, which are of class TCrossNet.

The idea is that a multimodal network is a heterogeneous network where each node belongs to a particular mode, and edges belong to a particular cross net (that is, a particular kind of interaction between two modes). For example, in a biological dataset, genes, diseases and drugs might be the modes, and disease-disease interactions, disease-gene interactions and gene-drug interactions might be the links, or crossnets. To represent this in SNAP, we would build a TModeNet for each mode – each TModeNet would contain only the nodes belonging to that mode. Next, we would build a TCrossNet for every kind of link between modes. All edges should be added to the appropriate TCrossNet. Note that there can be a TCrossNet for links from the same mode to itself, and that there can be multiple TCrossNet objects linking the same pair of modes (for example, in a multimodal social network dataset with a mode corresponding to users, and another corresponding to photos, there can be one TCrossNet connecting users to photos they took, and another connecting users to photos they are tagged in).

The TMMNet class allows the construction of such multimodal networks in a modular fashion – the user adds the corresponding instances of TModeNet, specifying a name for each mode, and then adds the edges by adding instances of TCrossNet, specifying the name of the cross net and the modes which it links. (TCrossNet supports both undirected and directed multi-edges.)

MMNets can be loaded from a TTable, using functions LoadModeNetToNet() and LoadCrossNetToNet().

A TMMNet can also be converted into a TNEANet, using the ToNetwork() method (documented below), after which all the SNAP algorithms that work on regular networks can be run on it. The method allows us to specify exactly which crossnets we want to include in the network, so that we can simply pull out the subgraph of interest to us.

The following code shows example usage of TMMNet to construct a toy multimodal network. (All the methods used in this example are documented in detail below.)

import snap

mmnet = snap.TMMNet.New()

# Create a new modenet
mmnet.AddModeNet("TestMode1")

# Add a crossnet which has directed links from TestMode1 to itself.
mmnet.AddCrossNet("TestMode1", "TestMode1", "TestCross1", snap.TBool(True))

# Add a crossnet which has undirected links from TestMode1 to itself.
mmnet.AddCrossNet("TestMode1", "TestMode1", "TestCross2", snap.TBool(False))

# Add a second mode
mmnet.AddModeNet("TestMode2")

# Add a directed, and then an undirected crossnet from TestMode1 to TestMode2.
mmnet.AddCrossNet("TestMode1", "TestMode2", "TestCross3", snap.TBool(True))
mmnet.AddCrossNet("TestMode1", "TestMode2", "TestCross4", snap.TBool(False))

# Get the mode net objects, and add nodes to them.
modenet1 = mmnet.GetModeNetByName("TestMode1")
modenet2 = mmnet.GetModeNetByName("TestMode2")
for i in range(1000):
    modenet1.AddNode(i)
    modenet2.AddNode(i*2)

# Get the cross net objects, and add edges to them.
crossnet1 = mmnet.GetCrossNetByName("TestCross1")
crossnet2 = mmnet.GetCrossNetByName("TestCross2")
crossnet3 = mmnet.GetCrossNetByName("TestCross3")
crossnet4 = mmnet.GetCrossNetByName("TestCross4")
for i in range(1000):
    crossnet1.AddEdge(i, (i+1)%1000, i)
    crossnet2.AddEdge((i+5)%1000, i, i)
    crossnet3.AddEdge(i, (i%1000)*2, i)
    crossnet4.AddEdge((i+5)%1000, (i%1000)*2, i)

# Iterate over modes
modeneti = mmnet.BegModeNetI()
while modeneti < mmnet.EndModeNetI():
    print(modeneti.GetModeName())
    modeneti.Next()

# Iterate over crossnets
crossneti = mmnet.BegCrossNetI()
while crossneti < mmnet.EndCrossNetI():
    print(crossneti.GetCrossName())
    crossneti.Next()

# Get a subgraph
crossnets = snap.TStrV()
crossnets.add("TestCross1")
sub_mmnet = mmnet.GetSubgraphByCrossNet(crossnets)

# Convert to TNEANet

crossnetids = snap.TIntV()
crossnetids.Add(mmnet.GetCrossId("TestCross1"))
crossnetids.Add(mmnet.GetCrossId("TestCross2"))
crossnetids.Add(mmnet.GetCrossId("TestCross3"))

# These are mappings consisting of triples of (modeid, old attribute name, new attribute name)
nodeattrmapping = snap.TIntStrStrTrV()
edgeattrmapping = snap.TIntStrStrTrV()

pneanet = mmnet.ToNetwork(crossnetids, nodeattrmapping, edgeattrmapping)

TModeNet

class TModeNet

class TModeNet(ModeId)

class TModeNet(Nodes, Edges)

class TModeNet(Nodes, Edges, ModeId)

class TModeNet(Graph)

Returns a new directed multigraph with node and edge attributes that represents a mode in a TMMNet. If no parameters are provided, an empty graph is created. If Nodes and Edges are specified, space is preallocated for Nodes nodes and Edges edges. If Graph is specified, the new graph is a copy of the input graph. ModeId provides the integer id for the mode the TModeNet represents.

In general, a TModeNet should not be created directly and instead should be added to a multimodal network using the TMMNet method AddModeNet().

TModeNet inherits from TNEANet and therefore has all the same methods. In addition, it has the following multimodal related functions:

GetCrossNetNames(Names)

Gets a list of CrossNets that have this Mode as either a source or destination type.

GetNeighborsByCrossNet(NId, Name, Neighbors, isOutEId=False)

For the given node with id NId, gets all the neighbors for crossnet type with name Name. If this mode is both the source and dest type, the flag isOutEId specifies direction.

BegMMNI(SIn)

Returns an iterator referring to the first node in the graph.

EndMMNI(SOut)

Returns an iterator referring to the past-the-end node in the graph.

GetMMNI()

Returns an iterator referring to the node of ID NId in the graph.

TModeNetNodeI

class TModeNetNodeI

Returns a new node iterator for TModeNet. Normally, these objects are not created directly, but obtained via a call to the network class TModeNet method, such as BegMMNI(), that returns a node iterator.

TModeNetNodeI provides the following methods:

Next()

Moves the iterator to the next node in the graph.

GetId()

Returns node ID of the current node.

GetDeg()

Returns degree of the current node, the sum of in-degree and out-degree.

GetInDeg()

Returns in-degree of the current node.

GetOutDeg()

Returns out-degree of the current node.

GetInNId(NodeN)

Returns ID of NodeN-th in-node (the node pointing to the current node).

GetOutNId(NodeN)

Returns ID of NodeN-th out-node (the node the current node points to).

GetNbrNId(NodeN)

Returns ID of NodeN-th neighboring node.

IsInNId(NId)

Tests whether node with ID NId points to the current node.

IsOutNId(NId)

Tests whether the current node points to node with ID NId.

IsNbrNId(NId)

Tests whether node with ID NId is a neighbor of the current node.

GetCrossNetNames(Names)

Gets a list of CrossNets that include the mode this node belongs to as either a source or destination type.

GetNeighborsByCrossNet(Name, Neighbors, isOutEId=False)

For the given node, gets all the neighbors for crossnet type with name Name. If this mode is both the source and dest type, the flag isOutEId specifies direction.

TCrossNet

class TCrossNet

class TCrossNet(SrcModeId, DstModeId, CrossNetId)

class TCrossNet(SrcModeId, DstModeId, IsDir, CrossNetId)

class TCrossNet(Graph)

Returns a new crossnet, which consists of the edges between two different modes in a multimodal network. If no parameters are provided, an empty crossnet is created. SrcModeId and DstModeId provide the ids for the source and destination mode id. IsDir indicates whether the edges in the crossnet are directed. CrossNetId gives the id for this crossnet. If Graph is specified, the new crossnet is a copy of the input crossnet.

A TCrossNet should not be created directly and instead should be added to a multimodal network using the TMMNet method AddCrossNet().

Methods for TCrossNet are presented in two groups. The first group of methods deal with graph structure which includes edges. The second group of methods deal with edge attributes.

TCrossNet provides iterators for fast traversal of edges and attributes. Iterator classes are TCrossNetEdgeI for iterating over edges, and TCrossNetAIntI, TCrossNetAFltI, TCrossNetAStrI for iterating over integer, float or string attributes, respectively.

TCrossNet methods for graph structure are the following:

Save(SOut)

Saves the crossnet to a binary stream SOut.

GetEdges()

Returns the number of edges in the crossnet.

AddEdge(SrcNId, DstNId, EId=-1)

Adds an edge with ID EId between node IDs SrcNId and DstNId to the crossnet. Returns the ID of the edge being added. If EId is -1, edge ID is automatically assigned. Throws an exception, if an edge with ID EId already exists or if either SrcNId or DstNId does not exist.

DelEdge(EId)

Deletes an edge with id EId from the crossnet.

IsEdge(EId)

Tests whether an edge with id EId exists in the graph.

BegEdgeI()

Returns an edge iterator referring to the first edge in the crossnet.

EndEdgeI()

Returns an edge iterator referring to the past-the-end edge in the crossnet.

GetEdgeI(EId)

Returns an edge iterator referring to edge with id EId in the crossnet.

Clr()

Deletes all edges from the graph.

GetMode1()

Returns the id of the source mode.

GetMode2()

Returns the id of the destination mode.

IsDirected()

Returns whether edges in the crossnet are directed.

TCrossNet methods for edge attributes support attributes of different types. Integer, float and string attributes are implemented. Each attribute type has its own method for a particular task. Attributes are named via string names.

TCrossNet methods for attributes are the following:

AddIntAttrE(Attr)

AddFltAttrE(Attr)

AddStrAttrE(Attr)

Defines a new integer, float or string edge attribute, respectively.

DelAttrE(Attr)

Deletes edge attribute Attr.

AddIntAttrDatE(EdgeI, Value, Attr)

AddFltAttrDatE(EdgeI, Value, Attr)

AddStrAttrDatE(EdgeI, Value, Attr)

Sets the value of attribute named Attr for the edge referred to by edge iterator EdgeI to Value. Value is an integer, a float, or a string, respectively.

AddIntAttrDatE(EId, Value, Attr)

AddFltAttrDatE(EId, Value, Attr)

AddStrAttrDatE(EId, Value, Attr)

Sets the value of attribute named Attr for the edge with edge id EId to Value. Value is an integer, a float, or a string, respectively.

GetIntAttrDatE(EdgeI, Attr)

GetFltAttrDatE(EdgeI, Attr)

GetStrAttrDatE(EdgeI, Attr)

Returns the value of attribute named Attr for the edge referred to by edge iterator EdgeI*. Result is an integer, a float, or a string, respectively.

GetIntAttrDatE(EId, Attr)

GetFltAttrDatE(EId, Attr)

GetStrAttrDatE(EId, Attr)

Returns the value of attribute named Attr for the edge with edge id EId. Result is an integer, a float, or a string, respectively.

BegEAIntI(Attr)

BegEAFltI(Attr)

BegEAStrI(Attr)

Returns an integer, float, or string attribute iterator, respectively, of the attribute named Attr referring to the first edge.

EndEAIntI(Attr)

EndEAFltI(Attr)

EndEAStrI(Attr)

Returns an integer, float, or string attribute iterator, respectively, of the attribute named Attr referring to the past-the-end edge.

GetEAIntI(Attr, EId)

GetEAFltI(Attr, EId)

GetEAStrI(Attr, EId)

Returns an integer, float, or string attribute iterator, respectively, of the attribute named Attr referring to the edge with edge ID EId.

DelAttrDatE(EdgeI, Attr)

Deletes the value of attribute named Attr for the edge referred to by edge iterator EdgeI.

DelAttrDatE(EId, Attr)

Deletes the value of attribute named Attr for the edge with edge ID EId.

IsIntAttrDeletedE(EId, Attr)

IsFltAttrDeletedE(EId, Attr)

IsStrAttrDeletedE(EId, Attr)

Returns whether the int, float, or string attribute, respectively has been deleted.

TCrossNetEdgeI

class TCrossNetEdgeI

Returns a new edge iterator for TCrossNet. Normally, these objects are not created directly, but obtained via a call to the graph class TCrossNet method, such as BegEdgeI(), that returns an edge iterator.

TCrossNetEdgeI provides the following methods:

Next()

Moves the iterator to the next edge in the graph.

GetId()

Returns the the edge id.

GetSrcNId()

Returns the ID of the source node of the edge.

GetDstNId()

Returns the ID of the destination node of the edge.

GetSrcModeId()

Returns the ID of the source mode of the edge.

GetDstModeId()

Returns the ID of the destination mode of the edge.

IsDirected()

Returns whether the edge is directed.

TCrossNetAIntI, TCrossNetAFltI, TCrossNetAStrI

class TCrossNetAIntI

class TCrossNetAFltI

class TCrossNetAStrI

Returns a new integer, float or string attribute iterator for TCrossNet. Normally, these objects are not created directly, but obtained via a call to the graph class TCrossNet method, such as BegEAIntI(), which returns an integer edge iterator, or BegEAFltI(), which returns a float edge iterator.

Attribute iterators provide the following methods:

Next()

Moves the iterator to the next node or edge in the graph.

GetDat()

Returns an attribute of the node or edge.

IsDeleted()

Returns true if the attribute has been deleted.

TMMNet

class TMMNet

class TMMNet(Graph)

Returns a new directed multimodal network, consisting of different modes and the edges between them.

Modes have user-specified names and SNAP-assigned integer IDs, which are arbitrary non-negative integers. Cross-nets, which store the edges between two modes, also have user-specified names and SNAP-assigned integer IDs. Cross-nets are, by default, directed but can also be undirected. The same source mode can be used as the destination mode for a given cross-net.

TMMNet provides iterators for fast traversal of modes and cross-nets. Iterator classes are TMMNetModeNetI for iterating over modes and TMMNetCrossNetI for iterating over edges.

TMMNet methods are the following:

New()

Returns a pointer to a new multimodal network.

Load(SIn)

Loads the multimodal network from a binary stream SIn and returns a pointer to it.

Save(SOut)

Saves the multimodal network to a binary stream SOut.

GetModeNets()

Returns the number of modes in the graph.

AddModeNet(ModeName)

Adds a mode with name ModeName to the multimodal network. Returns the id for the mode.

DelModeNet(ModeId)

DelModeNet(ModeName)

Deletes the mode with id ModeId or name ModeName, respectively, from the multimodal network.

BegModeNetI()

Returns a mode iterator referring to the first mode in the graph.

EndModeNetI()

Returns a mode iterator referring to the past-the-end mode in the graph.

GetModeNetI(MId)

Returns a mode iterator referring to the mode with ID MId in the graph.

GetModeId(ModeName)

Returns the id of the mode with name ModeName.

GetModeName(ModeId)

Returns the name of the mode with id ModeId.

GetModeNetByName(ModeName)

GetModeNetById(ModeId)

Returns a reference to the mode with name ModeName or id ModeId, respectively, in the multimodal network.

GetCrossNets()

Returns the number of crossnets in the graph.

AddCrossNet(ModeName1, ModeName2, CrossNetName, IsDir=True)

AddCrossNet(ModeId1, ModeId2, CrossNetName, IsDir=True)

Adds a crossnet with name CrossNetName from the modes specified with the given names or ids. IsDir indicates whether the edges in the crossnet are directed.

DelCrossNet(CrossId)

DelCrossNet(CrossName)

Deletes the crossnet with id CrossId or name CrossName, respectively, from the multimodal network.

BegCrossNetI()

Returns a crossnet iterator referring to the first crossnet in the graph.

EndCrossNetI()

Returns a crossnet iterator referring to the past-the-end crossnet in the graph.

GetCrossNetI(CId)

Returns a crossnet iterator referring to the crossnet with ID CId in the graph.

GetCrossId(CrossName)

Returns the id of the crossnet with name CrossName.

GetCrossName(CrossId)

Returns the name of the crossnet with id CrossId.

GetCrossNetByName(CrossName)

GetCrossNetById(CrossId)

Returns a reference to the crossnet with name CrossName or id CrossId, respectively, in the multimodal network.

ToNetwork(TIntV& CrossNetTypes, TIntStrStrTrV& NodeAttrMap, TIntStrStrTrV& EdgeAttrMap)

Converts the MMNet to a TNEANet (which flattens out the multimodal nature of the network), adding only the crossnets (and corresponding modenets) whose ids are specified in the vector of integer ids, CrossNetTypes.

As attribute names can collide (since different modes can have the same attribute name in a TMMNet, but can’t anymore once it is converted to a TNEANet), two attribute maps are passed, one for modes and one for crossnets. Each attribute map is passed as a vector of triples. Each triple has the mode id, the attribute name in the TMMNet, and the attribute name to be used in the newly created TNEANet.

TMMNetModeNetI

class TMMNetModeNetI

Returns a new mode iterator for TMMNet. Normally, these objects are not created directly, but obtained via a call to the network class TMMNet method, such as BegModeNetI(), that returns a mode iterator.

TMMNetModeNetI provides the following methods:

Next()

Moves the iterator to the next mode in the graph.

GetModeId()

Returns the ID of the current mode.

GetModeName()

Returns the name of the current mode.

GetModeNet()

Returns a reference to the current mode.

TMMNetCrossNetI

class TMMNetCrossNetI

Returns a new crossnet iterator for TMMNet. Normally, these objects are not created directly, but obtained via a call to the graph class TMMNet method, such as BegCrossNetI(), that returns an crossnet iterator.

TMMNetCrossNetI provides the following methods:

Next()

Moves the iterator to the next crossnet in the graph.

GetCrossId()

Returns the ID of the current crossnet.

GetCrossName()

Returns the name of the current crossnet.

GetCrossNet()

Returns a reference to the current crossnet.