Abstract: In a distributed information network, a method and system for assembling messages in an input set of messages into one or more message assemblies. The method involves constructing an initial directed graph of the set of messages, wherein the directed graph consists of pairs of a first vertex and a second vertex, wherein, for each message in the set of messages, a first vertex is an the out node of the message and the second vertex is the in node of the message. A connectivity matrix is defined on the initial directed graph that is modified in an iterative process. A final directed graph is produced that is partitioned into disjoint subgraphs, each subgraph being an assembly of messages from the input message set.

Abstract: The invention provides a method for learning and identification of message types in a network in which the learning and identification is based solely upon information contained in messages in the network. In one embodiment, the method includes comparing a current message with one or more previously defined message structures and applying a comparison method in order to infer whether the structure of the current message conforms to any one of the previously defined message structures. The invention also provides a method for classifying a collection of messages input or output from a computational node in which the set of messages is portioned into subsets of similar messages, and a structure of the messages of each subset is characterized. The invention further provides a method for inferring causal relations between a message type input to a computational node in a network and message type output from the computational node in the network using one or more statistical tests.