<< back

Traditional applications cope with graph structured data by engaging a preprocessing phase which map the graph structured information to a simpler representation, e.g. vectors of reals.
However, important information, e.g., the topological dependency of information on node n may be lost during the preprocessing stage and the final result may depend, in an unpredictable manner, on the details of the preprocessing algorithm.
More recently, there are various approaches [1, 2] attempting to preserve the graph structured nature of the data for as long as required before processing the data. In other words, these approaches attempt to avoid the preprocessing step of “squashing” the graph structured data into a vector of reals first, and to deal with the preprocessed data using a list based data processing technique, rather than paying special attention to the underlying graph structured nature of the data. In these recent approaches, the idea is to encode the underlying graph structured data using the topological relationship among the nodes of the graph in order to incorporate the graph structured information in the data processing step. Recursive neural networks [1, 3, 4] belong to this set of techniques and are commonly applied in this kind of problems. Our research group has actively partecipated to definition of this new computation paradigm, contributing to the theoretical foundations [9, 10, 11, 12] and the experimental assessment of the model [13, 14, 15, 16].

  credits .: Artificial Intelligence Research Group of Siena :.
history The key ideas Demo