IReNNS

Indipendent Recurrent Neural Network for Structures

IReNNS is an open source project developed at Politecnico di Milano.

IReNNS is the implementration of a recurrent, self-organizing map neural network derived from the Cascade-Correlation architecture. IReNNS can learn on structured data, i.e. data organized in graph structures: in particular, IReNNS can compute scalar functions on DOAGs (Directed Ordered Acyclic Graphs).

A schematic representation of the Neural Network architecture.

IReNNS was developed as software library to be used and integrated in other software or controlled by a user written "main" program. Neverthless, IReNNS has been provided with a simple user interface which lets you access easily to the main tasks and configuration parameters. IReNNS has been also provided with some libraries for the importation of datasets from CSV files. Structures can be represented as nested strings or in a SMILES-like syntax.

IReNNS is going to be released on this site in a few days. If you need any informations contact us!

 

IReNNS - (C)Copyright Antonino Trovato, 2008