Modular Labeling RAAM

Artificial Neural Nets and Genetic AlgorithmsPublished 1 January 1995

Alessandro Sperduti, Antonina Starita

Citations12

TL;DR

A modular Labeling RAAM system is proposed to speed up the training procedure and for reducing in size the developed compressed representations, by resorting to the BAMs derived from each LRAAM module.

Abstract

The Labeling RAAM model is a neural network able to encode labeled graphs in fixed size representations. In order to speed up the training procedure and for reducing in size the developed compressed representations, we propose a modular Labeling RAAM. In order to develop the modular system, we face two main problems: the mapping problem, i.e., how to map components of the structures into modules; and the membership problem, i.e., discovering which module must be used for decoding a compressed representation. The mapping between components and modules can be decided on the basis of the strongly connected components of the structures. The membership problem is solved by resorting to the BAMs derived from each LRAAM module. Preliminary results on the modular system are encouraging.

Keywords

Computer Science

Journal of the Operational Research SocietyIntroduction to Algorithms

16,942 Citations1991V. J. Rayward‐Smith, Thomas H. Cormen +2 more

Introduction to algorithms

3,066 Citations1990Cormen, Thomas H, Leiserson, Charles E +1 more

The Journal of the Acoustical Society of America<i>Neural Networks and Fuzzy Systems</i>

2,357 Citations1998Bart Kosko, John C. Burgess

Artificial IntelligenceTensor product variable binding and the representation of symbolic structures in connectionist systems

959 Citations1990Paul Smolensky

The tensor product representation rests on a principled analysis of structure; it saturates gracefully as larger structures are represented; it permits recursive construction of complex representations from simpler ones; it extends naturally to continuous structures and continuous representational patterns.

Artificial IntelligenceRecursive distributed representations

866 Citations1990Jordan Pollack

This paper presents a connectionist architecture which automatically develops compact distributed representations for variable-sized recursive data structures, as well as efficient accessing mechanisms for them.

IEEE Transactions on Neural NetworksHolographic reduced representations

660 Citations1995Tony Plate

This paper describes a method for representing more complex compositional structure in distributed representations that uses circular convolution to associate items, which are represented by vectors.

Artificial IntelligenceMapping part-whole hierarchies into connectionist networks

311 Citations1990Geoffrey E. Hinton

Three different ways of mapping part-whole hierarchies into connectionist networks are described, suggesting that neural networks have two quite different methods for performing inference.

Artificial IntelligenceBoltzCONS: Dynamic symbol structures in a connectionist network

116 Citations1990David S. Touretzky

The point of the work is to show how neural networks can exhibit compositionality and distal access, two properties that distinguish symbol processing from lower-level cognitive functions such as pattern recognition.

View PDF

Encoding pyramids by Labeling RAAM

3 Citations2002S. Lonardi, Alessandro Sperduti +1 more

The LRAAM is a neural network able to develop reduced representations of labeled directed graphs that preserve structural similarities on a pyramid and it is suggested to exploit this skill in data compression and/or to discover scale affine auto-similarities.

Encoding Conceptual Graphs by Labeling RAAM

3 Citations1994M. de Gerlache, Alessandro Sperduti +1 more

The possibility to memorize and retrieve natural language sentences and especially medical language sentences given in this kind of formalism with the use of the LRAAM model and a generalization of the access by content procedures introducing the concept of Generalized Hopfield Network is discussed.