The following information was submitted:
Transactions: WSEAS TRANSACTIONS ON CIRCUITS AND SYSTEMS
Transactions ID Number: 27-567
Full Name: Daniele Casali
Position: Doctor (Researcher)
Age: ON
Sex: Male
Address: Via del Politecnico, 1
Country: ITALY
Tel:
Tel prefix:
Fax:
E-mail address: daniele.casali@uniroma2.it
Other E-mails:
Title of the Paper: SVM-based Method for Associative Memories Implementation
Authors as they appear in the Paper: Daniele Casali, Giovanni Costantini, Renzo Perfetti, Elisa Ricci
Email addresses of all the authors: daniele.casali@uniroma2.it, costantini@uiroma2.it, perfetti@diei.unipg.it, elisa.ricci@diei.unipg.it
Number of paper pages: 10
Abstract: Generalized brain-state-in-a-box (GBSB) is one of the most known kind of Associative Memories. GBSB is based on a neural model, and needs an algorithm to perform the "training" operation that stores the patterns as equilibrium points of a system with a given state equation. Finding the optimum weight matrix to store a given set of patterns is the main objective of the training algorithm, and it can be seen as a classification problem. One of the most effective classification algorithms that we can find in literature is the Support Vector Machine (SVM). In this paper, we describe an algorithm that exploits capabilities of SVM to implement an efficient GBSB Associative Memory. Some properties of the networks designed in this way are evidenced, like a surprising generalized Hebb's law. The performance of the SVM approach is compared to existing methods with non-symmetric connections, by some design examples.
Keywords: Associative memories, neural networks, support vector machines
EXTENSION of the file: .rtf
Special (Invited) Session: A new Algorithm for Implementing BSB-based Associative Memories
Organizer of the Session: 591-551
How Did you learn about congress:
IP ADDRESS: 160.80.81.59
