Academia.eduAcademia.edu

Outline

Title
Abstract
Introduction
Experimental Results
Conclusions
References

A Modification of the Lernmatrix for Real Valued Data Processing

Profile image of LUIS PEREZLUIS PEREZ

2012

https://doi.org/10.1007/978-3-642-33275-3_60
visibility

…

description

8 pages

link

1 file

Abstract

An associative memory is a binary relationship between inputs and outputs, which is stored in an M matrix. In this paper, we propose a modification of the Steinbuch Lernmatrix model in order to process real-valued patterns, avoiding binarization processes and reducing computational burden. The proposed model is used in experiments with noisy environments, where the performance and efficiency of the memory is proven. A comparison between the proposed and the original model shows a good response and efficiency in the classification process of the new Lernmatrix.

Related papers

Neural network models for pattern recognition and associative memory
Gail Carpenter

Neural Networks, 1989

View PDFchevron_right
On the performance issues of the bidirectional associative memory
John Bragansa

1993

View PDFchevron_right
Designing associative memories implemented via recurrent neural networks for pattern recognition
Edgar Hernandez Duran

2011

In this paper a recurrent neural network is used as associative memory for pattern recognition. The goal of associative memory is to retrieve a stored pattern when enough information is presented in the network input. The network is training with twelve bipolar patterns to determine the corresponding weights. The weights are calculated by means of support vector machines training algorithms as the optimal hyperplane and soft margin hyperplane. Once the neural network is trained its performance is evaluated to retrieval stored patterns which correspond to characters encoded as bipolar vectors.

View PDFchevron_right
A Neural Net Associative Memory for Real-Time Applications
Giorgos Papadourakis

Neural Computation - NECO, 1990

A parallel hardware implementation of the associative memory neural network introduced by Hopfield is described. The design utilizes the Geometric Arithmetic Parallel Processor (GAPP), a commercially available single-chip VLSI general-purpose array processor consisting of 72 processing elements. The ability to cascade these chips allows large arrays of processors to be easily constructed and used to implement the Hopfield network. The memory requirements and processing times of such arrays are analyzed based on the number of nodes in the network and the number of exemplar patterns. Compared with other digital implementations, this design yields significant improvements in runtime performance and offers the capability of using large neural network associative memories in real-time applications.

View PDFchevron_right
A Method for the Improvement of the Behavior of Bidirectional Associative Memories as Pattern Classifiers
DIEGO AGUIRRE

Neural Processing Letters, 2003

We present a form of attaining success levels of up to 100% in character classification by the appropriate use of thresholds in the activity functions of the neurons making up the two-layer network with which bidirectional associative memories are implemented, together with a systematic method for generating the weight matrix. The system that is constructed includes a geometrical pre-processing stage that eliminates distortions, thereby improving the results. As a final characteristic, the functioning of the system presents a high level of immunity to noise or deformations. The system was evaluated using the two popular databases NIST#19 and UCI. There was found to be no misclassification in any case, whether under conditions of heavy contamination from noise or distortion of the image to be classified.

View PDFchevron_right
An Effective Associative Memory for Pattern Recognition
Boris Kryzhanovsky, Anatoly Fonarev, Leonid Litinskii

Lecture Notes in Computer Science, 2003

Neuron models of associative memory provide a new and prospective technology for reliable date storage and patterns recognition. However, even when the patterns are uncorrelated, the efficiency of most known models of associative memory is low. We developed a new version of associative memory with record characteristics of its storage capacity and noise immunity, which, in addition, is effective when recognizing correlated patterns.

View PDFchevron_right
A Neural Associative Pattern Classifier
DIEGO AGUIRRE

2002

In this work, we study the behaviour of the Bidirectional Associative Memory (BAM) in terms of the supporting neural structure, with a view to its possible improvements as a useful Pattern Classifier by means of class associations from unknown inputs, once mentioned classes have been previously defined by one or even more prototypes. The best results have been obtained by suitably choosing the training pattern pairs, the thresholds, and the activation functions of the network's neurones, by means of certain proposed methods described in the paper. In order to put forward the advantages of these proposed methods, the classifier has been applied on an especially popular hand-written character set as the well-known NIST#19 character database, and with one of the UCI's data bases. In all cases, the method led to a marked improvement in the performance achievable by a BAM, with a 0% error rate.

View PDFchevron_right
Intelligent Data Processing Based on Multi-Dimensional Numbered Memory Structures
Koen Vanhoof

IGI Global eBooks, 2013

The Multi-layer Pyramidal Growing Networks (MPGN) are memory structures based on multidimensional numbered information spaces (Markov, 2004), which permit us to create association links (bonds), hierarchically systematizing, and classification the information simultaneously with the input of it into memory. This approach is a successor of the main ideas of Growing Pyramidal Networks (Gladun, 2003), such as hierarchical structuring of memory that allows reflecting the structure of composing instances and gender-species bonds naturally, convenient for performing different operations of associative search. The recognition is based on reduced search in the multi-dimensional information space hierarchies. In this chapter, the authors show the advantages of using the growing numbered memory structuring via MPGN in the field of class association rule mining. The proposed approach was implemented in realization of association rules classifiers and has shown reliable results.

View PDFchevron_right
Associative memory models for data processing
Victor Quintana

International Journal of Electrical Power Energy Systems, 1992

This paper is an extension of work on a new framework for solving data acquisition and processing problems in power systems. The proposed methodology applies pattern analysis techniques to solve the network configuration, observability analysis and bad data processing problems. Other associative memory models are investigated for the solution of the observability analysis and bad data processing tasks. Special emphasis has been given to pattern analysis tools suitable for massively parallel implementations,such as artificial neural network models. Test results have been obtained for the IEEE 24-and ll8-bus test systems.

View PDFchevron_right
A binary correlation matrix memory k-nn classifier with hardware implementation
Jim Austin

British Machine Vision Conference, 1998

This paper describes a generic and fast classifier that uses a binary CMM (Correlation Matrix Memory) neural network for storing and matching a large amount of patterns efficiently, and a k-NN rule for classification. To meet CMM input requirements, a robust encoding method is proposed to convert numerical inputs into binary ones with the maximally achievable uniformity. To reduce the execution bottleneck, a hardware implementation of the CMM is described, which shows the network with on-board training and testing operates at over 200 times the speed of a current mid-range workstation, and is scaleable to very large problems. The CMM classifier has been tested on several benchmarks and, comparing with a simple k-NN classifier, it gave less than 1% lower accuracy and over 4 and 12 times speed-ups in software and hardware respectively.

View PDFchevron_right

References (25)

  1. Abu, M., Jacques, J.: Information capacity of the Hopfield model. IEEE Transactions on Information Theory, IT-31(4), 461-464 (1985)
  2. Acevedo, M.: Memorias asociativas bidireccionales alfa -beta. Ph.D. Thesis, Centre of Computer Research, Institute Polytechnic National, Mexico (2006)
  3. Amari, S.: Learning patterns and pattern sequences by self-organizing nets of threshold elements. IEEE Transaction on Computers C-21(11), 1197-1206 (1972)
  4. Anderson, J., Bower, G.: Memoria Asociativa, México, Limusa (1977)
  5. Austin, J.: ADAM: a distributed associative memory for scene analysis. In: Proceedings of First International Conference on Neural Networks, pp. 285-295 (1987)
  6. Flores, R., Yáñez, C.: Memorias asociativas alfa beta basadas en código Johnson Möbius modificado. Ms.C. Thesis. Center of Computer Research, Institute Polytechnic National, Mexico (2006)
  7. Hassoun, M.: Associative Neural Memories. Oxford University Press, New York (1993)
  8. Hopfield, J.J.: Neural networks and physical systems with emergent collective computational abilities. Proceedings of the National Academy of Sciences 79, 2554-2558 (1982)
  9. Kohonen, T.: Correlation matrix memories. IEEE Transactions on Computers C-2I(4), 353-359 (1972)
  10. Kosko, B.: Bidirectional associative memories. IEEE Transactions on Systems, Man and Cybernetics 18(1), 49-60 (1998)
  11. Lidner, D.: Introducción a las señales y los sistemas. McGraw Hill (1997)
  12. Palm, G., Schwenker, F., Sommer, F.T., Strey, A.: Neural associative memories. In: Krikelis, A., Weems, C.C. (eds.) Associative Processing and Processors, pp. 307-326. IEEE Computer Society, Los Alamitos (1997)
  13. Ritter, G., Sussner, P., Días de León, J.: Morphological associative memories. IEEE Transactions on Neural Networks 9, 281-293 (1998)
  14. Román, G., López, I., Yáñez, M.: A new classifier based on Associative Memories. In: Proc. 15th International Conference on Computing CIC, pp. 55-59. IEEE Computer Society, Los Alamitos (2006)
  15. Sánchez, F., Días de León, J., Yáñez, C.: Lernmatrix de Steinbuch: condiciones necesarias y suficientes para recuperación perfecta de patrones. Research on Computing Science 10, 91-102 (2002)
  16. Sánchez, A., Díaz de León, S., Yáñez, C.: New results on the Lernmatrix properties. Research in Computing Science 10, 91-102 (2004a)
  17. Sánchez, A., Díaz de León, S., Yáñez, C.: Lernmatrix de Steinbuch: Avances Teóricos. Computación y Sistemas 7(3), 175-189 (2004b)
  18. Simpson, P.K.: Artificial Neural Systems. Pergamon Press, New York (1990)
  19. Steinbuch, K.: Die Lernmatrix. Kybernetik 1(1), 36-45 (1961)
  20. Steinbuch, K., Frank, H.: Nichtdigitale Lernmatrizen als perzeptoren. Kybernetik 1(3), 117-124 (1961)
  21. Vázquez, H., Sossa, H.: A bidirectional hetero-associative memory for true -color patterns. Neural Processing Letters 28, 131-153 (2008)
  22. Willshaw, D., Buneman, O., Longuet-Higgins, H.: Non-holographic associative memory. Nature 222, 960-962 (1969)
  23. Yáñez, C.: Memorias Asociativas basadas en relaciones de orden y operadores binarios. Ph.D. Thesis. Center of Computer Research, Institute Polytechnic National, Mexico (2002)
  24. Yáñez, C., Díaz de León, J.: 2003a. Memorias Autoasociativas Morfológicas max: condiciones suficientes para convergencia, aprendizaje y recuperación de patrones. IT-I75, Serie Azul, CIC-IPN, Mexico (2003) ISBN 970-36-0035-2
  25. Yáñez, C., Díaz de León, J.: Memorias Autoasociativas Morfológicas min: condiciones suficientes para convergencia, aprendizaje y recuperación de patrones. IT-I77, Serie Azul, CIC-IPN, Mexico (2003b) ISBN 970-36-0036-0

Related papers

A New Classifier Based on Associative Memories
Itzamá López Yáñez

IEEE Computer Society, 2006

The Lernmatrix, which is the first known model of associative memory, is an heteroassociative memory, but it can also act as a binary pattern classifier depending on the choice of the output patterns. However, this model suffers two great problems: saturation and imperfect recall of some of the associations, even in the fundamental set, depending on the associations. In this work, a modification to the original Lernmatrix recall phase algorithm is presented. This modification improves the recalling capacity of the original model. Experimental results show this improvement

View PDFchevron_right
Performance of a Digital Associative Memory Model for Pattern Recognition
Raida Al-Alawi

Journal of Intelligent and Fuzzy Systems, 1997

This article presents a digital associative memory (DAM) with pyramids of probabilistic logic nodes used as the basic processing element. The DAM can be applied to various pattern recognition systems or image classifiers. A reward/ penalty error back propagation algorithm used to train the model will be described. Computer simulations are done to evaluate the peTj'ormance of the model by training the network on associating a number of patterns from each class of the numerals 0-9 with their prototype model. The effect of the size of the training set on the convergence of the training algorithm is investigated.

View PDFchevron_right
New Associative Memories to Recall Real-Valued Patterns
Ricardo Barrón

2004

In this note we describe a new set of associative memories able to recall patterns in the presence of mixed noise. Conditions are given under which the proposed memories are able to recall patterns either from the fundamental set of patterns and from distorted versions of them. Numerical and real examples are also provided to show the efficiency of the proposal.

View PDFchevron_right
Pattern classification and associative recall by neural networks
Tzi-dar Chiueh

2012

The first part of this dissertation discusses a new classifier based on a multilayer feed-forward network architecture. The main idea is to map irregularly-distributed prototypes in a classification problem to codewords that are organized in some way. Then the pattern classification problem is transformed into a threshold decoding problem, which is easily solved using simple hard-limiter neurons. At first we propose the new model and introduce two families of good "internal representation" codes. Then some analyses and software simulation concerning the storage capacity of this new model are done. The results show that the new classifier is much better than the classifier based on the Hopfield model in terms of both the storage capacity and the ability to classify correlated prototypes. A general model for neural network associative memories with a feedback. structure is proposed. Many existing neural network associative memories can be expressed as special cases of this g...

View PDFchevron_right
Pattern classification using smallest normalized difference associative memory
Mario Aldape-Pérez

Pattern Recognition Letters, 2017

In this paper a new associative classification algorithm is presented. The proposed algorithm overcomes the limitations of the original Alpha-Beta associative memory, while maintaining the fundamental set recalling capacity. This algorithm has two phases. The first phase is based on an Alpha-Beta auto-associative memory, which works in the domain of real numbers, unlike the traditional Alpha-Beta associative memories. In the second phase, normalized difference between the results of first phase and every pattern of the fundamental set is calculated. In order to demonstrate the behaviour and accuracy of the algorithm, multiple well known datasets and classification algorithms have been used. Experimental results have shown that our proposal achieved the best performance in three of the eight pattern classification problems in the medical field, using Stratified 10 Fold cross-validation. Our proposal achieved the best classification accuracy averaged over the all datasets addressed in the present work. Experimental results and statistical significance tests, allow us to affirm that the proposed model is an efficient alternative to perform pattern classification tasks.

View PDFchevron_right
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved