Cross-Validation and Information Measures for Ram Based Neural Networks

Jørgensen, Thomas Martini; Christensen, Steen Sloth; Liisberg, C.

doi:10.1142/9789812816849_0008

Cited by 6 publications

(7 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Va.c =O=p(cIa1)=, o<&<<l. (6) The next problem one is confronted with is how to combine the probability estimates of the individual features. As the features and therefore also the probabilities in general are not independent we can not just multiply the individual probability estimates p(cJa1 ) to obtain p(c{a1 }) -the probability of class c given the whole feature set {a1 } sampled by the n-tuples.…”

Section: Standard Scheme Versus Frequentist Schemementioning

confidence: 99%

<title>Cross-validation techniques for n-tuple-based neural networks</title>

Linneberg¹,

Jørgensen

1999

SPIE Proceedings

View full text Add to dashboard Cite

In spite of the simple classification concept, impressing performances have been reported using the n-tuple architecture in combination with a very simple training strategy. In general, however, the performance of the n-tuple classifier is highly dependent on the choice of input connections and on the encoding of the input data. Accordingly, the simple architecture needs to be accompanied with design tools for obtaining a suitable architecture. Due to the simplicity of the architecture, it is simple to perform leave-one-out cross-validation tests and extensions of the concept. Therefore, it is also possible to operate with design methods that make extensively use of such tests. This paper describes such design algorithms and especially introduces a simple design strategy that allows the n-tuple architecture to perform satisfactorily in cases with skewed class priors. It can also help to resolve conflicts in the training material. The described methods are evaluated on classification problems from the European StatLog project. It is hereby shown that the design tools extends the competitiveness of the n-tuple classification method.

show abstract

Section: Standard Scheme Versus Frequentist Schemementioning

confidence: 99%

<title>Cross-validation techniques for n-tuple-based neural networks</title>

Linneberg¹,

Jørgensen

1999

SPIE Proceedings

View full text Add to dashboard Cite

show abstract

“…There will be a vast number of possible connections for a matrix with the dimension like 32 by 32. The classification and generalization performance are highly dependent on these input mappings [5] [13]. A random map is suitable for an unoptimised problem [3] as it samples the point throughout the pattern matrix.…”

Section: Introductionmentioning

confidence: 99%

Hybridisation of GA and PSO to optimise N-tuples

Azhar

Deravi

Dimond

2009

2009 IEEE International Conference on Systems, Man and Cybernetics

View full text Add to dashboard Cite

Among numerous pattern recognition methods the neural network approach has been the subject of much research due to its ability to learn from a given collection of representative examples. This paper is concerned with the design of a Weightless Neural Network, which decomposes a given pattern into several sets of n points, termed n-tuples. Considerable research has shown that by optimising the input connection mapping of such n-tuple networks classification performance can be improved significantly. This paper investigates the hybridisation of Genetic Algorithm (GA) and Particle Swarm Optimisation (PSO) techniques in search of better connection maps to the N-tuples. Experiments were conducted to evaluate the proposed method by applying the trained classifier to recognise hand-printed digits from a widely used database compiled by U.S. National Institute of Standards and Technology (NIST).

show abstract

“…There will be a vast number of possible connections for a matrix with the dimension like 32 by 32. The classification and generalization performance are highly dependant on these input mappings (Bishop, 1990;Jorgensen et al, 1995). A random map is suitable for an un-optimized problem as it samples the point throughout the pattern matrix (Aleksander, 1979).…”

Section: Introductionmentioning

confidence: 99%

“…A random map is suitable for an un-optimized problem as it samples the point throughout the pattern matrix (Aleksander, 1979). Considerable research shows that by optimizing the connections classification performance can be improved significantly (Azhar & Dimond, 2004a, 2004b, 2004cBishop, 1990;Jorgensen et al, 1995). Among different optimization techniques, the Particle Swarm Optimization (PSO) (Kennedy & Eberhart, 1995) exhibits good performance in finding solutions to static optimization problems .…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Particle Swarm Intelligence to Optimize the Learning of N-tuples

Azhar¹,

Deravi²,

Dimond³

2008

Journal of Intelligent Systems

View full text Add to dashboard Cite

This paper concentrates on the swarm intelligence based bio-inspired approach to optimize N-tuple classifiers. We will explain the implementation of the particle swarm (PS) based training for N-tuple classifier that helps to recognize patterns more efficiently by learning the good sets of N-tuples. Various versions of the particle swarm based training will be explored by changing different parameters of the system. The swarm algorithm will also be hybridized with other bio-inspired techniques like Self-Organized Criticality (SOC) and the Fitness to Distance Ratio (FDR) based selection to add diversity in the swarm population. Both hybrid and pure particle swarm based training will be compared against standard algorithms like random sampling, the hill-climbing type stochastic method, and genetic algorithm (GA) based approach. Results will be shown on a subset of the NIST handwritten character set. This paper will describe in detail how particle swarm can be modeled to train an N-tuple classifier and how various parameters of the system should be chosen carefully to obtain an optimum set of tuples that achieves better recognition.

show abstract

Cross-Validation and Information Measures for Ram Based Neural Networks

Cited by 6 publications

References 1 publication

<title>Cross-validation techniques for n-tuple-based neural networks</title>

<title>Cross-validation techniques for n-tuple-based neural networks</title>

Hybridisation of GA and PSO to optimise N-tuples

Particle Swarm Intelligence to Optimize the Learning of N-tuples

Contact Info

Product

Resources

About