Minimal distance neural methods

Dich, W.; Grudziński, Karol; Diercksen, Geerd H. F.

doi:10.1109/ijcnn.1998.685962

Cited by 11 publications

(11 citation statements)

References 11 publications

(15 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This framework includes typical feedforward neural network models (MLP, RBF, SBF) , some novel networks (Distance-Based Multilayer Perceptrons (D-MLPs, [41]) and the nearest neighbor or minimum-distance networks [33,43]), as well as many variants of the nearest neighbor methods, improving upon the traditional approach by providing more flexible decision borders. This framework has been designed to enable meta-learning based on a search in the space of all possible models that may be systematically constructed.…”

Section: Similarity-based Framework For Meta-learningmentioning

confidence: 99%

Towards Comprehensive Foundations of Computational Intelligence

Duch

2007

Studies in Computational Intelligence

View full text Add to dashboard Cite

Abstract. Although computational intelligence (CI) covers a vast variety of different methods it still lacks an integrative theory. Several proposals for CI foundations are discussed: computing and cognition as compression, meta-learning as search in the space of data models, (dis)similarity based methods providing a framework for such meta-learning, and a more general approach based on chains of transformations. Many useful transformations that extract information from features are discussed. Heterogeneous adaptive systems are presented as particular example of transformation-based systems, and the goal of learning is redefined to facilitate creation of simpler data models. The need to understand data structures leads to techniques for logical and prototype-based rule extraction, and to generation of multiple alternative models, while the need to increase predictive power of adaptive models leads to committees of competent models. Learning from partial observations is a natural extension towards reasoning based on perceptions, and an approach to intuitive solving of such problems is presented. Throughout the paper neurocognitive inspirations are frequently used and are especially important in modeling of the higher cognitive functions. Promising directions such as liquid and laminar computing are identified and many open problems presented.

show abstract

Section: Similarity-based Framework For Meta-learningmentioning

confidence: 99%

Towards Comprehensive Foundations of Computational Intelligence

Duch

2007

Studies in Computational Intelligence

View full text Add to dashboard Cite

show abstract

“…values of all parameters and procedures employed. A general similarity-based model [6] of an adaptive system used for classification may include various types of parameters and procedures, such as: the set {R j } of reference vectors created from the set of training vectors {X i } by selection and/or optimization procedure; a similarity function D(·) (frequently a distance function or an activation function in neural networks) parameterized in various ways, or a table used to compute similarities for nominal attributes; a weighting function G(D(X, R)), estimating contribution of the reference vector R to the classification probability depending on its similarity to the vector X; the total cost function E[·] optimized during training. The cost function may include regularization terms, hints [3], it may depend upon a kernel function K(·), scaling the influence of the error, for a given training example, on the total cost, it may use a risk matrix R (C i |C j ) of assigning wrong classes or a matrix S(C i |C j ) measuring similarity of output classes.…”

Section: Distance Functions In Neural Networkmentioning

confidence: 99%

“…We will place our considerations in the general framework for similarity-based classification methods (SBMs) presented recently [6], [7]. Investigation of connections between neural network and similarity-based methods leads to a number of new neural network models.…”

Section: Introductionmentioning

confidence: 99%

Neural networks in non-Euclidean metric spaces

Duch

Adamczak²

IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)

View full text Add to dashboard Cite

show abstract

“…A general framework for similarity-based methods has been presented recently [3], [4], [5]. Except for classical minimal distance methods, such as the k-NN method, many popular neural network models, such as the Radial Basis Functions (RBFs), Multilayer Perceptrons (MLPs), Learning Vector Quantization (LVQ), may be presented in this form.…”

Section: Introductionmentioning

confidence: 99%

“…calculation of optimal weighting factors s i in the metric function Eq. (4). If the cost function is taken as the number of classification errors (i.e.…”

mentioning

confidence: 99%