A novel class of neural networks with quadratic junctions

DeClaris, N.; Su, Mu-Chun

doi:10.1109/icsmc.1991.169910

Cited by 20 publications

(7 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In fact, the producing of postsynaptic potential (PSP) in biological neurons is, most likely, a nonlinear process. This motivated us to build a new model of neurons that make good use of this nonlinearity [12]. In fact, the use of quadratic neural-type junctions was also inspired by intuitive concepts of cluster analysis [13].…”

Section: Brief Review Of Neurons With Quadratic Neural-type Junctionsmentioning

confidence: 99%

A Neural Tree with Partial Incremental Learning Capability

2007

2007 International Conference on Machine Learning and Cybernetics

View full text Add to dashboard Cite

This paper presents a new approach to constructing a neural tree with partial incremental learning capability. The proposed neural tree, called a quadratic-neuron-based neural tree (QUANT), is a tree structured neural network composed of neurons with quadratic neural-type junctions for pattern classification. The proposed QUANT integrates the advantages of decision trees and neural networks. Via a batch-mode training algorithm, the QUANT grows a neural tree containing quadratic neurons in its nodes. These quadratic neurons recursively partition the feature space into hyper-ellipsoidal-shaped sub-regions. The QUANT has the partial incremental capability so that it does not need to re-construct a new neural tree to accommodate new training data whenever new data are introduced to a trained QUANT. To demonstrate the performance of the proposed QUANT, several pattern recognition problems were tested.

show abstract

Section: Brief Review Of Neurons With Quadratic Neural-type Junctionsmentioning

confidence: 99%

A Neural Tree with Partial Incremental Learning Capability

2007

2007 International Conference on Machine Learning and Cybernetics

View full text Add to dashboard Cite

show abstract

“…This limitation has prompted researchers to experiment with neuron models with altered node connection functions. For instance, Yau and Manry [12], DeClaris [2], and Kavuri and Venkatasubramanian [5] report good results using neuron models that employ quadratic or ellipsoidal node connection functions. Lane et al [6] propose a neural architecture in which B-spline receptive fields are embedded in the node connection functions.…”

Section: Introductionmentioning

confidence: 94%

Generating nonlinear class boundaries using distributed encodings

Narayan

1995

Proceedings of the 33rd Annual on Southeast Regional Conference - ACM-SE 33

View full text Add to dashboard Cite

Multi-layer Perceptron (MLP) networks function as hyperplane classifiers when applied to classification problems. Therefore, MLP networks can be inefficient when applied to problems in which class boundaries are inadequately modeled by hyperplanes. Attempts to remedy this problem typically necessitate the introduction of a new neural network model in which alternative node connection functions are used to allow the formation of nonlinear class boundaries. In this paper we demonstrate the use of a biologically motivated data representation scheme which, while working within the constraints of the MLP model, permits the development of nonlinear class boundaries. The enhancements afforded by the data representation scheme are demonstrated and analyzed in the context of a classification problem.

show abstract

“…(1) which performs linear transformation of input data; the other one comprising of quadratic junctions given by equation. (2) which produces an adaptive hyper-ellipsoid. These types of quadratic junctions are capable of achieving hyper-ellipsoidal discriminants that can be varied in sizes and in locations.…”

Section: Quadratic Neuronsmentioning

confidence: 99%

“…Linear junctions are not good enough to represent the real process happening among neurons. Hence a new model of neurons that makes use of quadratic neural-type junctions [2] was developed. The use of quadratic neural-type junctions was also inspired by intuitive concepts of cluster analysis.…”

Section: Introductionmentioning

confidence: 99%

Classification and Clustering of Breast Cancer Images

Chandra

Nath

Malhothra³

2006

The 2006 IEEE International Joint Conference on Neural Network Proceedings

View full text Add to dashboard Cite

Medical image analysis is an important part of image analysis applications. Cluster analysis can be effectively used in pattern recognition and image processing. This work presents a novel way of classification and clustering medical image data using Self Organizing Neural networks with Quadratic neural type junctions. The medical image data comprised of breast image data which included images of both normal breasts and malignant breasts A large number of patterns comprising of left and right breast images were used for clustering and classification. It was observed that the Self organizing maps with quadratic neural type junctions is an effective tool for classification and clustering of medical images.

show abstract

A novel class of neural networks with quadratic junctions

Abstract: Abstracr-This paper discusses the architecture and training properties of a multilayer feedforward neural network class that uses quadratic junctions. Complexity and learning aspests of this class are examined and compared with traditional networks that use linear junctions.

Cited by 20 publications

References 1 publication

A Neural Tree with Partial Incremental Learning Capability

A Neural Tree with Partial Incremental Learning Capability

Generating nonlinear class boundaries using distributed encodings

Classification and Clustering of Breast Cancer Images

Contact Info

Product

Resources

About