Modular Neural Network Classifier Research Articles

Aircraft recognition using modular extreme learning machine

In this paper, a novel recognition scheme is proposed for identifying the aircrafts of different types based on multiple modular neural network classifiers. Three moment invariants including Hu moments, Zernike moments and Wavelet moments are extracted from the characteristics exhibited by aircrafts and used as the input variables of each modular neural network respectively. Each modular neural network consists of multiple single-hidden layer feedforward networks which are trained using the extreme learning machine and different clustering data subsets. A clustering and selection method is used to get the classification rate of each modular neural network and then based on their weighted sum the final classification output is obtained. The proposed recognition scheme is finally evaluated by recognizing six different types of aircraft models and the simulation results show the superiority of the proposed method compared with the single ELM classifier and other classification algorithms.

Feature Selection for Modular Neural Network Classifiers

Feature Selection for Modular Neural Network Classifiers

Feature-based decision aggregation in modular neural network classifiers

In several modular neural network (MNN) architectures, the individual decisions at the module level have to be integrated together using a voting scheme. All these voting schemes use the outputs of the individual modules to produce a global output without inferring explicit information from the problem feature space. This makes the choice of the aggregation procedure very subjective. In this work, a new MNN architecture will be presented. This architecture integrates learning into the voting scheme. We will be focusing on making the decision fusion a more dynamic process. In this context, dynamic means the aggregation procedure which has the flexibility to adapt to changes in the input. This approach requires the aggregation procedure to gather information about the input to help better understand how to dynamically aggregate decisions.

Automatic target recognition using a feature-decomposition and data-decomposition modular neural network

A modular neural network classifier has been applied to the problem of automatic target recognition using forward-looking infrared (FLIR) imagery. The classifier consists of several independently trained neural networks. Each neural network makes a decision based on local features extracted from a specific portion of a target image. The classification decisions of the individual networks are combined to determine the final classification. Experiments show that decomposition of the input features results in performance superior to a fully connected network in terms of both network complexity and probability of classification. Performance of the classifier is further improved by the use of multiresolution features and by the introduction of a higher level neural network on the top of the individual networks, a method known as stacked generalization. In addition to feature decomposition, we implemented a data-decomposition classifier network and demonstrated improved performance. Experimental results are reported on a large set of real FLIR images.

Modular Neural Network Classifiers: A Comparative Study

There is a wide variety of Modular Neural Network (MNN) classifiers in the literature. They differ according to the design of their architecture, task-decomposition scheme, learning procedure, and multi-module decision-making strategy. Meanwhile, there is a lack of comparative studies in the MNN literature. This paper compares ten MNN classifiers which give a good representation of design varieties, viz., Decoupled; Other-output; ART-BP; Hierarchical; Multiple-experts; Ensemble (majority vote); Ensemble (average vote); Merge-glue; Hierarchical Competitive Neural Net; and Cooperative Modular Neural Net. Two benchmark applications of different degree and nature of complexity are used for performance comparison, and the strength-points and drawbacks of the different networks are outlined. The aim is to help a potential user to choose an appropriate model according to the application in hand and the available computational resources.

Robust Classifiers without Robust Features

We develop a two-stage, modular neural network classifier and apply it to an automatic target recognition problem. The data are features extracted from infrared and TV images. We discuss the problem of robust classification in terms of a family of decision surfaces, the members of which are functions of a set of global variables. The global variables characterize how the feature space changes from one image to the next. We obtain rapid training times and robust classification with this modular neural network approach.