In this paper Lyapunov Diagonally Stable matrices are used to design Cellular neuralnetworks for associative memories. The proposed technique, which guarantees the global asymptotic stability of the equilibrium point...
详细信息
ISBN:
(纸本)0780348672
In this paper Lyapunov Diagonally Stable matrices are used to design Cellular neuralnetworks for associative memories. The proposed technique, which guarantees the global asymptotic stability of the equilibrium point, generates neural circuits where the input data are fed via external inputs, rather than initial conditions. This feature makes the suggested approach particularly suitable for hardware implementation techniques. Simulations results are reported to show the advantages and the usefulness of the proposed design method.
Insight into the core of the Pipelined Recurrent neural Network (PRNN) in prediction applications is provided. It is shown that modules of the PRNN contribute to the final predicted value at the output of the PRNN in ...
详细信息
Insight into the core of the Pipelined Recurrent neural Network (PRNN) in prediction applications is provided. It is shown that modules of the PRNN contribute to the final predicted value at the output of the PRNN in two ways, namely through the process of nesting, and through the process of learning. A measure of the influence of the output of a distant module to the amplitude at the output of the PRNN is analytically found, and the upper bound for it is derived. Furthermore, an analysis of the influence of the forgetting factor in the cost function of the PRNN to the process of learning is undertaken, and it is found that for the PRNN, the forgetting factor can even exceed unity in order to obtain the best predictor. Simulations on three speech signals support that approach, and outperform the other stochastic gradient based schemes.
Most imageprocessing tasks like pattern matching are defined in terms of large-neighborhood DTCNN templates, while most hardware implementations support only direct-neighborhood ones (3x3). Literature on DTCNN templa...
详细信息
ISBN:
(纸本)0780348672
Most imageprocessing tasks like pattern matching are defined in terms of large-neighborhood DTCNN templates, while most hardware implementations support only direct-neighborhood ones (3x3). Literature on DTCNN template decomposition shows that such large-neighborhood functions can be implemented as a sequence of successive direct-neighborhood templates, However, for this procedure the number of templates in the decomposition is exponential in the size of the original template. This paper shows how template decomposition is induced by the decomposition of structuring elements in the morphological design process. Ir is proved that an upper bound for the number of templates found this way is quadratic in the size of the original template. For many cases more efficient and even optimal decompositions can be obtained.
There has been a lot of interest in adaptive symbolic and neural agents for different tasks, for instance speech/language integration and image/text integration in various multimedia applications. Hybrid neural symbol...
There has been a lot of interest in adaptive symbolic and neural agents for different tasks, for instance speech/language integration and image/text integration in various multimedia applications. Hybrid neural symbolic methods have been shown to be able to reach a level where they can actually be further developed in real-world scenarios. A combination of symbolic and neural agents is possible in various neural symbolic processing architectures, which contain both symbolic and neural agents appropriate for to a specific task, e.g. integrating speech, text and images for multimedia. We concentrate on general principles of neural and hybrid architectures for multimedia in general. From the perspective of knowledge engineering, hybrid symbolic/neural agents are advantageous since different mutually complementary properties can be combined. Symbolic representations have advantages with respect to easy interpretation, explicit control, fast initial coding, dynamic variable binding and knowledge abstraction. On the other hand, neural agents show advantages for gradual analog plausibility, learning, robust fault-tolerant processing, and generalization to similar input. Since these advantages are mutually complementary, a hybrid symbolic neural architecture can be useful if different processing strategies have to be supported.
3D reconstruction of human body parts, and faces in particular, is catalysing growing interest in many disciplines ranging from basic imageprocessing to video conferencing, constructive and plastic surgery, rehabilit...
3D reconstruction of human body parts, and faces in particular, is catalysing growing interest in many disciplines ranging from basic imageprocessing to video conferencing, constructive and plastic surgery, rehabilitation and virtual clones. A host of devices (3D scanners), which provide these 3D models, have come to the market in the last few years. They are based on sampling a large number of 3D data points over the surface and of fitting a suitable analytical model to them. There are two main problems which have to be faced: filtering of the noise associated to sampling and interpolation between the samples. These two problems can be reframed in the domain of regularisation. It is shown how a regularised model can be efficiently obtained by using a new neural network called hierarchical radial basis function network (HRBF).
EEG events are widely used to diagnose patients who suffer from different diseases, including epilepsy. The EEG during a seizure exhibits characteristic temporal and spectral properties, depending upon the seizure typ...
EEG events are widely used to diagnose patients who suffer from different diseases, including epilepsy. The EEG during a seizure exhibits characteristic temporal and spectral properties, depending upon the seizure type and its cause. Identifying an EEG with an ictal event of this nature can help to support diagnosis and may also be used to classify the type of seizure. From this work, based on time-frequency analysis pre-processing of EEG seizures, we obtained some good results about the best resolution of frequency changes for feature extraction used for neural net input. Together with the other features (from the same data mining), the system performs a neural net-based and knowledge-based detection. There has been no such method reported previously in the literature about how to determine a signature for an EEG event.
Insight into the core of the pipelined recurrent neural network (PRNN) in prediction applications is provided. It is shown that modules of the PRNN contribute to the final predicted value at the output of the PRNN in ...
Insight into the core of the pipelined recurrent neural network (PRNN) in prediction applications is provided. It is shown that modules of the PRNN contribute to the final predicted value at the output of the PRNN in two ways, namely through the process of nesting, and through the process of learning. A measure of the influence of the output of a distant module to the amplitude at the output of the PRNN is analytically found, and the upper bound for it is derived. Furthermore, an analysis of the influence of the forgetting factor in the cost function of the PRNN to the process of learning is undertaken, and it is found that for the PRNN, the forgetting factor can even exceed unity in order to obtain the best predictor. Simulations on three speech signals support that approach, and outperform the other stochastic gradient based schemes.
A method far automating the detection of tubercle bacilli in sputum specimens is described. A fluorescence microscope with an attached digital camera is used to manually locate and capture images of tubercle bacilli. ...
A method far automating the detection of tubercle bacilli in sputum specimens is described. A fluorescence microscope with an attached digital camera is used to manually locate and capture images of tubercle bacilli. The method comprises two phases: (a) imageprocessing and analysis techniques are applied to the images for enhancement and feature extraction; (b) object recognition techniques are used for the automatic identification of tubercle bacilli in the images. The eventual implementation of the system will be semi-automatic, where the best candidate images containing bacilli are presented to the medical technologist together with a bacillus count, confidence measures and recommended diagnosis. The final diagnosis could be performed by the technologist in less than a minute for typical cases. Furthermore, the results should be more accurate due to the higher number of view-fields processed. The study presented in this paper indicates that machine-assisted diagnosis of tuberculosis is certainly feasible.
The proceedings contains 11 papers from the 1997 ieecolloquium on imageprocessing for Security applications. Topics discussed include: face recognition;statistical models;integrated person identification;speech reco...
详细信息
The proceedings contains 11 papers from the 1997 ieecolloquium on imageprocessing for Security applications. Topics discussed include: face recognition;statistical models;integrated person identification;speech recognition;balanced uncertainty wavelets;fingerprint compression;automatic signature verification;surveillance imaging;significance testing;image analysis;Bayesian networks;automatic human head location;crowd density estimation;and advanced visual surveillance.
neuralnetworks are ideally suited to the processing of noisy or uncertain data as they operate within a probabilistic framework. They produce probability estimates at their output and so allowance must-be made for th...
详细信息
neuralnetworks are ideally suited to the processing of noisy or uncertain data as they operate within a probabilistic framework. They produce probability estimates at their output and so allowance must-be made for this. This is a very important consideration in the context of industrial applications and the talk will illustrate how this issue was addressed in the Sharp LogiCook (a neural network microwave oven) and in Oxford Medical's QUESTAR (a neural network system for the analysis of sleep disorders).
暂无评论