Topic
Hybrid neural network
About: Hybrid neural network is a research topic. Over the lifetime, 1305 publications have been published within this topic receiving 18223 citations.
Papers published on a yearly basis
Papers
More filters
••
01 Jan 1994TL;DR: Expert systems perform reasoning using pre-established rules for a well-defined and narrow domain and combine knowledge bases of rules and domain-specific facts with information from clients or users about specific instances of problems in the knowledge domains of the expert systems.
Abstract: Expert systems perform reasoning using pre-established rules for a well-defined and narrow domain. They combine knowledge bases of rules and domain-specific facts with information from clients or users about specific instances of problems in the knowledge domains of the expert systems. Ideally, reasoning can be explained and the knowledge bases easily modified, independent of the inference engine, as new rules become know.
3 citations
••
TL;DR: A novel neural network which could resolve the constraints of the finite response time and hologram erasure on the convergence property of the photorefractive perceptron learning are discussed and experimental results of image classification are presented.
3 citations
••
TL;DR: The experimental results illustrate that the proposed approach is robust, and has the advantages over existing algorithms in calibration precision, and orthogonality of rotational matrix, in particular the precision of intrinsic and extrinsic parameters of camera, which provides a practical scheme for calibrating camera with radial distortion model.
Abstract: In order to calibrate camera with radial distortion model, a novel approach based on the hybrid neural network with rotational weight matrix and self-adaptive genetic-annealing algorithm is proposed. Firstly two sorts of neural networks are structured, whose weights are corresponding to the camera's extrinsic parameters and intrinsic parameters without and with radial distortion, so the structured neural networks coincide with the camera's pin hole model and radial distortion model respectively. And the performance index is obtained from the square of 2-norm of the difference between the vector consisted of network's outputs and the tested retinal coordinates of corresponding feature points projected in image planes. At the same time, a genetic-annealing algorithm is introduced into the solving-programming, where the probabilities of crossover and mutation are tuned according to the distance density of individuals, and unequal probability matching strategy is adopted. Thus while the system come to the equilibrium, the camera's parameters with radial distortion are obtained in the light of network's weights. The experimental results illustrate that the proposed approach is robust, and has the advantages over existing algorithms in calibration precision, and orthogonality of rotational matrix, in particular the precision of intrinsic and extrinsic parameters of camera, which provides a practical scheme for calibrating camera with radial distortion model.
3 citations
••
TL;DR: This paper defines the DA task on multiple round conversations between humans, and proposes a hybrid neural network-based ensemble model for solving this problem, which can achieve state-of-the-art accuracy on the experimental dialogue corpus.
Abstract: In dialogue systems, understanding the user utterances is crucial for providing appropriate responses. A traditional dialogue act classification (DA) task is to classify each user reply into “ACCEPT, REJECT, PROPOSE, and others”. In contrast, in this paper, we define the DA task on multiple round conversations between humans. The re-defined task is to classify a full dialogue according to the intention of one participant. We term this task as intention classification (IC). We, then, propose a hybrid neural network-based ensemble model for solving this problem. Two novel ensemble schemes are introduced for combining the classification results or features from various classifiers. One is ensembling features from each individual classifier using stacking, and we term this scheme as SFE. The other is adding wrong examples' weight to loss functions of each individual classifier using the AdaBoost scheme, and we term this scheme as MN-Ada. We have empirically examined the performance of the proposed ensemble schemes by using three popular deep neural networks, as well as one newly modified networks for IC. Extensive experiments have been conducted on a Chinese dialogue corpus. Our model can achieve state-of-the-art accuracy on the experimental dialogue corpus.
3 citations
••
21 Jun 2006TL;DR: The results indicated that both parallel and serial hybrid neural networks can model for complicated systems well and transfer the solution of nonlinear control strategy into solving for linear systems based on the decomposed models.
Abstract: Based on the "divide and rule" idea, hybrid neural networks (HNNs), which consisted of linear dynamic neural network and nonlinear static neural network, was used to model for complicated nonlinear systems. By using hybrid neural networks, it can reduce the degree of difficulty for training a single network, e.g. long training time and lower accuracy; and also can transfer the solution of nonlinear control strategy into solving for linear systems based on the decomposed models. An industrial polymerization process was introduced as a powerful case-study for the demonstration of potential of neural modeling. Nonlinear predicative models, based on both serial and parallel neural networks, were applied to predict the dynamic viscosity of PET. And the results indicated that both parallel and serial hybrid neural networks can model for complicated systems well
3 citations