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.

Traffic accident duration prediction based on natural language processing and a hybrid neural network architecture

Abstract: Predicting the duration of traffic accidents can effectively help traffic management. To make a more accurate real-time prediction of traffic accident duration, and fully utilize the huge amount of traffic texts in social networks, in this paper, we consider this prediction task as a classification problem. First, the reported text of traffic accidents in social networks is obtained. After the data augmentation, the Bag-of-words model and Fisher optimal segmentation algorithm are combined to calculate the optimal classification threshold based on duration, and the accidents are classified into four classes. And then, the C-BiLSTM neural network is constructed by fusing convolutional neural network (CNN) and bidirectional long short term memory (Bi-LSTM) to predict the classes of accident durations, and the prediction accuracy of final trained model can reach 96.09%. Through experiments, the proposed method is proved to be practical and effective in solving traffic accident duration prediction.

Ill-posed Surface Emissivity Retrieval from Multi-Geometry Hyperspectral Images using a Hybrid Deep Neural Network.

Abstract: Atmospheric correction is a fundamental task in remote sensing because observations are taken either of the atmosphere or looking through the atmosphere. Atmospheric correction errors can significantly alter the spectral signature of the observations, and lead to invalid classifications or target detection. This is even more crucial when working with hyperspectral data, where a precise measurement of spectral properties is required. State-of-the-art physics-based atmospheric correction approaches require extensive prior knowledge about sensor characteristics, collection geometry, and environmental characteristics of the scene being collected. These approaches are computationally expensive, prone to inaccuracy due to lack of sufficient environmental and collection information, and often impossible for real-time applications. In this paper, a geometry-dependent hybrid neural network is proposed for automatic atmospheric correction using multi-scan hyperspectral data collected from different geometries. The proposed network can characterize the atmosphere without any additional meteorological data. A grid-search method is also proposed to solve the temperature emissivity separation problem. Results show that the proposed network has the capacity to accurately characterize the atmosphere and estimate target emissivity spectra with a Mean Absolute Error (MAE) under 0.02 for 29 different materials. This solution can lead to accurate atmospheric correction to improve target detection for real time applications.

A simple hierarchical approximation RBF neural network

Abstract: The approximation algorithm introduced by Asim Roy et al. (1997) generates a hybrid neural network with RBF neurons and other types of hidden neurons for function approximation. The network is trained in stages, with RBF neurons at the early stages corresponding to general features in the space and those in later stages corresponding to more specific features. The other types of hidden neurons are added with a view to improving generalization and reducing the number of RBF neurons. The algorithm uses linear programming to design and train the hybrid network. We investigate simplifying the algorithm with a view to eliminating the need for the other types of hidden neurons and linear programming. The simple hierarchical approximation algorithm ('SHA') achieves comparable results in terms of accuracy without the added complexity introduced by the other types of hidden neurons.

Optimization And Design Of Cooling SystemsUsing A Hybrid Neural Network And GeneticAlgorithm Methodology

Abstract: In this paper a novel method for the design and optimization of cooling systems is presented. The numerical solution of free convection from a heated horizontal cylinder confined between adiabatic walls obtained from a finite element solver is used to propose a non-linear heat transfer model of GMDH type approach. In the context of GMDH model, three different methods depending on the structure of neural network are implemented. The system of orthogonal equations is solved using a SVD scheme. The coefficients of second order polynomials are computed and their behavior is discussed. In addition, to demonstrate the performance of the predicted model, the numerical data are divided into trained and prediction data, respectively. The model is based on trained data and it is validated using the prediction data. In the next step, using the above-mentioned model and the genetic algorithms, the optimum coefficient of heat transfer is obtained. The results reveal the robustness and excellent performance of the hybrid procedure introduced in this paper.

A hybrid neural network method for simulating spatial variation in earthquake ground motion

Abstract: Different excitations for supports should be considered for the analysis of long-span structures. The excitation of each support has time delay and spatial variation relative to other support excitations. The present study aims to propose a new method for simulating accelerograms for various distances considering spatial variation of earthquake records. The accelerograms are simulated based on response or design spectra using the learning capabilities of neural networks. In this method, the response spectrum, and the distance parameter (distance from fault rupture) are the input, and the corresponding accelerograms are the output of the network. There are three stages involved in this study. In the first stage, a replicator neural network is used as a data compressor to increase capability of the simulation. In the second stage, a radial basis function neural network is employed to generate a compressed accelerogram for a certain distance and a response spectrum. In the third stage, the compressed...