中国科学院机构知识库网格系统: 检索

Generalized Embedding Machines for Recommender Systems 期刊论文 OAI收割
Machine Intelligence Research, 2024, 卷号: 21, 期号: 3, 页码: 571-584
作者: Enneng Yang; Xin Xin; Li Shen; Yudong Luo; Guibing Guo

|

收藏 | 浏览/下载：8/0 | 提交时间：2024/05/23

Feature interactions, high-order interaction, factorization machine (FM), recommender system, graph neural network (GNN)

Isogeometric Convolution Hierarchical Deep-learning Neural Network: Isogeometric analysis with versatile adaptivity 期刊论文 OAI收割
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING, 2023, 卷号: 417, 页码: 46
作者: Zhang, Lei; Park, Chanwook; Lu, Ye; Li, Hengyang; Mojumder, Satyajit

|

收藏 | 浏览/下载：28/0 | 提交时间：2024/01/08

Convolution isogeometric analysis (C-IGA) Convolution hierarchical deep-learning neural network (C-hiDeNN) Software 2.0 r-h-p-s-a adaptive finite element method (FEM) High-order smoothness and convergence

面向农作物病害识别的高阶残差卷积神经网络研究 期刊论文 OAI收割
中国科学技术大学学报, 2019, 卷号: 49
作者: 曾伟辉; 李淼; 张健; 黄小平; 王敬贤

|

收藏 | 浏览/下载：28/0 | 提交时间：2020/10/26

crop disease recognition high-order residual(HOR) robustness convolutional neural network(CNN) 农作物病害识别高阶残差鲁棒性卷积神经网络

Study of experimental design and Response Surface method for surrogate model of computational simulation (EI CONFERENCE) 会议论文 OAI收割
2nd Annual Conference on Electrical and Control Engineering, ICECE 2011, September 16, 2011 - September 18, 2011, Yichang, China
Xi R.; Jia H.; Xiao Q.

收藏 | 浏览/下载：25/0 | 提交时间：2013/03/25

While the high-precision simulation is widely used in science and technology Design of Experiment (DOE) based on Response Surface (RS) method can be employed in surrogate model to reduce the cost and error. In order to illustrate the relationship between parameters and response features several DOE methods and Response Surface (RS) method are studied. The author used polynomial regression and RBF neural network based on orthogonal array to build a rocket aerodynamic discipline surrogate model respectively which proved their feasibility. From the results of the test case conclusion is drawn that characteristic as well as acclimatization of DOE methods and different approximation should be considered for different issues so the factors of cost and accuracy could reach a balance synthetically. 2011 IEEE.

Research on the identification for a nonlinear system (EI CONFERENCE) 会议论文 OAI收割
International Conference on Optical, Electronic Materials and Applications 2011, OEMA 2011, March 4, 2011 - March 6, 2011, Chongqing, China
作者: Liu J.; Jia P.; Liu J.; Liu J.

收藏 | 浏览/下载：37/0 | 提交时间：2013/03/25

The characteristic of the drift error of inertial platform is a high-order nonlinear dynamic system using the neural networks' abilities of universal approximation of differentiable trajectory and capturing system dynamic information this paper presents the drift error identifying project of inertial platform based on Elman networks structure. First the drift error model of inertial platform is established after selecting the input and output for network momentum and alterable speed algorithm is used to speed up the network convergence. On the basis of the algorithm the extended nonlinear node function in the hidden network does not only improve the learning speed of network but also satisfies the need of accuracy on system identification. Through the drift error data measured on inertial platform the training result shows that the scheme achieves satisfied identification results. (2011) Trans Tech Publications.

Recurrent high order neural network modeling for wastewater treatment process 期刊论文 OAI收割
Journal of Computers, 2011, 卷号: 6, 期号: 8, 页码: 1570-1577
作者: Qiao JF(乔俊飞); Yang WW(杨维维); Yuan MZ(苑明哲)

收藏 | 浏览/下载：19/0 | 提交时间：2017/03/07

wastewater treatment recurrent high order neural network filtering

The research of nonlinear control based on fuzzy neural network (EI CONFERENCE) 会议论文 OAI收割
International Conference on Electrical and Control Engineering, ICECE 2010, June 26, 2010 - June 28, 2010, Wuhan, China
Fan Y.-Y.; Sang Y.-J.

收藏 | 浏览/下载：33/0 | 提交时间：2013/03/25

This paper discussed and researched the structure and algorithm of fuzzy neural network controller based on the character of fuzzy logic and neural network theory. For the nonlinear system characteristics of uncertainty high order and hysteresis this paper used the fuzzy neural network technology to control nonlinear system and improved the control quality obviously. Take the single inverted pendulum for example the paper constructed the nonlinear mathematicmodel realized the control with the method of the adaptive fuzzy neural network and compared with control method of liner quadratic regulator the simulation results indicate that the method of adaptive fuzzy neural network can realize the stabilization of control better without the linear model of system and has a higher robustness. 2010 IEEE.

The costs prediction of AOD furnace based on improved RBF neural network (EI CONFERENCE) 会议论文 OAI收割
2010 International Conference on Computer, Mechatronics, Control and Electronic Engineering, CMCE 2010, August 24, 2010 - August 26, 2010, Changchun, China
Na T.; Zhang D.-J.; Hui L.

收藏 | 浏览/下载：18/0 | 提交时间：2013/03/25

In order to predict the cost a model of cost prediction was set up based on adaptive hierarchical genetic algorithm and RBF neural network. Hierarchical genetic algorithm could optimize the topology and the parameters simultaneously. Compared with simple genetic algorithm it has more efficiency in not only accelerating and stabilizing the parameters training but also determining the structure of the network. Adaptive crossover and mutation probability could accelerate the speed and avoid prematurity. The model was tested by five samples. The results showed that the prediction model has high prediction accuracy which indicated that it was applicable to predict the cost by the model. 2010 IEEE.

Lossless wavelet compression on medical image (EI CONFERENCE) 会议论文 OAI收割
4th International Conference on Photonics and Imaging in Biology and Medicine, September 3, 2005 - September 6, 2005, Tianjin, China
作者: Liu H.; Liu H.; Liu H.

收藏 | 浏览/下载：53/0 | 提交时间：2013/03/25

An increasing number of medical imagery is created directly in digital form. Such as Clinical image Archiving and Communication Systems (PACS). as well as telemedicine networks require the storage and transmission of this huge amount of medical image data. Efficient compression of these data is crucial. Several lossless and lossy techniques for the compression of the data have been proposed. Lossless techniques allow exact reconstruction of the original imagery while lossy techniques aim to achieve high compression ratios by allowing some acceptable degradation in the image. Lossless compression does not degrade the image thus facilitating accurate diagnosis of course at the expense of higher bit rates i.e. lower compression ratios. Various methods both for lossy (irreversible) and lossless (reversible) image compression are proposed in the literature. The recent advances in the lossy compression techniques include different methods such as vector quantization wavelet coding neural networks and fractal coding. Although these methods can achieve high compression ratios (of the order 50:1 or even more) they do not allow reconstructing exactly the original version of the input data. Lossless compression techniques permit the perfect reconstruction of the original image but the achievable compression ratios are only of the order 2:1 up to 4:1. In our paper we use a kind of lifting scheme to generate truly loss-less non-linear integer-to-integer wavelet transforms. At the same time we exploit the coding algorithm producing an embedded code has the property that the bits in the bit stream are generated in order of importance so that all the low rate codes are included at the beginning of the bit stream. Typically the encoding process stops when the target bit rate is met. Similarly the decoder can interrupt the decoding process at any point in the bil stream and still reconstruct the image. Therefore a compression scheme generating an embedded code can start sending over the network the coarser version of the image first and continues with the progressive transmission of the refinement details. Experimental results show that our method can get a perfect performance in compression ratio and reconstructive image.

Multiresolution locally expanded HONN for handwritten numeral recognition 期刊论文 OAI收割
PATTERN RECOGNITION LETTERS, 1997, 卷号: 18, 期号: 10, 页码: 1019-1025
作者: Liu, CL; Kim, JH; Dai, RW

收藏 | 浏览/下载：14/0 | 提交时间：2015/11/08

handwritten numeral recognition high order neural network local expansion multiresolution representation