http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Comparative analysis of two recurrent neural networks for predicting the Lorenz system
Dmitriy Mikhailenko,Andrey Gavrilov,Artem Lenskiy 한국정보통신학회 2017 2016 INTERNATIONAL CONFERENCE Vol.9 No.1
In this paper, a brief comparison of two recurrent neural networks (RNN) is presented. We compare a single hidden layer network with lateral connection trained by backpropagation algorithm and the echo state network(ESN). Both networks have similar structures, yet different training algorithms. We focus on predicting the solution of Lorenz dynamical system. In the simulation, in order to make sure that both networks have equal number of trainable weights, the number of nodes was set different for each of two networks. As expected RNN with back-propagation outperforms ESN, however, the backpropagation algorithm is more computationally demanding and its convergence is not guaranteed, hence, ESN is appropriate choice as a first approximation in predicting dynamics systems.
Data Clustering Using Hybrid Neural Network
( Donghai Guan ),( Andrey Gavrilov ),( Weiwei Yuan ),( Sungyoung Lee ),( Young-koo Lee ) 한국정보처리학회 2007 한국정보처리학회 학술대회논문집 Vol.14 No.1
Clustering plays an indispensable role for data analysis. Many clustering algorithms have been developed. However, most of them suffer poor performance of learning. To archive good clustering performance, we develop a hybrid neural network model. It is the combination of Multi-Layer Perceptron (MLP) and Adaptive Resonance Theory 2 (ART2). It inherits two distinct advantages of stability and plasticity from ART2. Meanwhile, by combining the merits of MLP, it improves the performance for clustering. Experiment results show that our model can be used for clustering with promising performance.