期刊
IFAC PAPERSONLINE
卷 51, 期 13, 页码 485-489出版社
ELSEVIER
DOI: 10.1016/j.ifacol.2018.07.326
关键词
neural networks; Black-Box identification; LSTM
Long-Short Term Memory (LSTM) is a type of Recurrent Neural Networks (RNN). It takes sequences of information and uses recurrent mechanisms and gate techniques. LSTM has many advantages over other feedforward and recurrent NNs in modeling of time series, such as audio and video. However, in non-linear system modeling normal LSTM does not work well(Wang, 2017). In this paper, we combine LSTM with NN, and use the advantages. The novel neural model consists of hierarchical recurrent networks and one multilayer perceptron. We design a special learning algorithm which uses backpropagation, and backpropagation through time methods. We use two non-linear system examples to compare our neural modeling with other well known methods. The results show that for the simulation model (only the test input is used and the past test output is not used), the modified LSTM model proposed in this paper is much better than the other existing neural models. (C) 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据