Special Issue on Extreme Learning Machines for Pattern Classification and System Modelling
摘要截稿:
全文截稿: 2019-04-15
影响因子: 2.663
期刊难度:
CCF分类: 无
中科院JCR分区:
• 大类 : 计算机科学 - 3区
• 小类 : 计算机:硬件 - 3区
• 小类 : 计算机:跨学科应用 - 3区
• 小类 : 工程:电子与电气 - 3区
Overview
Extreme Learning Machine (ELM), as an effective training methodology for feed-forward neural networks (FNN), has been widely used to perform the universal approximation and parallel processing in science and engineering. In recent years, ELM’s applications have spread out from manufacturing, transportation, process control, dynamic system modelling, digital signal and image processing to information retrieval with vast amount of data. When a single hidden layered feed-forward neural network (SLFNN) is trained with ELM, the input weights are uniformly randomly selected in a range, and the output weights are then optimally designed by using the batch learning type of least squares or other regularization methods. In such a way, the input data are mapped to the feature space at the outputs of the hidden layer. In particular, when the number of the hidden nodes is greater than the number of input patterns, the data features are sparsely distributed in the feature space. This remarkable characteristic of SLFNNs trained with ELM makes it possible to linearly separate the features of the input data in the feature space, which may be nonlinearly separable in the input space. In addition, many results in signal/image processing and big data analysis have shown that the SLFNNs trained with ELM have excellent capabilities of interpolation, universal approximation and generalization, and has actually filled the gap between machine learning and biological learning. More recently, ELM has combined with deep-learning, Bayesian belief and fuzzy logic to exhibit a powerful role as biological learning in the area of artificial intelligence.
Authors are invited to submit their original research manuscripts with new findings in ELM theories and applications to pattern classification, system modelling and big data analysis to this special issue.
Topics:
The topics of interest are:
• Clustering, feature extraction and pattern classification
• Combination of deep learning and ELM
• Combination of Bayesian belief and ELM
• Financial modelling and data analysis
• Engineering system modelling
• Human computer interface and brain computer interface