Hierarchical ensemble of Extreme Learning Machine

Liu

IEEE Trans. Geosci. Remote Sensing

2020

Self Cite

169

Hyperspectral image (HSI) consists of hundreds of continuous narrow bands with high spectral correlation, which would lead to the so-called Hughes phenomenon and the high computational cost in processing. Band selection has been proven effective in avoiding such problems by removing the redundant bands. However, many of existing band selection methods separately estimate the significance for every single band and cannot fully consider the nonlinear and global interaction between spectral bands. In this paper, by assuming that a complete HSI can be reconstructed from its few informative bands, we propose a general band selection framework, Band Selection Network (termed as BS-Net). The framework consists of a band attention module (BAM), which aims to explicitly model the nonlinear inter-dependencies between spectral bands, and a reconstruction network (RecNet), which is used to restore the original HSI cube from the learned informative bands, resulting in a flexible architecture. The resulting framework is end-to-end trainable, making it easier to train from scratch and to combine with existing networks. We implement two BS-Nets respectively using fully connected networks (BS-Net-FC) and convolutional neural networks (BS-Net-Conv), and compare the results with many existing band selection approaches for three real hyperspectral images, demonstrating that the proposed BS-Nets can accurately select informative band subset with less redundancy and achieve significantly better classification performance with an acceptable time cost.

Section: B Convolutional Neural Networkmentioning

confidence: 99%

BS-Nets: An End-to-End Framework for Band Selection of Hyperspectral Image

Liu

IEEE Trans. Geosci. Remote Sensing

2020

Self Cite

169

Mathematical Problems in Engineering

“…ELM was first proposed by Huang et al [45], which was developed for single hidden layer feedforward networks (SLFNs). By comparing with traditional neural networks, it requires great efforts in the adjustment of hyperparameter [46], ELM can provide good generalization ability and extremely fast learning speed. ELM contains input, hidden layers, and output nodes, and only hidden layer nodes required to be set in ELM.…”

Section: Extreme Learning Machine (Elm)mentioning

confidence: 99%

A Novel Key Influencing Factors Selection Approach of P2P Lending Investment Risk

Xia

Zhu

et al. 2019

Recent frequent “thunderstorm incidents” of P2P lending industry have caused the panic of industry investors. To predict the investment risk of P2P lending, we should scientifically and rationally analyze the key influencing factors of P2P lending investment risk. Existing key influencing factors selection methods mainly involve traditional statistical approaches and artificial intelligence methods. The traditional statistical approaches cannot deal with the high-dimensional nonlinear problems, and it cannot find the exact key influencing factors of the P2P lending investment risk. The artificial intelligence methods cannot recognize and learn the application background, and the selected attributes without active thinking and personal perception may not be the key influencing factors of P2P lending investment risk. To address the above issues, a novel key influencing factors selection approach of P2P lending investment risk is proposed by combining the proposed fireworks coevolution binary glowworm swarm optimization (FCBGSO), multifractal dimension (MFD), probit regression, and artificial prior knowledge. First, multifractal dimension combined with the proposed FCBGSO is used to select the preliminary influencing factors of the investment risk; second, the nonsignificant relevant attributes in the preliminary influencing factors are removed using the probit regression, and we add the influencing factors extracted from the original dataset of P2P lending using the artificial prior knowledge into the retaining influencing factors after removing one by one. A small and reasonable number of influencing factor subsets are achieved. Finally, we evaluate each influencing factors subset using extreme learning machine (ELM), and the subset with the best classification accuracy is efficiently achieved, i.e., it is the key influencing factors of P2P lending investment risk. Experimental results on the real P2P lending dataset from the Renrendai platform demonstrate that the proposed approach performs better than other state-of-the-art methods and that it has validity and effectiveness. It provides a new research idea for the key influencing factors selection of P2P lending investment risk.

“…The clustering-based BS method enables the intra-cluster variation of spectral bands to be considered, however, there is often a need for each band cluster to do a second selection. The searching-based method treats BS of HSI as a combinational optimization problem, which usually can be solved by heuristic searching methods, such as evolution algorithms [5], [18], [19]. Theoretically, although this type of BS method can select the global optimal band subset, the heuristic searching is often time-consuming.…”

Section: Introductionmentioning

confidence: 99%

Efficient Graph Convolutional Self-Representation for Band Selection of Hyperspectral Image

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Zhang

Liu

et al. 2020

Self Cite

Hyperspectral image (HSI) band selection (BS) is an important task for HSI dimensionality reduction, whose goal is to select an informative band subset containing less redundancy. However, traditional BS methods basically work in the Euclidean domain and thus often neglect to consider the structural information of spectral bands. In this paper, to make full use of the structural information, a novel BS method termed as Efficient Graph Convolutional Self-Representation (EGCSR) is proposed by incorporating graph convolution into the selfrepresentation model. Since the proposed method is typically modeled in the non-Euclidean domain, it tends to result in a more robust self-representation coefficient matrix. We provide a closed-form solution to the EGCSR model, which leads to high computational efficiency. We further propose two strategies to determine the informative band subset from the coefficient matrix. The first is a ranking-based strategy (EGCSR-R), which ranks every band by calculating the cumulative contribution, and the second is a clustering-based strategy (EGCSR-C), which treats band selection as a band clustering task based on using subspace segmentation. Extensive experimental results on three real HSI datasets show that the proposed EGCSR model is dramatically superior to many existing band selection methods and with high computational efficiency.