Pseudoinverse Matrix Decomposition Based Incremental Extreme Learning Machine with Growth of Hidden Nodes

Kassani, Peyman Hosseinzadeh; Kim, Euntai

doi:10.5391/ijfis.2016.16.2.125

Cited by 13 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Over the last few years, machine learning methods have successfully solved the problem of finding hidden patterns or structures in data [ 13 , 14 ]. Several methods for ALL detection have been proposed in the literature.…”

Section: Introductionmentioning

confidence: 99%

An Aggregated-Based Deep Learning Method for Leukemic B-lymphoblast Classification

2020

View full text Add to dashboard Cite

Leukemia is a cancer of blood cells in the bone marrow that affects both children and adolescents. The rapid growth of unusual lymphocyte cells leads to bone marrow failure, which may slow down the production of new blood cells, and hence increases patient morbidity and mortality. Age is a crucial clinical factor in leukemia diagnosis, since if leukemia is diagnosed in the early stages, it is highly curable. Incidence is increasing globally, as around 412,000 people worldwide are likely to be diagnosed with some type of leukemia, of which acute lymphoblastic leukemia accounts for approximately 12% of all leukemia cases worldwide. Thus, the reliable and accurate detection of normal and malignant cells is of major interest. Automatic detection with computer-aided diagnosis (CAD) models can assist medics, and can be beneficial for the early detection of leukemia. In this paper, a single center study, we aimed to build an aggregated deep learning model for Leukemic B-lymphoblast classification. To make a reliable and accurate deep learner, data augmentation techniques were applied to tackle the limited dataset size, and a transfer learning strategy was employed to accelerate the learning process, and further improve the performance of the proposed network. The results show that our proposed approach was able to fuse features extracted from the best deep learning models, and outperformed individual networks with a test accuracy of 96.58% in Leukemic B-lymphoblast diagnosis.

show abstract

Section: Introductionmentioning

confidence: 99%

An Aggregated-Based Deep Learning Method for Leukemic B-lymphoblast Classification

2020

View full text Add to dashboard Cite

show abstract

“…The basic RVFLN algorithm inverse matrix can be calculated by the online method with the help of the Moore-Penrose pseudo inverse matrix method and it is good for online models, but it is not suitable for offline models. 33 Various adapted working conditions are presented in the online model, due to this cause, online sequential algorithm is preferred for the improvement of the accuracy. The proposed DC ring microgrid model produces huge data, and OS-RVFLN is more suitable 34 for demonstration.…”

Section: Os-rvfln Algorithm With Forgetting Factormentioning

confidence: 99%

Effective fault distance estimation and diagnosis in the PV based DC‐ring network using Hilbert Huang transform and weighted online sequential random vector functional network

Anjaiah

Dash

2021

Int Trans Electr Energ Syst

View full text Add to dashboard Cite

This paper proposes a novel approach for fault detection, classification, and distance estimation of five kinds of faults (ie, cable and Arc faults) in the PV-based DC-ring microgrid. The proposed approach is based on the Hilbert Huang Transform (HHT) and online sequential random vector functional link network (OS-RVFLN). Initially, differential current signals are passed through empirical mode decomposition (EMD) to yield intrinsic mode functions (IMFs). Among the series of IMFs, more efficient IMFs are selected based on the threshold of maximum weighted kurtosis index (M-WKI) value for fault diagnosis. The selected IMFs are passed through Hilbert transform (HT) for determining the instantaneous frequency, magnitude spectrum, and Hilbert energy. Then, the Teager energy operator (TEO) is applied to the more effective IMF among the selected IMFs of each DC fault for fault detection. The most suitable target features are extracted from the selected efficient IMFs. This feature set data is initialized to the proposed HHT-OSRVFLN algorithm for five kinds of faults classification and its location estimation. The performance of the proposed algorithm in the proposed microgrid is proved in terms of low fault detection time, superior classification accuracy, and low relative error through different case studies by consuming minimum computation time in MATLAB/Simulink environment. Finally, the digital architecture of the proposed method is designed and simulated in Xilinx ISE 14.5 software environment to estimate the fault location. The Chip scope hardware debugging output proves the correctness, simplicity, feasibility, practicability, and robustness of the proposed method for online fault diagnosis.List of Symbols and Abbreviations: R f , fault resistance; Idc 1 , Idc 2 , DC currents; Vdc 1 , Vdc 2 , DC voltages; α 1 , α 2 , attenuation constants; ωd 1 , ωd 2 , damping frequencies; ω n1 , ω n2 , natural frequencies; I 3 , differential current; I avg , average current; SD (k), squared difference; M e (t), mean envelope; E [C(t)], Teager energy of the signal; ω(t), instantaneous frequency; θ(t), phase angle; I b , best IMF; H, hidden layers; β 0 , output weight matrix; Z, output of the signal; d pfd , predicted fault distance; d actl , actual fault distance; t c , computation time; CC, correlation coefficient; EMD, empirical mode decomposition; FPGA, field programmable gate array; HHT, Hilbert Haung transform; ICON, integrator controller; ILA, integrated logic analyzer; IMF, intrinsic mode functions; KI, Kurtosis index; MAC, multiply and accumulate unit; OS-RVFLN, online sequential random vector functional link; TEO, Teager energy operator; VHDL, hardware description language; WKI, weighted kurtosis index.

show abstract

“…Various efforts have been made to explore the relations between the approximation ability and the number of nodes of some specific neural network, such as singlehidden-layer feedforward neural networks (SLFNs), and two-hidden-layer feedforward neural networks with specific or conditional activation functions [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]. For example, it was proved that N arbitrary distinct samples can be learned precisely by standard SLFNs with N hidden neurons (including biases) and the signum activation function in [12].…”

Section: Introductionmentioning

confidence: 99%

“…Later, Huang [17] proved that if the number of hidden nodes is equal to the number of distinct training samples, SLFNs with random input weight vectors and hidden biases can approximate the training samples with zero error. Furthermore, it was proved by [18][19][20] that for SLFNs, the approximation error is monotonically decreasing with gradually adding nodes in hidden layer.…”

Section: Introductionmentioning

confidence: 99%

Upper bounds on the node numbers of hidden layers in MLPs

Liu¹,

Feng²,

Du³

et al. 2021

NNW

View full text Add to dashboard Cite

It is one of the fundamental and challenging problems to determine the node numbers of hidden layers in neural networks. Various efforts have been made to study the relations between the approximation ability and the number of hidden nodes of some specific neural networks, such as single-hidden-layer and two-hiddenlayer feedforward neural networks with specific or conditional activation functions. However, for arbitrary feedforward neural networks, there are few theoretical results on such issues. This paper gives an upper bound on the node number of each hidden layer for the most general feedforward neural networks called multilayer perceptrons (MLP), from an algebraic point of view. First, we put forward the method of expansion linear spaces to investigate the algebraic structure and properties of the outputs of MLPs. Then it is proved that given k distinct training samples, for any MLP with k nodes in each hidden layer, if a certain optimization problem has solutions, the approximation error keeps invariant with adding nodes to hidden layers. Furthermore, it is shown that for any MLP whose activation function for the output layer is bounded on R, at most k hidden nodes in each hidden layer are needed to learn k training samples.

show abstract

Pseudoinverse Matrix Decomposition Based Incremental Extreme Learning Machine with Growth of Hidden Nodes

Cited by 13 publications

References 13 publications

An Aggregated-Based Deep Learning Method for Leukemic B-lymphoblast Classification

An Aggregated-Based Deep Learning Method for Leukemic B-lymphoblast Classification

Effective fault distance estimation and diagnosis in the PV based DC‐ring network using Hilbert Huang transform and weighted online sequential random vector functional network

Upper bounds on the node numbers of hidden layers in MLPs

Contact Info

Product

Resources

About