Modeling and optimal regulation of erythropoiesis subject to benzene intoxication

  • Received: 01 February 2004 Accepted: 29 June 2018 Published: 01 March 2004
  • MSC : 34K60, 34K35, 65M60, 92C37, 93C20.

  • Benzene (C6H6) is a highly flammable, colorless liquid. Ubiquitous exposures result from its presence in gasoline vapors, cigarette smoke, and industrial processes. Benzene increases the incidence of leukemia in humans when they are exposed to high doses for extended periods; however, leukemia risks in humans subjected to low exposures are uncertain. The exposure-dose- response relationship of benzene in humans is expected to be nonlinear because benzene undergoes a series of metabolic transformations, detoxifying and activating, resulting in various metabolites that exert toxic e ffects on the bone marrow.
        Since benzene is a known human leukemogen, the toxicity of benzene in the bone marrow is of most importance. And because blood cells are produced in the bone marrow, we investigated the eff ects of benzene on hematopoiesis (blood cell production and development). An age-structured model was used to examine the process of erythropoiesis, the development of red blood cells. This investigation proved the existence and uniqueness of the solution of the system of coupled partial and ordinary di fferential equations. In addition, we formulated an optimal control problem for the control of erythropoiesis and performed numerical simulations to compare the performance of the optimal feedback law and another feedback function based on the Hill function.

    Citation: H. T. Banks, Cammey E. Cole, Paul M. Schlosser, Hien T. Tran. Modeling and optimal regulation of erythropoiesis subject to benzene intoxication[J]. Mathematical Biosciences and Engineering, 2004, 1(1): 15-48. doi: 10.3934/mbe.2004.1.15

    Related Papers:

    [1] Chunmei He, Hongyu Kang, Tong Yao, Xiaorui Li . An effective classifier based on convolutional neural network and regularized extreme learning machine. Mathematical Biosciences and Engineering, 2019, 16(6): 8309-8321. doi: 10.3934/mbe.2019420
    [2] Mengya Zhang, Qing Wu, Zezhou Xu . Tuning extreme learning machine by an improved electromagnetism-like mechanism algorithm for classification problem. Mathematical Biosciences and Engineering, 2019, 16(5): 4692-4707. doi: 10.3934/mbe.2019235
    [3] Vinh Huy Chau . Powerlifting score prediction using a machine learning method. Mathematical Biosciences and Engineering, 2021, 18(2): 1040-1050. doi: 10.3934/mbe.2021056
    [4] Michele La Rocca, Cira Perna . Nonlinear autoregressive sieve bootstrap based on extreme learning machines. Mathematical Biosciences and Engineering, 2020, 17(1): 636-653. doi: 10.3934/mbe.2020033
    [5] Yunpeng Ma, Shilin Liu, Shan Gao, Chenheng Xu, Wenbo Guo . Optimizing boiler combustion parameters based on evolution teaching-learning-based optimization algorithm for reducing NOx emission concentration. Mathematical Biosciences and Engineering, 2023, 20(11): 20317-20344. doi: 10.3934/mbe.2023899
    [6] Ming Zhu, Kai Wu, Yuanzhen Zhou, Zeyu Wang, Junfeng Qiao, Yong Wang, Xing Fan, Yonghong Nong, Wenhua Zi . Prediction of cooling moisture content after cut tobacco drying process based on a particle swarm optimization-extreme learning machine algorithm. Mathematical Biosciences and Engineering, 2021, 18(3): 2496-2507. doi: 10.3934/mbe.2021127
    [7] Sungwon Kim, Meysam Alizamir, Youngmin Seo, Salim Heddam, Il-Moon Chung, Young-Oh Kim, Ozgur Kisi, Vijay P. Singh . Estimating the incubated river water quality indicator based on machine learning and deep learning paradigms: BOD5 Prediction. Mathematical Biosciences and Engineering, 2022, 19(12): 12744-12773. doi: 10.3934/mbe.2022595
    [8] Yufeng Qian . Exploration of machine algorithms based on deep learning model and feature extraction. Mathematical Biosciences and Engineering, 2021, 18(6): 7602-7618. doi: 10.3934/mbe.2021376
    [9] Wang Cai, Jianzhuang Wang, Longchao Cao, Gaoyang Mi, Leshi Shu, Qi Zhou, Ping Jiang . Predicting the weld width from high-speed successive images of the weld zone using different machine learning algorithms during laser welding. Mathematical Biosciences and Engineering, 2019, 16(5): 5595-5612. doi: 10.3934/mbe.2019278
    [10] Anastasia-Maria Leventi-Peetz, Kai Weber . Probabilistic machine learning for breast cancer classification. Mathematical Biosciences and Engineering, 2023, 20(1): 624-655. doi: 10.3934/mbe.2023029
  • Benzene (C6H6) is a highly flammable, colorless liquid. Ubiquitous exposures result from its presence in gasoline vapors, cigarette smoke, and industrial processes. Benzene increases the incidence of leukemia in humans when they are exposed to high doses for extended periods; however, leukemia risks in humans subjected to low exposures are uncertain. The exposure-dose- response relationship of benzene in humans is expected to be nonlinear because benzene undergoes a series of metabolic transformations, detoxifying and activating, resulting in various metabolites that exert toxic e ffects on the bone marrow.
        Since benzene is a known human leukemogen, the toxicity of benzene in the bone marrow is of most importance. And because blood cells are produced in the bone marrow, we investigated the eff ects of benzene on hematopoiesis (blood cell production and development). An age-structured model was used to examine the process of erythropoiesis, the development of red blood cells. This investigation proved the existence and uniqueness of the solution of the system of coupled partial and ordinary di fferential equations. In addition, we formulated an optimal control problem for the control of erythropoiesis and performed numerical simulations to compare the performance of the optimal feedback law and another feedback function based on the Hill function.


    Artificial Neural Networks (ANNs) simulate the process of human brain information processing through a large number of neurons that are interconnected in a certain way and efficient network learning algorithms [1]. Over the past few decades, ANNs have been widely used in various fields of human needs due to their powerful nonlinear mapping capabilities and parallel computing capabilities [2].

    So far, many neural network learning algorithms have been proposed and improved. Jian et al. summarized well-known learning algorithms among them [3]. Among them, the backpropagation algorithm is one of the most mature neural network learning algorithms, and it is also a famous representative of all iterative gradient descent algorithms for supervised learning in neural networks. It was first proposed by Paul Werbos in the 1970s [4] and widely used after it was rediscovered by Rumelhart and McClelland in 1986 [5]. However, the BP algorithm is not perfect, and there are still inevitable defects, such as easy to fall into a local minimum during the training process, and the convergence speed is slow. In response to these shortcomings, researchers have proposed many methods to improve the backpropagation technique [6]. For example, genetic algorithm is used to optimize the connection weights of BP network [7], add white noise to the weighted sum of BP [8], add momentum term, regularization operator and Adaboost integration algorithm [9], and dynamically change the learning rate according to the change of mean square error [10]. However, these problems of the BP algorithm still restrict its development in many application fields, especially since the learning speed cannot meet the actual needs. Recently, Huang et al. proposed extreme learning machine (ELM), which is a simple and efficient learning algorithm for single hidden layer feedforward neural network (SLFN) [11]. The core idea of the ELM algorithm is that the input weights and hidden layer biases of the network are randomly selected and kept unchanged during the training process, and then the output weights are directly obtained by Moore-Penrose generalized inverse operation. The advantages of ELM are: only the number of hidden layer neurons needs to be optimized, with less human intervention; it avoids the process of iterative optimization of parameters in traditional SLFN training algorithm, which greatly saves training time; the resulting solution is the only optimal solution, which guarantees the generalization performance of the network [12]. Therefore, ELM has been widely used in many fields such as disease diagnosis [13,14,15], traffic sign recognition [16,17], and prediction [18,19].

    In recent years, the significant advantages of ELM have attracted the attention of a large number of researchers in academia and industry, and the research on this algorithm and model has achieved fruitful results [20,21]. The learning process of the standard ELM algorithm can be considered to be based on empirical risk minimization, which tends to produce overfitted models. In addition, since ELM does not consider heteroskedasticity in practical applications, its generalization ability and robustness will be greatly affected when there are many outliers in the training samples. To effectively overcome the above problems, regularization methods are applied to ELM [22,23,24]. Lu et al. proposed a probabilistic RELM method to reduce the influence of noise data and outliers on the model [25]. Yıldırım and Revan Özkale combined the advantages of Liu estimator and Lasso regression method to deal with the shortcomings of traditional ELM instability and poor generalization [26]. Huang et al. [27] introduced the kernel function into ELM and proposed a general framework-KELM that can be used for regression, binary classification and multi-classification problems, which effectively improved the problem of generalization and stability degradation caused by random parameters. However, kernel selection is an important part of KELM, and it may not always be appropriate to choose empirically. Liu and Wang [28] proposed a multiple kernel extreme learning machine (MK-ELM) to solve this problem. However, MK-ELM cannot effectively handle large-scale datasets, because it needs to optimize more parameters, resulting in high computational complexity of the algorithm. To meet the needs of online real-time applications, online sequential extreme learning machine algorithms (OS-ELM) have been proposed [29], and OS-ELM was improved [30,31,32]. Online sequential class-specific extreme learning machine (OSCSELM) supports online learning techniques of both chunk-by-chunk and one-by-one learning modes, which is used to solve the class imbalance problem of small and large data sets [33]. Lu et al. used the OS-ELM training method of Kalman filter to improve its stability [34]. OS-ELM was extended to solve the problem of increasing classes [35].

    Before the training starts, the user needs to specify the number of hidden layer neurons for ELM. However, how to choose the appropriate number of hidden layer neurons for different applications has always been a difficult and hot topic in the field of neural network research. At present, the methods used by ELM to adjust the hidden layer structure mainly include swarm intelligence optimization [36,37,38,39,40], incremental method [41,42,43,44,45], pruning [46,47,48,49,50] and adaptive [51,52,53]. With the advent of the era of big data, storing and processing large-scale data has become an urgent need for enterprises, and the ensemble and parallelism of ELM have therefore become a research hotspot [54,55,56,57]. Lam and Wunsch learn features through unsupervised feature learning (UFL) algorithm, and then train features through fast radial basis function (RBF) extreme learning machine (ELM), which improves the accuracy and speed of the algorithm [58]. Yao and Ge proposed distributed parallel extreme learning machine (dp-ELM) and hierarchical extreme learning machine [59]. Duan et al. proposed an efficient ELM with three parallel sub-algorithms based on the Spark framework (SELM) for big data classification. [60]. Many researchers have turned their attention to deep ELM and conducted some innovative research works [61,62]. Dai et al. proposed multilayer one-class extreme learning machine (OC-ELM) [63]. Zhang et al. proposed multi-layer extreme learning machine (ML-ELM) [64]. Yahia et al. proposed a new structure based on extreme learning machine auto-encoder with deep learning structure and a composite wavelet activation function for hidden nodes [65].

    The inappropriate initial parameters of the hidden layer (input weights, hidden layer biases and the number of nodes) in the original ELM will lead to poor classification results of ELM [21], although the improved algorithms mentioned above for the original ELM improve its generalization performance, they greatly increase the computational complexity. Therefore, we need a network learning algorithm with fast learning speed and higher generalization performance.

    In this paper, we propose a new regression and classification model without iterative optimization parameters in the spirit of extreme learning, called functional extreme learning machine (FELM). FELM aims to use functional neurons (FNs) model as the basic units, and use functional equation-solving theory to guide the modeling process of functional extreme learning machine [66,67,68,69]. Like ELM, the FELM parameter matrix is obtained by solving the generalized inverse of the hidden layer neuron output matrix. FELM is a generalization of ELM. Its unique network structure and simple and efficient learning algorithm make it not only solve the problems that ELM can solve, but also solve many problems that ELM cannot solve. However, FELM is also different from ELM. The activation function of ELM is fixed, and ELM has weights and biases. The neuron function of FELM is not fixed, and there are no weights and biases, only parameters (coefficients), so it avoids the influence of random parameters (input weights, hidden layer biases) on the generalization performance and stability of ELM model. Its neuron functions are linear combinations of given basic functions, and the basic functions are selected according to the problem to be solved without specifying the number. The learning essence of FELM is the learning of parameters, and the parameter learning algorithm proposed in this paper does not need iterative calculation and has high accuracy. FELM is compared with other popular technologies in terms of generalization performance and training time on several artificial datasets, benchmark regression and classification datasets. The results show that FELM is not only fast, but also has good generalization performance.

    The rest of this paper is organized as follows: Section 2 provides an overview of FN and ELM. In Section 3, the topology of functional extreme learning machine, the theory of structural simplification, and the parameter learning algorithm are described. Section 4 presents the performance comparison results of FELM, classical ELM, OP-ELM, classical SVM and LSSVM on regression and classification problems. Section 5 draws conclusions and discusses future research directions.

    Functional neuron model and extreme learning machine (ELM) will be briefly discussed in the following section.

    Functional neuron was proposed by Enrique Castillo [66]. Figure 1(a) is a functional neuron model, Figure 1(b) is its expansion model and Figure 1(c) is the expanded model for the green part of Figure 1(b). The mathematical expression of the functional neuron is:

    O=f(X) (1)
    Figure 1.  (a) Functional neuron model. (b) Expansion model of functional neuron. (c) The expansion of the green part in (b).

    where, X={x1,x2,...,xm}, O={o1,o2,...,om}, f() is functional neuron function, X and O are the input and output of functional neuron function, respectively. Functional neuron function can be expressed by a linear combination of basic functions:

    f(X)=nj=1aijφij(X)=aTiφi(X) (2)

    where, {φi(X)=(φi1(X),φi2(X),...,φin(X))|i=1,2,...,m} is any given basic function family, and different function families can be selected according to specific problems and data. φ1(X),...,φm(X) are pairwise independent basic function families. Commonly used basic functions are the trigonometric function family and the Fourier family. {ai=(ai1,ain,...,ain)T|i=1,2,...,m} is a learnable set of parameters.

    Based on the generalized inverse matrix theory, Huang et al. proposed a new type of single hidden layer feedforward neural network algorithm with excellent performance-extreme learning machine (ELM) [11]. The extreme learning machine network structure is shown in Figure 2.

    Figure 2.  Extreme learning machine network model.

    Given N different training samples {xi,tixiRD,tiRm,i=1,2,,N}, xi=[xi1,xi2,,xiD]T as the input vector, ti=[ti1,ti2,,tim]T as the corresponding expected output. g(x) is an activation function, which is a nonlinear piecewise continuous function that satisfies the ELM general approximation ability theorem. The commonly used functions are Sigmoid function, Gaussian function, etc. So the mathematical model in Figure 2 is expressed as follows:

    Hβ=T (3)

    where, H=[h1(x1)hL(x1)h1(xN)hL(xN)]N×L=[g(ω1x1+b1)g(ωLx1+bL)g(ω1xN+b1)g(ωLxN+bL)]N×L.

    In ELM, H is called a random feature mapping matrix, ωi=[ωi1,ωi2,...,ωiD] represents the input weight that connects the ith hidden layer neuron and the input layer neuron, bi represents the bias of the ith hidden layer neuron, and β=[β1,,βL]T represents the weight matrix between the output layer and the hidden layer. Hidden layer node parameters (ωi,bi) are randomly generated and remain unchanged.

    Calculate the output weight:

    β=H+T (4)

    where, H+ represents the Moore-Penrose generalized inverse of the hidden layer output matrix H.

    According to the example of functional extreme learning machine in Figure 3(a), it can be seen that a functional extreme learning machine network consists of the following elements:

    Figure 3.  A functional extreme learning machine network structure. (a) The initial structure of the functional extreme learning machine. (b) The equivalent simplified network.

    1) Several layers of storage units: One layer of input units ({x1,x2,x3}), one layer of output storing units ({d}). Several intermediate storage units ({G(x1,x2),x3}), they are used to store intermediate information produced by functional neurons. Storage units are represented by solid circles with corresponding names.

    2) One or more layers of processing units (i.e., functional neurons): Each functional neuron is a computing unit, which processes the input values from input units or the previous layer of functional neurons, and provides input data to the next layer of neurons or output units. As in Figure 3(a) {G,I,F}.

    3) A set of directed links: They connect the input units to the first layer of processing units, one layer of processing units to the next layer of processing units, and the last layer of computing units to the output units. The arrows indicate the direction in which information flows. Information flows only from the input layer to the output layer.

    All these elements together constitute the network architecture of the functional extreme learning machine (FELM). The network architecture corresponds to the functional equation one by one. The functional equation is the key to the FELM learning process. Therefore, the network structure is determined, and the generalization ability of the FELM is also defined.

    Note the following differences between standard neural networks and FELM networks:

    1) The functional neuron as shown in Figure 1(a) is the basic computing unit of FELM. It is different from the M-P neuron (Figure 4), which has no weights {wk} and biases, only parameters, and can have multiple outputs {O1,O2,...,O3}.

    Figure 4.  M-P neuron model.

    2) In standard neural networks, the functions are given and the weights must be learned. In FELM networks, the functional neuron functions can be linear combinations of any nonlinear correlation basic function families f(x1,x2,,xk)=nj=1ajφj(x1,x2,,xk). Where, {φj(x1,x2,...,xk)|j=1,2,...,n} is a given appropriate basic function family, which means that FELM can choose different basic function families for functional neurons depending on the specific problem and data. n represents the number of basic functions. {aj|j=1,2,...,n} is the learnable parameter set. The following are some commonly used function families: Polynomial family {1,x,x2,...,xm}, Fourier family {1,sin(x),cos(x),...,,sin(mx),cos(mx)} and exponential family {1,ex,ex,...,emx,emx}.

    Structural simplification: Each initial network structure corresponds to a functional equation set, then the functional equation set solution method is used to simplify the initial structure to obtain an equivalent FELM which is optimal. The functional equation corresponding to Figure 3(a) is:

    d=F(G(x1,x2),x3) (5)

    Theorem 1 of [66]: The general solution is continuous on a real rectangle of the functional equation F[G(x,y),z]=K[x,N(y,z)] and is G invertible in two variables. For a fixed value of the second variable, F is invertible in the first variable. For a fixed value of the first variable, K and N are invertible in the second variable:

    F(x,y)=k[f(x)+g(y)],G(x,y)=f1[p(x)+q(y)],K(x,y)=k[p(x)+n(y)],N(x,y)=n1[q(x)+g(y)], (6)

    where, f,k,n,p,q and g are arbitrary continuous and strictly monotonic functions. Therefore, according to Theorem 1 of [66], the general solution of functional Eq (5) is Eq (7):

    F(x1,x2)=k[f(x1)+r(x2)],G(x1,x2)=f1[p(x1)+q(x2)] (7)

    According to Eq (7), Eq (5) can be written as:

    d=F(G(x1,x2),x3)=k[p(x1)+q(x2)+r(x3)] (8)

    According to Eq (8), the corresponding topology structure can be drawn, as shown in Figure 3(b). Figure 3(b), (a) are equivalent, indicating that they get the same output when they have the same input. In the initial structure of FELM, the functional neuron function is multi-parameter. In the simplified FELM structure, the functional neuron function is a single parameter.

    Expression uniqueness of FELM: After structural simplification, the functional equation corresponding to the simplified functional network is d=k[p(x1)+q(x2)+r(x3)], but whether the expression of the functional equation is unique needs to be verified. The following is the verification process, assuming that there are two functional neuron function sets {k1,p1,q1,r1} and {k2,p2,q2,r2}, such that :

    k1[p1(x1)+q1(x2)+r1(x3)]=k2[p2(x1)+q2(x2)+r2(x3)].x1,x2,x3 (9)

    Let k2(v)=k1(vbcda), then v=p2(x1)+q2(x2)+r2(x3), vbcda=p1(x1)+q1(x2)+r1(x3), and v=ap1(x1)+aq1(x2)+ar1(x3)+b+c+d. So the solution of the functional equation is:

    p2(x1)=ap1(x1)+b;q2(x2)=aq1(x2)+c;r2(x3)=ar1(x3)+d (10)

    where a,b,c,d are arbitrary constants. Because any values (a,b,c,d), Eq (10) into Eq (7), will get the following result:

    F(x1,x2)=k2[ap1(x1)+aq1(x2)+ar1(x3)+b+c+d]=k1[p1(x1)+q1(x2)+r1(x3)] (11)

    Therefore, the expression of Eq (8) is unique.

    The FELM in Figure 3(b) is taken as an example to illustrate its parameter learning process. Write Eq (8) as follows

    k1(x4)=p(x1)+q(x2)+r(x3), (12)

    where x4 represents d.

    Each neuron function is a linear combination of given nonlinear correlation basic functions, that is

    p(x1)=m1j=1a1jφ1j(x1);q(x2)=m2j=1a2jφ2j(x2),r(x3)=m3j=1a3jφ3j(x3);k1(x4)=m4j=1a4jφ4j(x4), (13)

    where m1, m2, m3 and m4 are the numbers of basic functions of p, q, r and k respectively, and aij is the parameter coefficient of FELM.

    Let ai=[ai1,ai2,,aimi],i=1,2,3,4.A=[a1,a2,a3,a4]T is the parameter to be optimized. ffir=[φi1(xir),φi2(xir),,φimi(xir)],Ffr=[ff1r,ff2r,ff3r,ff4r];r=1,2,,n.Ff=[Ff1;Ff2;;Ffn]; B=[k1(x41),k1(x42),,k1(x4n)]T, n is the number of observed samples. Then Eq (14) is obtained:

    FfA=B (14)

    The parameters of FELM can be obtained by Eq (15).

    A=Ff+B (15)

    where Ff+ is the generalized inverse of Ff.

    The above example illustrates the process of model learning. The steps of constructing and simplifying the FELM network and then performing parameter learning are as follows:

    Step 1: Based on the characteristics of the problem to be solved, the initial network model is established;

    Step 2: Write the functional equation corresponding to the initial network model;

    Step 3: Using the functional equation solving method to solve the functional equation and obtain the general solution expression;

    Step 4: Based on the general solution expression, use its one-to-one correspondence with the FELM to redraw the corresponding FELM network (simplified FELM);

    Step 5: The FELM learning algorithm is used to obtain the optimal parameters of the model.

    In this section, on many benchmark practical problems in the field of function approximation and classification, the performance of the proposed FELM learning algorithm is compared with the commonly used network algorithms (ELM, OP-ELM, SVM, LS-SVM) on two artificial datasets, 20 different datasets (16 for regression, 4 for classification) and XOR classification problem to verify the effectiveness and superiority of FELM. Experimental environment description for FELM and comparison algorithms: 11th Gen Intel (R) Core (TM) i5-11320H @ 3.20 GHz, 16 GB RAM and MATLAB 2019b. ELM source code used in all experiments: http://www.ntu.edu.sg/home/egbhuang/, OP-ELM source code: https://research.cs.aalto.fi//aml/software.shtml, SVM source code: http://www.csie.ntu.edu.tw/cjlin/libsvm/, and the most popular LS-SVM implementation: http://www.esat.kuleuven.ac.be/sista/lssvmlab/. The sigmoidal activation function is used for ELM, the Gaussian kernel function is used for OP-ELM, and the radial basis function is used for SVM and LS-SVM. The basic functions of our proposed algorithm FELM will be set according to the following specific problems to be solved. In Section 3.1 and Section 3.2, FELM adopts the network structure of Figure 3(b).

    It is well known that the performance of SVM is sensitive to the combination of (C,γ). Similar to SVM, the generalization performance of LS-SVM is also closely dependent on the combination of (C,γ). Therefore, in order to achieve good generalization performance, it is necessary to select appropriate cost parameter C and kernel parameter γ for SVM and LS-SVM in each dataset. We tried 17 different values of C and γ, that is, for each dataset, we used 17 different C values and 17 different γ values, a total of 289 pairs (C,γ). Each problem is tested 50 times, and the training data set and the test data set are randomly generated from the entire dataset. For each dataset, two-thirds is the training set and the rest is the test set. This section gives the simulation results, including average training and test accuracy, corresponding standard deviation (Dev), and training time. In experiments, all inputs (attributes) and outputs (targets) have been normalized into [-1, 1].

    To test the performance of FELM on regression problems, we first use the objective function 'sinc' function, which is defined as:

    y=Sinc(x)=sinxx,x[4π,4π].

    To effectively reflect the performance of our algorithm, some different forms of zero mean Gaussian noise pollution are added to the training data points. In particular, we have the following training samples (xi,yi),i=1,2,...,l.

    (Type A) yi=sinxixi+ξi,xiU[4π,4π],ξiN(0,0.12)

    (Type B) yi=sinxixi+ξi,xiU[4π,4π],ξiN(0,0.22)

    Next, we proceed to compare the performance of the proposed FELM with other algorithms using the following two synthetic datasets.

    g(x)=|x14|+|sin(π(1+x14))|+1,10x10.

    Again, all training data points are shifted by adding different forms of Gaussian noise below.

    (Type C) yi=g(xi)+ξi,xiU[10,10],ξiN(0,0.22)

    (Type D) yi=g(xi)+ξi,xiU[10,10],ξiN(0,0.42)

    where U[a,b] and N(c,d2) denote the uniform random variable in [a,b] and the Gaussian random variable with mean c and variance d2, respectively. The training set and the test set have 5000 data respectively, which are evenly and randomly distributed on the interval [a,b]. To make the regression problem 'real', Types A–D added four different forms of Gaussian noise to all training samples, while the test data remained noise-free.

    As shown in Table 1, appropriate basic functions are assigned to our FELM algorithm on four different synthetic datasets. The initial node number of ELM algorithm is 5, and the optimal number of hidden layer nodes is found by interval 5 nodes in 5–100, and the initial maximum number of neurons for OP-ELM is 100. The results of 50 experiments on all algorithms are shown in Table 2, where bold indicates optimal test accuracy. Figure 5 plots the one-time fit curves of FELM and other regressors on these synthetic datasets with different noise types. It can be seen from Table 2 that the proposed FELM learning algorithm achieves the highest test accuracy (root mean square error, RMS) on artificial datasets with noise types A, B and D. On the artificial dataset with noise type C, FELM is superior to ELM, OP-ELM and LSSVR, and is only lower than SVR. Table 2 also shows the optimal parameter combinations of SVR and LSSVR on these synthetic datasets and the required support vectors (SVs), the network complexity (nodes) of FELM, ELM and OP-ELM. In addition, Table 2 also compares the training and testing time of these five methods. It can be seen that the proposed FELM is the fastest learning method, which is several times or dozens of times faster than ELM, and hundreds of times faster than OP-ELM, SVR and LSSVR. This is because compared with ELM, SVR and LSSVR, FELM has the smallest network complexity, so it requires less learning time. Compared with OP-ELM, FELM does not need to cut out redundant nodes, so it requires less training time. Since the number of support vectors required for SVR and the support vectors required for LSSVR are much larger than the network complexity of FELM, they all take more test time than FELM, at least 60 times more than it does, which means that FELM trained in actual deployment may respond to new external unknown data much faster than SVM and LS-SVM. In short, the proposed FELM outperforms the other four comparison algorithms in approaching four artificial datasets with different types of noise.

    Table 1.  Basic functions given by FELM on 4 synthetic datasets.
    Datasets Basic functions
    Types A and B {1,x2,x4,x6,x8,x10,cos(x),cos(3x),cos(5x)}
    Type C {x,x2,x3,sin(x2),cos(x2),sin(x),cos(x),cos(2x),ex3,ex3,ex5,cos(π2(1+x)), cos(π(1+x)),cos(2π(1+x)),cos(3π(1+x)),arctan(x7),arctan(x6), arctan(x5),arctan(x4),arctan(x3),arctan(x2),arctan(x),arctan(2x), arctan(3x),arctan(4x),arctan(5x)}
    Type D {x,sin(x2),cos(x2),cos(x3),sin(x),sin(2x),cos(x),cos(2x),cos(3x),ex3,ex3,ex5, cos(π2(1+x)),cos(π(1+x)),cos(2π(1+x)),cos(3π(1+x))}

     | Show Table
    DownLoad: CSV
    Table 2.  Performance comparison of FELM, ELM, OP-ELM, SVR and LSSVR on four datasets with different types of noises.
    Noise Regressor Time(s) Testing SVs/nodes
    (C, γ) Training Testing RMS DEV
    Type A FELM 0.0024 0.0014 0.0065 0.0006 9
    ELM 0.0141 0.0047 0.0065 0.0012 20
    OP-ELM 0.8194 0.0025 0.0060 0.0010 15.50
    SVR (27, 2-2) 2.2711 0.3150 0.0145 0.0020 1613.46
    LSSVR (28, 20) 2.0607 0.3824 0.0087 0.0010 5000
    Type B FELM 0.0027 0.0014 0.0098 0.0015 9
    ELM 0.0138 0.0084 0.0127 0.0021 20
    OP-ELM 0.7905 0.0024 0.0115 0.0024 14.70
    SVR (22, 2-2) 1.7248 0.6016 0.0188 0.0026 3089.38
    LSSVR (28, 20) 1.9938 0.3710 0.0177 0.0018 5000
    Type C FELM 0.0053 0.0024 0.0290 0.0011 26
    ELM 0.4159 0.0194 0.0328 0.0012 90
    OP-ELM 0.8390 0.0031 0.0686 0.0029 19.20
    SVR (28, 2-1) 35.1289 0.5969 0.0283 0.0014 3107.00
    LSSVR (28, 2-8) 2.2814 0.3572 0.0601 0.0022 5000
    Type D FELM 0.0042 0.0015 0.0384 0.0019 16
    ELM 0.4656 0.0209 0.0429 0.0032 100
    OP-ELM 0.7586 0.0027 0.0725 0.0046 18.10
    SVR (27, 2-1) 15.4804 0.8234 0.0434 0.0040 4020.82
    LSSVR (28, 20) 1.9812 0.3792 0.0396 0.0037 5000

     | Show Table
    DownLoad: CSV
    Figure 5.  Predictions of different regressors on four synthetic datasets with different forms of noise.

    For further evaluation, 16 different regression datasets are selected. These datasets are usually used to test machine learning algorithms, mainly from UCI Machine Learning repository [70] and StatLib [71]. The different attributes of 16 datasets are summarized in Table 3.

    Table 3.  Examples of actual regression.
    Datasets #Train #Test #Total #Features
    Abalone 2784 1393 4177 8
    Mpg 261 131 392 7
    Autoprice 106 53 159 15
    Balloon 1334 667 2001 2
    Baskball 64 32 96 4
    Cleveland 202 101 303 13
    Cloud 72 36 108 9
    Concrete CS 686 344 1030 8
    Diabetes 28 15 43 2
    Housing 337 169 506 13
    Machine CPU 139 70 209 6
    Mg 923 462 1385 6
    Quake 1452 726 2178 3
    Servo 111 56 167 4
    Strike 416 209 625 6
    Wisconsin B.C. 129 65 194 32

     | Show Table
    DownLoad: CSV

    As shown in Table 4, we assign appropriate basic functions to our FELM algorithm on 16 different datasets. It can be seen that these basic functions are relatively short in length, indicating that the structural complexity of the networks is low. The initial number of nodes in ELM is 5, and the optimal number of hidden layer nodes is found at intervals of 5 nodes within 5–100. The optimal number of nodes obtained by ELM on each dataset is shown in Table 5. The table also shows the best parameter combination and support vector number of SVR and LSSVR on each dataset, the initial maximum number of neurons and the number of neurons after pruning of OP-ELM.

    Table 4.  Basic functions given by FELM on 16 regression datasets.
    Datasets Basic functions Datasets Basic functions
    Abalone {ex,ex} Diabetes {ex,ex}
    Mpg {1,ex2,ex3,ex4} Housing {1,ex,ex}
    Autoprice {1,ex4} Machine CPU {1,ex,ex}
    Balloon {ex,ex,e2x,e2x,1,x,x,x2,x3} Mg {sin(3x),sin(5x),x3,ex2,ex2,ex5,ex5}
    Baskball {ex5,ex6} Quake {1,sin(x),sin(3x),sin(5x),x,x3,x5,ex,ex}
    Cleveland {1,ex2} Servo {1,e3x,e3x,sin(x4),cos(x4),sin(x6),cos(x6)}
    Cloud {1,ex5} Strike {1,sin(x4)}
    Concrete CS {1,sin(x),sin(3x),sin(5x),x,x2,x3,ex,ex} Wisconsin B.C. {sin(x4),cos(x4)}

     | Show Table
    DownLoad: CSV
    Table 5.  Comparison of network complexity.
    Algorithm SVR LSSVR ELM OP-ELM
    ε (C,γ) # SVs (C,γ) # SVs # nodes init final
    Abalone 25 (21,20) 1051.28 (25,21) 2784 35 100 33
    Mpg 28 (20,21) 104.08 (22,21) 261 30 100 36
    Autoprice 26 (23,23) 42.66 (26,22) 106 15 100 14
    Balloon 23 (27,22) 6 (28,2-1) 1334 20 100 41
    Baskball 28 (23,24) 44.38 (22,22) 64 10 62 7
    Cleveland 26 (21,27) 142.72 (24,28) 198 20 100 9
    Cloud 26 (27,28) 20.62 (28,26) 72 15 70 20
    Concrete CS 28 (24,21) 174.42 (27,2-1) 686 90 100 87
    Diabetes 25 (21,22) 21.12 (22,20) 28 5 26 6
    Housing 28 (23,22) 123.42 (28,22) 337 80 100 58
    Machine CPU 27 (28,26) 8.74 (27,23) 139 30 100 15
    Mg 26 (21,20) 592.84 (21,2-2) 923 70 100 84
    Quake 21 (24,27) 493.28 (20,28) 1452 30 100 11
    Servo 27 (25,23) 46.36 (26,21) 111 25 100 41
    Strike 27 (28,28) 88.74 (22,22) 416 10 100 11
    Wisconsin B.C. 28 (27,28) 14.92 (28,26) 129 75 100 51

     | Show Table
    DownLoad: CSV

    The results of 50 trials on 16 datasets by the proposed FELM and other comparison algorithms are shown in Tables 68. The bold body in Table 6 indicates the optimal test accuracy. The comparison of FELM and the other four comparison algorithms on testing RMSE is shown in Table 6. The minimum test RMSE is obtained on 10 datasets of Autoprice, Balloon, Baskball, Cleveland, Cloud, Diabetes, Machine CPU, Servo, Strike and Wisconsin B.C. On other datasets, although the accuracy obtained by our algorithm is lower than SVR and LSSVR, it is higher than ELM and OP-ELM. The comparison of the five algorithms in training and testing time is shown in Table 7. The table shows that our FELM spends similar training and testing time as ELM, but much less than OP-ELM, SVR and LSSVR. The comparison results of FELM and other algorithms on the standard deviation of testing RMSE are shown in Table 8. According to the table, our FELM is a stable learning method. Figure 6 shows the test RMSE comparison of FELM and other comparison algorithms running 50 times on four datasets (Autoprice, Cleveland, Abalone and Quake). In short, combined with Tables 68 and the intuitive display of Figure 6, we can know that the proposed method FELM not only has good versatility and stability, but also has fast training speed.

    Table 6.  Comparison of testing RMSE.
    Datasets ELM OP-ELM SVR LSSVR FELM
    Abalone 0.1566 0.2186 0.1513 0.1501 0.1525
    Mpg 0.1575 0.1568 0.1403 0.1389 0.1415
    Autoprice 0.1927 0.1804 0.1712 0.1669 0.1600
    Balloon 0.0157 0.0175 0.0420 0.0097 0.0072
    Baskball 0.2631 0.2669 0.2623 0.2576 0.2477
    Cleveland 0.4538 0.4359 0.4370 0.4281 0.4204
    Cloud 0.1458 0.1761 0.1240 0.1234 0.1168
    Concrete CS 0.1249 0.1196 0.1008 0.0792 0.1122
    Diabetes 0.3787 0.4635 0.3456 0.3242 0.2938
    Housing 0.1972 0.2305 0.1458 0.1478 0.1754
    Machine CPU 0.0474 0.0933 0.0731 0.0398 0.0395
    Mg 0.2748 0.2722 0.2719 0.2657 0.2678
    Quake 0.3475 0.3466 0.3420 0.3441 0.3436
    Servo 0.2300 0.1979 0.1819 0.1785 0.1694
    Strike 0.1517 0.1614 0.1541 0.1423 0.1320
    Wisconsin B.C. 0.0782 0.0269 0.0670 0.0264 0.0210

     | Show Table
    DownLoad: CSV
    Table 7.  Comparison of training and testing time.
    Datasets ELM (s) OP-ELM (s) SVR(s) LSSVR(s) FELM (s)
    Train Test Train Test Train Test Train Test Train Test
    Abalone 0.0020 0.0008 0.5504 0.0019 0.3782 0.0756 0.5977 0.0791 0.0021 0.0006
    Mpg 0.0004 0.0004 0.0405 0.0008 0.0040 0.0007 0.0066 0.0017 0.0006 0.0002
    Autoprice 0.0002 0.0004 0.0248 0.0003 0.0010 0.0002 0.0037 0.0012 0.0007 0.0004
    Balloon 0.0006 0.0005 0.1419 0.0014 0.0015 0.0003 0.1311 0.0203 0.0011 0.0003
    Baskball 0.0001 0.0004 0.0108 0.0002 0.0006 0.0001 0.0027 0.0012 0.0001 0.0001
    Cleveland 0.0003 0.0007 0.0325 0.0003 0.0047 0.0010 0.0052 0.0018 0.0004 0.0002
    Cloud 0.0001 0.0004 0.0136 0.0004 0.0005 0.0001 0.0027 0.0011 0.0002 0.0001
    Concrete CS 0.0024 0.0008 0.1046 0.0025 0.0427 0.0028 0.0355 0.0077 0.0033 0.0009
    Diabetes 0.0001 0.0004 0.0048 0.0002 0.0002 0.0000 0.0024 0.0011 0.0001 0.0000
    Housing 0.0013 0.0005 0.0516 0.0015 0.0099 0.0012 0.0073 0.0029 0.0011 0.0003
    Machine CPU 0.0003 0.0004 0.0260 0.0004 0.0004 0.0001 0.0038 0.0020 0.0002 0.0001
    Mg 0.0016 0.0006 0.1538 0.0029 0.0628 0.0114 0.0634 0.0089 0.0038 0.0016
    Quake 0.0010 0.0005 0.2497 0.0005 0.0656 0.0130 0.1651 0.0186 0.0023 0.0006
    Servo 0.0002 0.0004 0.0268 0.0008 0.0027 0.0002 0.0035 0.0011 0.0004 0.0002
    Strike 0.0002 0.0003 0.0577 0.0003 0.0139 0.0012 0.0125 0.0039 0.0004 0.0001
    Wisconsin B.C. 0.0012 0.0004 0.0289 0.0010 0.0007 0.0001 0.0041 0.0019 0.0014 0.0006

     | Show Table
    DownLoad: CSV
    Table 8.  Comparison of the standard deviation of testing RMSE.
    Datasets ELM OP-ELM SVR LSSVR FELM
    Abalone 0.0068 0.2685 0.0040 0.0043 0.0043
    Mpg 0.0134 0.0176 0.0135 0.0130 0.0123
    Autoprice 0.0340 0.0556 0.0364 0.0397 0.0240
    Balloon 0.0102 0.0121 0.0121 0.0019 0.0005
    Baskball 0.0285 0.0288 0.0293 0.0272 0.0278
    Cleveland 0.0312 0.0356 0.0351 0.0343 0.0316
    Cloud 0.0309 0.0672 0.0444 0.0363 0.0317
    Concrete CS 0.0083 0.0085 0.0085 0.0084 0.0050
    Diabetes 0.1095 0.3457 0.0575 0.0524 0.0443
    Housing 0.0356 0.1718 0.0192 0.0189 0.0197
    Machine CPU 0.0620 0.0623 0.0191 0.0249 0.0221
    Mg 0.0100 0.0088 0.0105 0.0079 0.0083
    Quake 0.0088 0.0093 0.0112 0.0090 0.0085
    Servo 0.0407 0.0442 0.0584 0.0405 0.0423
    Strike 0.0383 0.0312 0.0305 0.0380 0.0331
    Wisconsin B.C. 0.0183 0.0091 0.0097 0.0030 0.0041

     | Show Table
    DownLoad: CSV
    Figure 6.  Comparison of test RMSE of FELM, ELM, OP-ELM, SVR and LSSVR running 50 times on four datasets.

    The XOR problem dataset randomly generates 1000 samples here, with one class containing 478 samples and the other containing 522 samples. The binary problem is not linearly separable. On this problem, if FELM adopts the structure shown in Figure 3(b), it is not easy to find suitable basic function families for its neurons. Therefore, we adopt the multi-input single-output FELM structure shown in Figure 7 to achieve better generalization performance by increasing the number of hidden layer nodes. FELM sets the number of hidden layer nodes on this problem to be 73, and the node functions correspond to the following:

    Figure 7.  Multiple-input single-output functional extreme learning machine.

    {sin(x2); sin(3x2); sin(5x2); x2; e2x2; sin(x1)sin(x2); sin(x1)sin(3x2); sin(x1)sin(5x2); sin(x1)x2; sin(x1)x32; sin(x1)ex2; sin(x1)ex2; sin(3x1)sin(5x2); sin(3x1)x2; sin(3x1)x22; sin(3x1)e2x2; sin(3x1)e2x2; sin(5x1)sin(x2); sin(5x1)sin(5x2); sin(5x1)x2; sin(5x1)x22; sin(5x1)ex2; sin(5x1)e2x2; sin(5x1)e2x2; x1; x1sin(x2); x1sin(3x2); x1sin(5x2); x1x2; x1x22; x1e2x2; x21; x21sin(3x2); x21x2; x21x22; x21ex2; x21e2x2; x21e2x2; x31sin(x2); x31sin(5x2); x31x2; x31x22; x31ex2; x31ex2; x31e2x2; ex1sin(x2); ex1sin(3x2); ex1sin(5x2); ex1x2; ex1x32; ex1ex2; ex1ex2; ex1e2x2; ex1e2x2; ex1x2; ex1ex2; ex1e2x2; e2x1; e2x1sin(x2); e2x1sin(3x2); e2x1sin(5x2); e2x1x2; e2x1x22; e2x1ex2; e2x1e2x2; e2x1e2x2; e2x1x2; e2x1x22; e2x1x32; e2x1ex2; e2x1ex2; e2x1e2x2; e2x1e2x2}.

    The initial maximum number of neurons for OP-ELM is 100. This section also adds an ELM comparison with an activation function of RBF. The initial number of nodes of the ELM is 5, and the optimal number of hidden layer nodes is found at intervals of 10 nodes within 5–1000, and the optimal number of nodes obtained on each dataset is shown in Table 9. The table also shows the optimal parameter combination and support vector number of SVM and LSSVM on the problem, and the final number of neurons of OP-ELM.

    Table 9.  Performance comparison of FELM, ELM, OP-ELM, SVM and LSSVM on 'XOR' dataset.
    Regressor (C, γ) Time(s) Testing SVs/nodes
    Training Testing Rate (%) DEV (%)
    FELM 0.0704 0.0017 98.82 0.62 73
    ELM (sig) 0.0054 0.0013 97.29 0.86 155
    ELM (RBF) 0.0020 0.0013 97.38 0.90 75
    OP-ELM 0.0704 0.0015 97.24 1.06 49
    SVM (24, 26) 0.0188 0.0033 97.29 0.87 269.86
    LSSVM (23, 2-4) 0.0219 0.0062 97.89 0.78 666

     | Show Table
    DownLoad: CSV

    The average results of 50 trials conducted by FELM and other models on the XOR dataset are shown in Table 9. The data in the table show that the performance of FELM is better than ELM, OP-ELM, SVM and LSSVM. Figure 8 shows the boundaries of different classifiers on the XOR problem. It can be seen that, similar to ELM, OP-ELM, SVM and LS-SVM, FELM can solve the XOR problem well.

    Figure 8.  Separating boundaries of different classifiers in XOR case.

    The newly proposed FELM algorithm is compared with four other popular algorithms (ELM, OP-ELM, SVM and LSSVM) on four classification problems: Iris, WDBC, Diabetes and Wine. These four datasets are from UCI Machine Learning repository [70], and the number of samples, attributes and classes are shown in Table 10. The ELM algorithm sets the initial number of nodes to 5, and finds the optimal number of hidden layer nodes at intervals of 10 nodes within 5–1000. As shown in Table 11, we assign appropriate basic functions to these datasets for our FELM algorithm, and the optimal number of nodes obtained by the ELM algorithm on each dataset. The table also shows the optimal parameter combination and support vector number of SVM and LSSVM, the initial maximum number of neurons and the number of neurons after pruning of OP-ELM.

    Table 10.  Examples of actual classification. WDBC stands for Wisconsin Breast Cancer. Diabetes stands for Pima Indians Diabetes.
    Datasets #Train #Test #Total #(Featdures/Classes)
    Iris 100 50 150 4/3
    WDBC 379 190 569 30/2
    Diabetes 512 256 768 8/2
    Wine 118 60 178 13/3

     | Show Table
    DownLoad: CSV
    Table 11.  Comparison of network complexity.
    Algorithm FELM SVM LSSVM ELM OP-ELM
    basic functions (C, γ) # SVs (C, γ) # SVs # nodes init final
    Iris {1,ex,ex,e2x,e2x} (26, 2-5) 20.48 (2-4, 23) 100 15 90 16.80
    WDBC {1,e3x,e5x} (22, 2-3) 55.04 (22, 25) 379 65 90 32.90
    Diabetes {ex,ex} (24, 2-5) 281.7 (24, 23) 512 15 90 19.90
    Wine {ex,e2x} (2-1, 2-2) 63.64 (2-3, 23) 118 15 90 33.40

     | Show Table
    DownLoad: CSV

    The performance comparison between all algorithms is shown in Tables 1214. In the comparison of these five algorithms, obviously, better test results are given in bold. In Table 12, our FELM compared with four other algorithms on testing correct classification rate, and FELM achieved the highest test correct classification rate. The comparison results of the five algorithms for training and testing time are shown in Table 13. The learning speed of FELM is similar to the ELM, which is much faster than the OP-ELM, SVM and LSSVM. In Table 14, FELM is compared with other comparison algorithms on the standard deviation of testing correct classification rate, and the results show that FELM has good stability. Figure 9 shows the successful classification rate comparison of FELM and the other four algorithms running 50 times on four classification datasets. It can be seen that FELM obtains the highest number of higher classification rates. Compared with other algorithms, the curve fluctuation is smaller, indicating that its stability is better. In summary, combined with Tables 1214 and Figure 9, it can be seen that FELM not only guarantees the learning speed in all cases, but also achieves better generalization performance.

    Table 12.  Comparison of testing correct classification Rate.
    Datasets ELM OP-ELM SVM LSSVM FELM
    Iris 95.36 95.36 96.40 91.32 97.76
    WDBC 95.16 95.83 97.78 95.01 97.81
    Diabetes 76.83 76.89 77.59 73.70 77.85
    Wine 96.03 95.73 97.90 94.50 98.87

     | Show Table
    DownLoad: CSV
    Table 13.  Comparison of training and testing time.
    Datasets ELM (s) OP-ELM (s) SVM (s) LSSVM (s) FELM (s)
    Train Test Train Test Train Test Train Test Train Test
    Iris 0.0002 0.0005 0.0217 0.0004 0.0004 0.0001 0.0022 0.0036 0.0002 0.0001
    WDBC 0.0013 0.0009 0.0519 0.0010 0.0047 0.0010 0.0079 0.0161 0.0043 0.0010
    Diabetes 0.0004 0.0007 0.0716 0.0006 0.0199 0.0044 0.0120 0.0078 0.0005 0.0002
    Wine 0.0002 0.0005 0.0255 0.0007 0.0011 0.0003 0.0024 0.0076 0.0004 0.0002

     | Show Table
    DownLoad: CSV
    Table 14.  Comparison of the standard deviation of testing correct classification rate.
    Datasets ELM OP-ELM SVM LSSVM FELM
    Iris 3.19 2.95 2.62 3.89 1.84
    WDBC 1.63 1.45 0.91 1.56 1.00
    Diabetes 2.21 2.15 2.15 3.15 2.17
    Wine 3.31 2.43 1.64 2.82 1.41

     | Show Table
    DownLoad: CSV
    Figure 9.  Comparison of test successful classification rates for FELM, ELM, OP-ELM, SVM and LSSVM on four datasets run 50 times.

    In this paper, we propose a new method for data regression and classification called functional extreme learning machine (FELM). Different from the traditional ELM, FELM is problem-driven rather than model-driven, without the concept of weight and bias. It uses the functional neuron as the basic unit, and uses functional equation solving theory to guide its modeling process. The functional neuron of the learning machine is represented by a linear combination of any linearly independent basic functions, and infinitely approximates the desired accuracy by adjusting the coefficients of the basic functions in the functional neuron. In addition, the parameter fast learning algorithm proposed in this paper does not need iteration and has high accuracy. Its learning process is different from the ELM used by people at present. It is expected to fundamentally overcome the shortcomings of the random initial parameters of the hidden layer (connection weights, bias values, number of nodes) in the current ELM theory that significantly affect the classification accuracy of ELM. Like ELM, FELM has less human intervention. It only needs to match the appropriate basic functions for the problem, and can obtain the optimal parameters according to the parameter learning algorithm without iteration. Simulation results show that compared with ELM, FELM has better performance and similar learning speed in regression and classification. Compared to SVM and LS-SVM, FELM can run stably with faster learning speed (up to several hundred times) while guaranteeing generalization performance. The proposed FELM theory provides a new idea for tapping the potential of extreme learning and broadening the application of extreme learning, which has important theoretical significance and broad application prospects. In the future work, we will use the parameter screening algorithm to further improve the generalization ability and stability of FELM and broaden its practical application range. These are the author's next works.

    This research is funded by the National Natural Science Foundation of China, Grant number 62066005, U21A20464. Project of the Guangxi Science and Technology under Grant No. 2019KY0185.

    The authors declare there is no conflict of interest.

  • This article has been cited by:

    1. Maryam Sayadi, Behzad Hessari, Majid Montaseri, Amir Naghibi, Enhanced TDS Modeling Using an AI Framework Integrating Grey Wolf Optimization with Kernel Extreme Learning Machine, 2024, 16, 2073-4441, 2818, 10.3390/w16192818
    2. Rongchao Yang, Qingbo Zhou, Lei Xu, Yi Zhang, Tongyang Wei, Forecasting the Total Output Value of Agriculture, Forestry, Animal Husbandry, and Fishery in Various Provinces of China via NPP-VIIRS Nighttime Light Data, 2024, 14, 2076-3417, 8752, 10.3390/app14198752
    3. Qasem Abu Al-Haija, Shahad Altamimi, Mazen AlWadi, Analysis of Extreme Learning Machines (ELMs) for intelligent intrusion detection systems: A survey, 2024, 253, 09574174, 124317, 10.1016/j.eswa.2024.124317
    4. Xiangming Kong, Ningyun Lu, Baoli Zhang, Jianfei Chen, 2023, RUL prediction of rolling element bearings based on ACO-FELM, 979-8-3503-3775-4, 1, 10.1109/SAFEPROCESS58597.2023.10295598
    5. Shahad Altamimi, Qasem Abu Al-Haija, Maximizing intrusion detection efficiency for IoT networks using extreme learning machine, 2024, 4, 2730-7239, 10.1007/s43926-024-00060-x
    6. Arthi. A, Kapil Aggarwal, R Karthikeyan, S. Kayalvili, Srimathi S, Aviral Srivastava, 2023, Hybrid Multimodal Machine Learning Driven Android Malware Recognition and Classification Model, 979-8-3503-4060-0, 1555, 10.1109/ICECA58529.2023.10394838
    7. Abera Abiyo Dofee, Pritam Chand, Spatiotemporal dynamics of land use land cover patterns in the middle Omo-Gibe River Basin, Ethiopia: machine learning, geospatial, and field survey integrated approach, 2025, 11, 2363-6203, 10.1007/s40808-024-02185-y
    8. José Ignacio Amorena, Elvira Fernández-Ahumada, Dolores María Eugenia Álvarez, Preventing adulteration: authentication and identification of camelids and other textile fibers with NIR spectroscopy, 2025, 0040-5000, 1, 10.1080/00405000.2024.2447616
    9. Senhui Wang, Applying genetic algorithm to extreme learning machine in prediction of tumbler index with principal component analysis for iron ore sintering, 2025, 15, 2045-2322, 10.1038/s41598-025-88755-1
    10. Sheng Li, Shunli Wang, Wen Cao, Liya Zhang, Carlos Fernandez, An improved pelican optimization-kernel extreme learning machine for highly accurate state of charge estimation of lithium-ion batteries in energy storage systems, 2025, 0947-7047, 10.1007/s11581-025-06327-9
  • Reader Comments
  • © 2004 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(4328) PDF downloads(1036) Cited by(23)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog