Figure 1.
The process of particle movement.
Citation: Fumiaki Uchiumi, Akira Sato, Masashi Asai, Sei-ichi Tanuma. An NAD+ dependent/sensitive transcription system: Toward a novel anti-cancer therapy[J]. AIMS Molecular Science, 2020, 7(1): 12-28. doi: 10.3934/molsci.2020002
| [1] | Xianli Liu, Yongquan Zhou, Weiping Meng, Qifang Luo . Functional extreme learning machine for regression and classification. Mathematical Biosciences and Engineering, 2023, 20(2): 3768-3792. doi: 10.3934/mbe.2023177 |
| [2] | 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 |
| [3] | Yan Yan, Yong Qian, Hongzhong Ma, Changwu Hu . Research on imbalanced data fault diagnosis of on-load tap changers based on IGWO-WELM. Mathematical Biosciences and Engineering, 2023, 20(3): 4877-4895. doi: 10.3934/mbe.2023226 |
| [4] | 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 |
| [5] | 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 |
| [6] | 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 |
| [7] | Jian-xue Tian, Jue Zhang . Breast cancer diagnosis using feature extraction and boosted C5.0 decision tree algorithm with penalty factor. Mathematical Biosciences and Engineering, 2022, 19(3): 2193-2205. doi: 10.3934/mbe.2022102 |
| [8] | Yufeng Li, Chengcheng Liu, Weiping Zhao, Yufeng Huang . Multi-spectral remote sensing images feature coverage classification based on improved convolutional neural network. Mathematical Biosciences and Engineering, 2020, 17(5): 4443-4456. doi: 10.3934/mbe.2020245 |
| [9] | Xiao Liang, Taiyue Qi, Zhiyi Jin, Wangping Qian . Hybrid support vector machine optimization model for inversion of tunnel transient electromagnetic method. Mathematical Biosciences and Engineering, 2020, 17(4): 3998-4017. doi: 10.3934/mbe.2020221 |
| [10] | Hong Lu, Hongxiang Zhan, Tinghua Wang . A multi-strategy improved snake optimizer and its application to SVM parameter selection. Mathematical Biosciences and Engineering, 2024, 21(10): 7297-7336. doi: 10.3934/mbe.2024322 |
Classification is a very important issue in many fields such as face detection, big data, and disease diagnosis. Especially in recent years, with the development of internet and smart devices, various types of data are exploding. In order to obtain accurate results more efficiently, the traditional image analysis method and signal detection are being replaced by machine learning methods gradually. In all these methods, artificial neural networks (ANNs) [1] and support vector machine (SVM) [2] are the two most popular methods. For ANNs, many neural network models have been developed, such as back propagation algorithm (BP) [3] and convolutional neural networks (CNN) [4]. However, these techniques are time-consuming, easy to be trapped in local optima and require the setting of many parameters. To overcome these disadvantages, the extreme learning machine (ELM) method has been proposed for single-hidden layer feed-forward neural network [5].
ELM has advantages of high learning speed and excellent classification performance owing to its inherent characteristics of simple structure. Due to the above advantages, ELM has been widely used in various fields, such as localization [6], industrial production [7], solar radiation prediction [8], finite-time optimal control of nonlinear systems [9], etc. In addition, ELM has several variants to solve complex problems. Zhang et al. proposed a multilayer probability extreme learning machine for device-free localization [10]. Youngmin Park combined convolutional neural network and ELM for image classification [11]. In ELM, the input weights and hidden bias are randomly generated without iterative learning [12]. Although these settings bring certain advantages, they also increase the risk of overfitting. Besides, the hidden neurons are sensitive to unknown testing data. Traditionally, choice of these parameters mainly depends on prior knowledge and expertise. To solve these problems, it is important to optimize the input weights, hidden bias and the structure of the network.
Intelligent algorithms are naturally considered as the solution to the above problems, such as particle swarm optimization (PSO) [13], ant colony optimization (ACO) [14], and artificial bee colony algorithm (ABC) [15]. Electromagnetism-like mechanism (EM) algorithm, which was developed by Birbil and Fang in 2003 [16], is a population-based random search algorithm similar to genetic algorithm (GA). Because of its strong search capability and easy implementation, EM has been successfully applied to optimization problems [16,17,18,19,20,21,22,23], such as function optimization [17] and flow shop scheduling [18,19,20,21]. All these previous studies have demonstrated the excellent optimization performance of EM. Therefore, the integration of EM and ELM should be a promising approach in training feedforward neural network.
In this study, an improved EM algorithm called DAEM is proposed by incorporating some theories of dragonfly algorithm (DA) [24] into EM approach. By using the new algorithm, we optimized the input weights and hidden biases, and minimum norm least-square scheme was used to analytically determine the output weights in ELM. In the selection of input weights and hidden biases, the improved EM considers not only the classification error rate but also the norm of the output weights as well as constrains the input weights and hidden biases within a reasonable range. In addition, the k-fold cross-validation method is adopted to avoid the problem of overfitting.
The rest of the paper is organized as follows. The theories related to ELM and EM are briefly introduced in Section 2. Section 3 describes the establishment of the DAEM-ELM algorithm. Section 4 presents the results and discussion on eight classification problems to demonstrate the effectiveness of the proposed algorithm. Finally, the conclusions are summarized in Section 5.
The core idea of ELM is to transform the training process of traditional SLFN model into solving the least square solution problem. The main process of ELM consists of random generation of the parameters of hidden neurons, followed by fixing of the hidden layer parameters and then algebraically solving the output weights. The specific theoretical basis of ELM is as follows.
For N arbitrary distinct samples (xi, ti), where xi = [xi1, xi2, …, xin]T∈Rn, ti = [ti1, ti2, …, tim]T∈Rm. The i-th sample xi is an n × 1 feature vector, and ti is an m × 1 target vector. The standard mathematical model of SLFNs with L hidden neurons and activation function g (x) is as follows:
| Oj=L∑i=1βig(ωi⋅xj+bi),j=1,2,...,N | (1) |
where Oj denotes the corresponding actual output vector of xj, ωi = [ωi1, ωi2, …, ωin]T indicates the weight vector connecting the i-th hidden neuron and input neurons, βi = [βi1, βi2, …, βim]T represents the weight vector connecting the i-th hidden neuron and output neurons, bi is the bias of i-th hidden neuron, also known as the threshold, and ωi·xj is the inner product of ωi and xj. The purpose of training SLFNs is to minimize the error of output value, which means:
| minL∑j=1‖Oj−tj‖ | (2) |
Then, the N equations represented by equation (1) can be expressed in matrix form as follows:
| Hβ=T | (3) |
where H=H(ω1,ω2,...,ωL,b1,b2,...,bL,x1,x2,...,xN)=[g(ω1⋅x1+b1)...g(ω1⋅x1+bL).........g(ω1⋅xN+b1)...g(ωL⋅xN+bL)]N×L, β=[βT1⋮βTL]L×m, T=[tT1⋮tTN]N×m, H is the hidden layer output matrix, β is the output weight matrix, and T is the output matrix.
In the algorithm of ELM, when the input weights and hidden layer biases are randomly generated, the determination of the output weights is to find the least-square (LS) solution to the linear system:
| β=H+T | (4) |
where H+ is obtained by singular value decomposition of Moore-Penrose (MP) generalized inverse matrix.
The pseudo-code of ELM is as follows:
| Input: (xi, ti), L, g(x) |
| Generate the input weights ωi and the biases bi of hidden |
| neurons randomly; |
| H ← Compute the hidden layer output matrix; |
| β ← Compute the output weights by formula (4); |
| Output: β |
The basic principle of EM is that every feasible solution is compared to a charged particle and the charge of each particle is calculated by the value of the preparative optimization objective function [16]. The charge determines not only the type of the force between two particles (either attraction or repulsion), but also the strength of the force. Under the action of attraction and repulsion forces, the population moves to a new generation. The whole process of particle movement under the force in a population is shown in Figure 1. As can be clearly seen from Figure 1, the blue particle is subject to the forces of other particles in the population, both attraction and repulsion. Besides, the optimal particle in the population will always attract other particles to move towards it. Specifically, the EM algorithm mainly includes the following four steps.
(1) Initialization: m particles are randomly selected from the feasible region as initial population. Each coordinate of the particle is uniformly distributed between corresponding upper and lower bounds. Then, the objective function value is calculated for each particle and the particle with the best objective function value is stored in Xbest.
(2) Local search: The procedure of the local search conducted on a single particle is to improve the solution obtained. Each dimension of the current optimal particle Xbest is searched according to a certain step size. Once a better solution is found, the optimal particle is updated. The effective local information obtained from this procedure can contribute to the abilities of EM of both exploration and exploitation.
(3) Calculation of the resultant force: The magnitude of the force of particle i is strongly related to its charge qi, which can be calculated by formula (5):
| qi=exp(−nf(Xi)−f(Xbest)m∑k=1(f(Xk)−f(Xbest))),∀i | (5) |
The resultant force Fi exerted on particle i can be calculated as follows:
| Fi=m∑j≠i{(Xj−Xi)qiqj‖Xj−Xi‖2,if f(Xj)<f(Xi)(Xi−Xj)qiqj‖Xj−Xi‖2,if f(Xj)⩾ | (6) |
According to the above formula, the particle with better and poorer objective function values attracts and repulses other particles, respectively. The better the objective function value is, the stronger the attraction will be, and vice versa.
(4) Movement of the population: After calculating the resultant force, the particles are moved in the direction of the force, thus forming a new generation of population. The direction and step of the movement are determined by the following formula:
| (7) |
λ is a random number of 0 to 1, which guarantees that the particles with a nonzero probability move to unvisited regions. RNG is a movable range between the upper and lower sectors.
The pseudo-code of EM algorithm is as follows:
| EM (M, MAXITER, LSITER, δ) |
| M is the number of particles; MAXITER is the maximum number of iterations; LSITER is the maximum number of iterations in local search; δ is the local search parameter, δ∈[0, 1]. |
| Initialization of the population and parameters |
| iteration 0 |
| Do { |
| Local search (LSITER, δ) |
| q Calculation of charge of each particle (f(X)) |
| F Calculation of resultant force (q) |
| Move each particle (F) |
| iteration iteration + 1 |
| } while iteration < MAXITER |
From the introduction of ELM, it can be seen that the input weights and hidden layer biases are randomly generated at the initialization stage. The network constructed in this way may give rise to a problem of overfitting. More specifically, ELM usually requires more hidden neurons than conventional neural networks to achieve the expected performance. Larger network size results in longer running time of the testing phase of ELM, which may hinder its efficient development in some test time sensitive scenarios [25]. To solve this problem, an improved approach designated as DAEM-ELM, which combines EM with ELM, is proposed in this paper. This new ELM adopts a novel EM called DAEM to optimize the input weights and biases of ELM to improve the generalization performance and the conditioning of the SLFN. In this section, we will first provide a detailed description of the DAEM algorithm, and then present the DAEM-ELM algorithm.
The convergence speed of EM slows down gradually during the iterations and the algorithm easily falls into the local optimal solution, that is, prematurity. In addition, the position and number of adjacent particles influence the step length and position update of the population according to formula (1). But it remains unclear under what conditions individual particles can be defined as adjacent to each other. In order to solve these problems, we propose an improved EM algorithm called DAEM, by incorporating some theories of dragonfly algorithm (DA) into EM.
First, the adjacency of particles is defined. A neighborhood (circle in a 2D, sphere in a 3D space, or hypersphere in an nD space) with a certain radius r is assumed around each particle [24]. If the Euclidean distance between particle i and particle j is less than r, particle i and particle j are considered as adjacent. In order to accelerate the convergence speed, the radius r increases with increasing number of iterations. The specific calculation formula of r is as follows:
| (8) |
where iter is the number of current iterations, Max_iteration is the maximum number of iterations, ub is the upper limit of variables, and lb is the lower limit of variables. The neighborhood can be represented as (r1, r2, ..., rn) and n represents the number of dimensions.
Taking particle i as an example, it can be expressed as Xi = (xi1, xi2, ..., xin), and then the neighborhood range of particle i in the j-th dimension (j = 1, 2, …, n) is [xij-rj, xij + rj]. When another particle Xk is within the neighborhood range of particle i of each dimension, it is considered that Xk is adjacent to Xi.
For different problems, the fixed solution updating strategy of EM may not be always reasonable, and cannot guarantee the discovery of global optimal solution or approximate global optimal solution. Therefore, DAEM provides three different solution updating strategies motivated by DA. In this way, the suitable updating strategy can be selected according to the prior information of different problems. Besides, variable searching step is adopted to solve the conflicts between solution accuracy and computation time in the optimization process. The three updating strategies are described as follows.
Strategy 1: when the distance between the current particle and the optimal particle in a certain dimension is smaller than the neighborhood radius, and there are adjacent particles in the neighborhood of the current particle, the updated formula of the particle is as follows:
| (9) |
where c = 0.9-iter × (0.5/Max_iteration), η is a random number between [0, 1]. Fi is still calculated by formula (6), but only the particles in the neighborhood instead of the entire population are considered.
Strategy 2: when the distance between the current particle and the optimal particle in a certain dimension is smaller than the neighborhood radius, and there is no particle adjacent to the current particle in its neighborhood, the updated formula of the particle is as follows:
| (10) |
| (11) |
| (12) |
| (13) |
where r1 and r2 are two random numbers in [0, 1], and α is a constant (equal to 1.5 in DAEM).
Strategy 3: when the distance between the current particle and the optimal particle in any dimension is greater than the neighborhood radius, the updated formula of the particle is as follows:
| (14) |
| (15) |
where and are random numbers in [0, 1], , , Xworst and Xbest represent the worst and best particle of current population respectively, l is a random numbers in [0, 1], L is the difference between the position of Xbest and position of Xi when Xbest is in the neighborhood of Xi, otherwise L is a m-dimensional vector of 0's, e = 0.1 – iter × (0.2/Max_iteration) (when e < 0, let e = 0), C is an n-dimensional vector generated randomly.
The basic process of DAEM-ELM is described below. Data samples are divided into training sample set and testing sample set. The training set is trained by DAEM to build the classification model. Then, the testing set is tested by the obtained classification model. The detailed steps of the proposed method are as follows.
Firstly, the population is randomly generated. Each particle in the population is composed of a set of input weights and hidden biases. The specific coding form is as follows:
| (16) |
where N is the number of input neurons, and L is the number of hidden neurons, that is, the dimension of each particle is (N + 1) × L. All components in the particle are randomly initialized within the range of [-1, 1].
Secondly, for each particle, the corresponding output weights are computed according to formula (4). Then, the fitness of each particle is evaluated. The fitness function formula of DAEM-ELM is:
| (17) |
where N is the number of training samples, and MCRi is the misclassification of the algorithm.
Neural network training should not solely rely on the misclassification of training set as the fitness function, because a higher training accuracy does not guarantee a higher test accuracy. It has been reported that neural networks tend to have better generalization performance with weights of smaller norms [26,27]. Hence, the output weights are also considered in selection strategy design in order to further enhance the performance of our algorithm. It is stipulated that when the fitness values of different particles are similar, the particle with smaller norm of output weights is chosen as a better solution. The algorithm introduces the tolerance rate λ to meet this requirement and the effect is better when λ is set to 0.04 as validated by experiment. Besides, to reduce the computational complexity of the algorithm, the following regulation formula of current optimal solution is used only after each iteration (the fitness function is still the misclassification, and only the tolerance rate is added for adjustment).
| (18) |
| (19) |
where f (Xi), f (Xibest) and f (Xgbest) are the corresponding fitness values for the i-th particle, the best position of the i-th particle and global best position of all particles, respectively. are the corresponding output weights obtained by MP generalized inverse when the input weights are set as the i-th particle, the best position of the i-th particle and global best position of all particles, respectively.
Thirdly, the k-fold cross-validation (k-fold CV) method is applied in order to get an unbiased estimate of the generalization accuracy and make full use of samples in case of insufficient sample. In k-fold CV, the data sample set is divided into k mutually disjoint subsets (approximately equal in size), such as S1, S2, …, Sk, and then DAEM-ELM is performed for k iterations. Sk is selected as the testing set and the rest of subsets are used as training set in the i-th iteration. Here, the parameter value of k is 5 and the final classification results are the average value of five iterations.
The pseudo-code of DAEM-ELM is as follows:
| DAEM-ELM: Performance estimation by k-fold CV where k=5; MAXITER is the maximum number of iterations. |
| begin |
| for i=1:k |
| Training set=k-1 subsets Testing set=remaining subsets |
| begin DAEM |
| Initialize the population with random numbers |
| iteration 0 |
| Do{ |
| Train the ELM on the training set |
| Calculate fitness value |
| Update the position of each particle |
| iteration iteration + 1 |
| } while iteration < MAXITER |
| end DAEM |
| Achieve the optimal input weights and hidden bias from the best |
| solution |
| Test the ELM with the optimal input weights and hidden bias |
| end |
| Return the average classification accuracy and standard deviation of |
| ELM |
| end |
In this section, the performance of the proposed algorithm is evaluated on eight real-world classification problems (Thyroid, WDBC, Wine, Bupa Liver, Australian, Breast Cancer, Parkinson and Iris), and all these data-sets are taken from the University of California Irvine (UCI) repository [28]. The specification of these datasets is listed in Table 1.
| Data-sets | Instances | Attributes | Classes | Missing Value |
| Thyroid | 215 | 5 | 3 | N |
| Parkinson | 195 | 22 | 2 | N |
| Iris | 150 | 4 | 3 | Y |
| Bupa Liver | 345 | 6 | 2 | N |
| Australian | 690 | 14 | 2 | Y |
| Breast Cancer | 699 | 9 | 2 | Y |
| Wine | 178 | 13 | 3 | N |
| WDBC | 569 | 30 | 2 | N |
DownLoad:
CSV
Iris, Australian and Breast cancer datasets have missing values (A data-set contains a certain number of instances, and an instance includes several attributes. If some instance in a data-set lack some attributes, it is said that the data-set has missing values.). The missing categorical attributes are replaced by the mode of the attributes, and the missing continuous attributes are replaced by the mean of the attributes in order to ensure the integrity of the sample data. Besides, normalization is employed to avoid the influence of the feature values in larger numerical ranges on those in smaller numerical ranges, which can also reduce the computational complexity. Every feature value can be normalized by scaling them into the interval of [-1, 1] according to:
| (20) |
where x′ is the normalization value, x is the original value, mini is minimum value of feature i and maxi is maximum value of feature i.
The results obtained with the proposed algorithm are then presented and analyzed, and compared with those obtained using other related algorithms. The parameters in all algorithms of experiments are determined by trial and error. For DAEM-ELM, the maximum optimization epochs are 70, and the population size is 50. The sigmoid function g(x) = 1/(1 + exp(-x)) is adopt as the ELM activation function to compute the hidden layer output matrix. All the programs are run in MATLAB 7.0 environment.
The performance of DAEM-ELM algorithm is tested with the number of hidden neurons increasing from 5 to 40 at a step size of 5. The reason for choosing this range is that an excessive number of hidden neurons will lead to an overfitting problem for ELM. In addition, better parameters more suitable for associated networks can be found with the DAEM algorithm. Hence, the ELM only needs a small number of hidden neurons to obtain better results. The experimental results are shown in Table 2.
| Hidden Neurons | Accuracy (%) | Dataset | |||||||
| Thyroid | Parkinson | Iris | Bupa Liver | Australian | Breast Cancer | Wine | WDBC | ||
| 5 | Training | 94.19 | 88.46 | 98.31 | 78.62 | 76.27 | 94.61 | 98.89 | 97.80 |
| Testing | 93.02 | 82.05 | 97.33 | 75.36 | 76.09 | 94.38 | 95.94 | 94.83 | |
| 10 | Training | 97.67 | 89.10 | 98.33 | 80.80 | 82.25 | 94.79 | 99.26 | 98.01 |
| Testing | 93.02 | 87.17 | 92.33 | 71.01 | 74.64 | 92.41 | 98.33 | 96.83 | |
| 15 | Training | 98.49 | 93.59 | 98.33 | 81.88 | 83.88 | 95.68 | 99.29 | 97.17 |
| Testing | 94.41 | 87.18 | 91.67 | 80.65 | 74.64 | 94.41 | 94.52 | 96.56 | |
| 20 | Training | 98.84 | 92.95 | 98.33 | 81.88 | 83.88 | 99.04 | 100 | 98.19 |
| Testing | 95.35 | 89.74 | 86.67 | 68.41 | 73.82 | 97.57 | 97.25 | 96.73 | |
| 25 | Training | 98.84 | 93.59 | 98.33 | 82.25 | 87.50 | 96.05 | 100 | 97.80 |
| Testing | 95.35 | 87.18 | 83.33 | 71.01 | 86.96 | 93.82 | 95.94 | 96.73 | |
| 30 | Training | 98.84 | 96.15 | 98.33 | 82.03 | 90.22 | 96.97 | 100 | 97.80 |
| Testing | 93.02 | 87.18 | 80.00 | 70.72 | 86.96 | 93.82 | 91.77 | 96.73 | |
| 35 | Training | 98.96 | 95.51 | 98.33 | 82.97 | 91.30 | 97.13 | 100 | 98.19 |
| Testing | 95.35 | 86.87 | 90.00 | 69.66 | 89.13 | 95.08 | 94.52 | 96.56 | |
| 40 | Training | 98.84 | 96.15 | 98.33 | 80.07 | 83.88 | 97.31 | 100 | 99.12 |
| Testing | 90.70 | 92.31 | 83.33 | 71.01 | 82.61 | 92.97 | 94.38 | 96.56 | |
DownLoad:
CSV
As shown in Table 2, the accuracy rate does not simply increase with the number of hidden neurons. When the hidden neurons increase to a certain number, further increase will lead to a decline in accuracy. Secondly, the optimal number of hidden neurons varies for different problems. Specifically, the optimal number is 20, 25 and 35 for the Thyroid dataset, and we chose the smallest number (20) to reduce the computational time of the network. For the Parkinson, Iris, Bupa Liver, Australian, Breast Cancer, Wine and WDBC datasets, the optimal number is 40, 5, 15, 35, 20, 10 and 10, respectively.
Table 3 and Table 4 show the results achieved with all seven investigated methods (ABC-ELM [29], ELM, PSO-ELM [29], IPSO-ELM [29], E-ELM [29], LM and DAEM-ELM) for the eight benchmark classification datasets based on five trials. The last column of Table 3 shows the smallest number of hidden neurons to be used in order to achieve the best results. It is evident that a higher accuracy rate and a smaller number of hidden neurons represent a better mode. In addition, to test the accuracy, the nearly optimal number of hidden neurons for these algorithms and the standard deviations are shown in the table in the form of mean ± standard deviation. From Table 3 and Table 4, we can draw the following conclusions.
| Dataset | Algorithm | Accuracy (%) | Hidden Neurons | |
| Training | Testing | |||
| Thyroid | ABC-ELM | 98.79 | 94.97±1.44 | 15 |
| ELM | 96.84 | 92.93±3.98 | 30 | |
| PSO-ELM | 97.92 | 94.14±3.67 | 30 | |
| IPSO-ELM | 98.33 | 94.31±2.65 | 25 | |
| E-ELM | 98.10 | 92.74±3.02 | 40 | |
| LM | 95.70 | 91.07±3.41 | 35 | |
| DAEM-ELM | 98.84 | 95.35 | 20 | |
| Parkinson | ABC-ELM | 95.12 | 89.11±3.02 | 15 |
| ELM | 92.28 | 86.15±5.79 | 40 | |
| PSO-ELM | 93.66 | 87.59±4.70 | 30 | |
| IPSO-ELM | 93.95 | 88.10±4.62 | 25 | |
| E-ELM | 94.17 | 87.08±5.70 | 35 | |
| LM | 89.24 | 82.38±4.65 | 35 | |
| DAEM-ELM | 96.15 | 91.80±2.30 | 40 | |
| Iris | ABC-ELM | 97.63 | 96.68±1.83 | 15 |
| ELM | 96.00 | 95.42±2.45 | 20 | |
| PSO-ELM | 96.38 | 95.89±1.13 | 15 | |
| IPSO-ELM | 96.76 | 96.13±1.64 | 10 | |
| E-ELM | 98.81 | 95.20±3.13 | 30 | |
| LM | 98.74 | 96.00±2.67 | 10 | |
| DAEM-ELM | 98.31 | 97.33±2.67 | 5 | |
| Bupa Liver | ABC-ELM | 77.94 | 72.83±4.21 | 15 |
| ELM | 76.58 | 71.30±5.14 | 30 | |
| PSO-ELM | 77.18 | 71.54±5.26 | 25 | |
| IPSO-ELM | 77.40 | 71.72±5.33 | 25 | |
| E-ELM | 76.26 | 71.19±5.70 | 20 | |
| LM | 74.91 | 69.37±6.04 | 35 | |
| DAEM-ELM | 81.88 | 80.65±0.41 | 15 | |
DownLoad:
CSV
| Dataset | Algorithm | Accuracy (%) | Hidden Neurons | |
| Training | Testing | |||
| Australian | ABC-ELM | 90.74 | 87.38±1.61 | 15 |
| ELM | 87.50 | 85.35±3.20 | 30 | |
| PSO-ELM | 89.37 | 86.04±2.31 | 40 | |
| IPSO-ELM | 89.65 | 86.41±2.72 | 15 | |
| E-ELM | 89.51 | 86.03±2.80 | 20 | |
| LM | 87.82 | 85.97±2.77 | 40 | |
| DAEM-ELM | 91.30 | 89.13±0.21 | 35 | |
| Breast Cancer | ABC-ELM | 98.54 | 96.97±1.09 | 10 |
| ELM | 97.42 | 96.05±1.02 | 40 | |
| PSO-ELM | 97.16 | 96.31±1.25 | 35 | |
| IPSO-ELM | 98.25 | 97.18±1.33 | 25 | |
| E-ELM | 97.88 | 96.45±1.67 | 35 | |
| LM | 96.21 | 95.96±2.24 | 40 | |
| DAEM-ELM | 99.04 | 97.54±0.35 | 20 | |
| wine | ABC-ELM | 99.97 | 98.43±1.11 | 10 |
| ELM | 99.86 | 97.98±2.08 | 25 | |
| PSO-ELM | 100 | 97.63±2.27 | 15 | |
| IPSO-ELM | 100 | 97.82±2.01 | 15 | |
| E-ELM | 100 | 98.02±1.69 | 25 | |
| LM | 99.40 | 98.05±2.55 | 30 | |
| DAEM-ELM | 99.26 | 98.33±1.33 | 10 | |
| WDBC | ABC-ELM | 98.85 | 96.82±1.23 | 10 |
| ELM | 96.43 | 96.13±1.64 | 30 | |
| PSO-ELM | 97.49 | 96.28±1.60 | 20 | |
| IPSO-ELM | 97.96 | 96.54±1.51 | 10 | |
| E-ELM | 98.03 | 96.10±1.93 | 20 | |
| LM | 96.11 | 95.17±2.22 | 30 | |
| DAEM-ELM | 98.01 | 96.83±0.42 | 10 | |
DownLoad:
CSV
For the Thyroid dataset, although the number of hidden neurons of DAEM-ELM is not the smallest (only slightly bigger than that of ABC-ELM), its classification accuracy is the highest among seven methods and the standard deviation of the acquired performance is also the smallest (equal to 0), indicating the consistency and stability of the proposed method. The results of the Thyroid dataset are shown in Figure 2(a).
For the Parkinson dataset, ABC-ELM has the fewest hidden neurons. But the proposed method outperforms other six methods in terms of classification accuracy by around 3% (more than 9% compared with that of LM). Besides, DAEM-ELM also has the smallest standard deviation among these methods. Figure 2(b) shows the accuracy obtained on the Parkinson dataset.
For the Iris dataset, DAEM-ELM only needs five hidden neurons to achieve the highest classification accuracy. In this way, the proposed method achieves both the highest classification accuracy and the most compact network structure. The results of the Iris dataset are shown in Figure 2(c).
For the Bupa Liver dataset and WDBC dataset, the proposed method outperforms others in all cases (classification accuracy, number of hidden neurons, standard deviation), as shown in Figure 2(d) and Figure 2(h).
For the Australian dataset and Breast Cancer dataset, the DAEM-ELM still maintains the highest classification accuracy and the smallest standard deviation with a medium number of hidden neurons. These results are also confirmed in Figure 2(e) and Figure 2(f).
For the Wine dataset, ABC-ELM is the algorithm with the best performance. The proposed algorithm has the same number of hidden neurons as ABC-ELM, and only slightly poorer performance in accuracy and standard deviation. Figure 2(g) illustrates the results on the Wine dataset.
In summary, the DAEM-ELM algorithm can achieve better performance by using DAEM to select the input weights and hidden biases of the SLFN than the ELM, PSO-ELM, IPSO-ELM, E-ELM and LM algorithms, indicating that the optimal network structure tuned by the DAEM algorithm contributes greatly to the reduction of hidden neurons in the models and a reasonable generalization performance for these datasets.
In this paper, a novel extreme learning machine based on electromagnetism-like mechanism (DAEM-ELM) is proposed. In the new algorithm, an improved EM is used to optimize the input weights and hidden biases, and minimum norm least-square scheme is employed to analytically determine the output weights. In the optimization process, the improved EM considers not only the misclassification but also the norm of the output weights as well as constrains the input weights and hidden biases within a reasonable range. In addition, the 5-fold CV method is adopted to prevent the overfitting problem. Experimental results show that DAEM-ELM outperforms other methods (original ELM, PSO-ELM, IPSO-ELM, E-ELM and LM) and has a more compact network structure. It is also confirmed that due to the selection of optimal parameters, the results are more stable with fewer hidden neurons. Hence, it can be concluded the developed DAEM-ELM algorithm can be a feasible and effective algorithm for classification problems. Future research work will be focused on the identification of the optimal hidden neuron number, input weights and hidden biases at the same time.
This research work is supported in part by National Natural Science Foundation of China (NSFC) under Grant No. 61603145.
All authors declare no conflict of interest in this paper.
| [1] |
Kandoth C, McLellan MD, Vandin F, et al. (2013) Mutational landscape and significance across 12 major cancer types. Nature 502: 333-339. doi: 10.1038/nature12634
|
| [2] |
Rahman N (2014) Realizing the promise of cancer predisposition genes. Nature 505: 302-308. doi: 10.1038/nature12981
|
| [3] |
Aronson S, Rehm H (2015) Building the foundation for genomics in precision medicine. Nature 526: 336-342. doi: 10.1038/nature15816
|
| [4] | Wishart DS, Mandal R, Stanislaus A, et al. (2016) Cancer metabolomics and the human metabolome database. Metabolomics 6: E10. |
| [5] |
Warburg O (1956) On the origin of cancer cells. Science 123: 309-314. doi: 10.1126/science.123.3191.309
|
| [6] |
Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the Warburg effect: The metabolic requirements of cell proliferation. Science 324: 1029-1033. doi: 10.1126/science.1160809
|
| [7] |
Seyfried TN, Flores RE, Poff AM, et al. (2014) Cancer as a metabolic disease: implications for novel therapeutics. Carcinogenesis 35: 515-527. doi: 10.1093/carcin/bgt480
|
| [8] |
Seyfried TN (2015) Cancer as a mitochondrial metabolic disease. Front Cell Develop Biol 3: 43. doi: 10.3389/fcell.2015.00043
|
| [9] |
Vafai SB, Mootha VK (2012) Mitochondrial disorders as windows into an ancient organelle. Nature 491: 374-383. doi: 10.1038/nature11707
|
| [10] |
Zhang J, Pavlova NN, Thompson CB (2017) Cancer cell metabolism: the essential role of the nonessential amino acid, glutamine. EMBO J 36: 1302-1315. doi: 10.15252/embj.201696151
|
| [11] |
Clunton AA, Lukey MJ, Cerione RA, et al. (2017) Glutamine metabolism in cancer: understanding the heterogeneity. Trends Cancer 3: 169-180. doi: 10.1016/j.trecan.2017.01.005
|
| [12] |
Martinez-Outschoorn UE, Peiris-Pagés M, Pestell RG, et al. (2017) Cancer metabolism: a therapeutic perspective. Nat Rev Clin Oncol 14: 11-31. doi: 10.1038/nrclinonc.2016.60
|
| [13] |
Spinelli JB, Yoon H, Ringel AE, et al. (2017) Metabolic recycling of ammonia via glutamate dehydrogenase supports breast cancer biomass. Science 358: 941-946. doi: 10.1126/science.aam9305
|
| [14] |
Mattaini KR, Sullivan MR, Vander Heiden MG (2016) The importance of serine metabolism in cancer. J Cell Biol 214: 249-257. doi: 10.1083/jcb.201604085
|
| [15] | Danenberg PV (1977) Thymidylate synthetase—A target enzyme in cancer chemotherapy. Biochim Biophys Acta 473: 73-92. |
| [16] |
Mathews CK (2015) Deoxyribonucleotide metabolism, mutagenesis and cancer. Nat Rev Cancer 15: 528-539. doi: 10.1038/nrc3981
|
| [17] |
Irwin CR, Hitt MM, Evans DH (2017) Targeting nucleotide biosynthesis: a strategy for improving the oncolytic potential of DNA viruses. Front Oncol 7: 229. doi: 10.3389/fonc.2017.00229
|
| [18] |
Shay JW (2016) Role of telomeres and telomerase in aging and cancer. Cancer Discovery 6: 584-593. doi: 10.1158/2159-8290.CD-16-0062
|
| [19] |
Sahin E, Colla S, Liesa M, et al. (2011) Telomere dysfunction induces metabolic and mitochondrial compromise. Nature 470: 359-365. doi: 10.1038/nature09787
|
| [20] |
Houtkooper RH, Mouchiroud L, Ryu D, et al. (2013) Mitochondrial protein imbalance as a conserved longevity mechanism. Nature 497: 451-457. doi: 10.1038/nature12188
|
| [21] |
Vogelstein B, Papadopoulos N, Velculescu VE, et al. (2013) Cancer genome landscapes. Science 339: 1546-1558. doi: 10.1126/science.1235122
|
| [22] | Gasparre G, Porcelli AM, Lenaz G, et al. (2014) Relevance of mitochondrial genetics and metabolism in cancer development. Mitochondria Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press, 235-251. |
| [23] |
Troulinaki K, Bano D (2012) Mitochondrial deficiency: a double-edged sword for aging and neurodegeneration. Front Genet 3: 244. doi: 10.3389/fgene.2012.00244
|
| [24] | Bender DA (2014) Micronutrients. Introduction to nutrition and metabolism Boca Raton, NW: CRC Press, Taylor & Francis Group, 343-349. |
| [25] |
Gholson RK (1966) The pyridine nucleotide cycle. Nature 212: 933-935. doi: 10.1038/212933a0
|
| [26] |
Rechsteiner M, Catanzarite V (1974) The biosynthesis and turnover of nicotinamide adenine dinucleotide in enucleated culture cells. J Cell Physiol 84: 409-422. doi: 10.1002/jcp.1040840309
|
| [27] |
Tanuma S, Sato A, Oyama T, et al. (2016) New insights into the roles of NAD+-poly (ADP-ribose) metabolism and poly (ADP-ribose) glycohydrolase. Curr Protein Pep Sci 17: 668-682. doi: 10.2174/1389203717666160419150014
|
| [28] |
Chen YR (2013) Mitochondrial dysfunction. Basis of oxidative stress: chemistry, mechanisms, and disease pathogenesis Hoboken, NJ: John Wiley & Sons, Inc, 123-135. doi: 10.1002/9781118355886.ch6
|
| [29] | Wünschiers R (2012) Carbohydrate Metabolism and Citrate Cycle. Biochemical Pathways: An Atlas of Biochemistry and Molecular Biology, 2nd ed Hoboken, NJ: John Wiley & Sons, Inc, 37-58. |
| [30] | Wünschiers R (2012) Nucleotides and Nucleosides. Biochemical Pathways: An Atlas of Biochemistry and Molecular Biology Hoboken, NJ: John Wiley & Sons, Inc, 124-133. |
| [31] |
Chaudhuri AR, Nussenzweig A (2017) The multifaceted roles of PARP1 in DNA repair and chromatin remodeling. Nat Rev Mol Cell Biol 18: 610-621. doi: 10.1038/nrm.2017.53
|
| [32] |
Maruta H, Okita N, Takasawa R, et al. (2007) The Involvement of ATP produced via (ADP-ribose) (n) in the maintenance of DNA replication apparatus during DNA Repair. Biol Pharm Bull 30: 447-450. doi: 10.1248/bpb.30.447
|
| [33] |
Wright RH, Lioutas A, Le Dily F, et al. (2016) ADP-ribose-derived nuclear ATP synthesis by NUDIX5 is required for chromatin remodeling. Science 352: 1221-1225. doi: 10.1126/science.aad9335
|
| [34] |
German NJ, Haigis MC (2015) Sirtuins and the metabolic hurdles in cancer. Current Biol 25: R569-R583. doi: 10.1016/j.cub.2015.05.012
|
| [35] |
O'Callaghan C, Vassilopoulos A (2017) Sirtuins at the crossroads of stemness, aging, and cancer. Aging Cell 16: 1208-1218. doi: 10.1111/acel.12685
|
| [36] | Hsu WW, Wu B, Liu WR (2016) Sirtuins 1 and 2 are universal histone deacetylases. ASC Chem Biol 11: 792-799. |
| [37] |
Tan B, Young DA, Lu ZH, et al. (2012) Pharmacological inhibition of nicotinamide phosphoribosyltransferase (NAMPT), an enzyme essential for NAD+ biosynthesis, in human cancer cells. J Biol Chem 288: 3500-3511. doi: 10.1074/jbc.M112.394510
|
| [38] |
Cambronne XA, Stewart ML, Kim D, et al. (2016) Biosensor reveals multiple sources for mitochondrial NAD+. Science 352: 1474-1477. doi: 10.1126/science.aad5168
|
| [39] |
Berridge G, Cramer R, Galione A, et al. (2002) Metabolism of the novel Ca2+-mobilizing messenger nicotinic acid-adenine dinucleotide phosphate via a 2′-specific Ca2+-dependent phosphatase. Biochem J 365: 295-301. doi: 10.1042/bj20020180
|
| [40] |
Tran MT, Zsengeller ZK, Berg AH, et al. (2016) PGC1α drives NAD biosynthesis linking oxidative metabolism to renal protection. Nature 531: 528-532. doi: 10.1038/nature17184
|
| [41] |
Gujar AD, Le S, Mao DD, et al. (2016) An NAD+-dependent transcriptional program governs self-renewal and radiation resistance in glioblastoma. Proc Natl Acad Sci USA 113: E8247-E8256. doi: 10.1073/pnas.1610921114
|
| [42] |
Zhao H, Tang W, Chen X, et al. (2017) The NAMPT/E2F2/SIRT1 axis promotes proliferation and inhibits p53-dependent apoptosis in human melanoma cells. Biochem Biophys Res Commun 493: 77-84. doi: 10.1016/j.bbrc.2017.09.071
|
| [43] |
Mutz CN, Schwentner R, Aryee DNT, et al. (2017) EWS-FLI1 confers exquisite sensitivity to NAMPT inhibition in Ewing sarcoma cells. Oncotarget 8: 24679-24693. doi: 10.18632/oncotarget.14976
|
| [44] |
Tan B, Dong S, Shepard RL, et al. (2015) Inhibition of nicotinamide phosphoribosyltransferase (NAMPT), an enzyme essential for NAD+ biosynthesis, leads to altered carbohydrate metabolism in cancer cells. J Biol Chem 290: 15812-15824. doi: 10.1074/jbc.M114.632141
|
| [45] |
Hufton SE, Moerkerk PT, Brandwijk R, et al. (1999) A profile of differentially expressed genes in primary colorectal cancer using suppression subtractive hybridization. FEBS Lett 463: 77-82. doi: 10.1016/S0014-5793(99)01578-1
|
| [46] |
Van Beijnum JR, Moerkerk PT, Gerbers AJ, et al. (2002) Target validation for genomics using peptide-specific phage antibodies: a study of five gene products overexpressed in colorectal cancer. Int J Cancer 101: 118-127. doi: 10.1002/ijc.10584
|
| [47] | Bi TQ, Che XM, Liao XH, et al. (2011) Over expression of Nampt in gastric cancer and chemopotentiating effects of the Nampt inhibitor FK866 in combination with fluorouracil. Oncol Rep 26: 1251-1257. |
| [48] |
Ogino Y, Sato A, Uchiumi F, et al. (2018) Cross resistance to diverse anticancer nicotinamide phosphoribosyltransferase inhibitors induced by FK866 treatment. Oncotarget 9: 16451-16461. doi: 10.18632/oncotarget.24731
|
| [49] |
Chowdhry S, Zanca C, Rajkumar U, et al. (2019) NAD metabolic dependency in cancer is shaped by gene amplification and enhancer remodelling. Nature 569: 570-575. doi: 10.1038/s41586-019-1150-2
|
| [50] |
Ulanovskaya OA, Zhul AM, Cravatt BF (2013) NNMT promotes epigenetic remodelling in cancer by creating a metabolic methylation sink. Nat Chem Biol 9: 300-306. doi: 10.1038/nchembio.1204
|
| [51] |
Eckert MA, Coscia F, Chryplewicz A, et al. (2019) Proteomics reveals NNMT as a master metabolic regulator of cancer-associated fibroblasts. Nature 569: 723-728. doi: 10.1038/s41586-019-1173-8
|
| [52] |
Cantó C, Houtkooper RH, Pirinen E, et al. (2012) The NAD+ precursor nicotinamide riboside enhances oxidative metabolism and protects against high-fat diet-induced obesity. Cell Metab 15: 838-847. doi: 10.1016/j.cmet.2012.04.022
|
| [53] |
Han X, Tai H, Wang X, et al. (2016) AMPK activation protects cells from oxidative stress-induced senescence via autophagic flux restoration and intracellular NAD elevation. Aging Cell 15: 416-427. doi: 10.1111/acel.12446
|
| [54] |
Yang Y, Sauve AA (2016) NAD+ metabolism: bioenergetics, signaling and manipulation for therapy. Biochim Biophys Acta 1864: 1787-1800. doi: 10.1016/j.bbapap.2016.06.014
|
| [55] |
Zhang H, Ryu D, Wu Y, et al. (2016) NAD+ repletion improves mitochondrial and stem cell function and enhances life span in mice. Science 352: 1436-1443. doi: 10.1126/science.aaf2693
|
| [56] |
Rehmani I, Liu F, Liu A (2013) Cell signaling and transcription. Molecular basis of oxidative stress: chemistry, mechanisms, and disease pathogenesis Hoboken, NJ: John Wiley & Sons, 179-201. doi: 10.1002/9781118355886.ch8
|
| [57] |
Yun J, Finkel T (2014) Mitohormesis. Cell Metab 19: 757-766. doi: 10.1016/j.cmet.2014.01.011
|
| [58] |
López-Otín C, Serrano M, Partridge L, et al. (2013) The hallmarks of aging. Cell 153: 1194-1217. doi: 10.1016/j.cell.2013.05.039
|
| [59] |
Santidrian AF, Matsuno-Yagi A, Ritland M, et al. (2013) Mitochondrial complex I activity and NAD+/NADH balance regulate breast cancer progression. J Clin Invest 123: 1068-1081. doi: 10.1172/JCI64264
|
| [60] |
Luo C, Lim JH, Lee Y, et al. (2016) PGC1α-mediated transcriptional axis suppresses melanoma metastasis. Nature 537: 422-426. doi: 10.1038/nature19347
|
| [61] |
Kamenisch Y, Fousteri M, Knoch J, et al. (2010) Proteins of nucleotide and base excision repair pathways interact in mitochondria to protect from loss of subcutaneous fat, a hallmark of aging. J Exp Med 207: 379-390. doi: 10.1084/jem.20091834
|
| [62] |
Guarente L (2014) Linking DNA damage, NAD+/SIRT1, and aging. Cell Metab 20: 706-707. doi: 10.1016/j.cmet.2014.10.015
|
| [63] |
Dang L, White DW, Gross S, et al. (2009) Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature 462: 739-744. doi: 10.1038/nature08617
|
| [64] |
Yan H, Parsons DW, Jin G, et al. (2009) IDH1 and IDH2 mutations in gliomas. N Engl J Med 360: 765-773. doi: 10.1056/NEJMoa0808710
|
| [65] |
Lu C, Ward PS, Kapoor GS, et al. (2012) IDH mutation impairs histonedemethylation and results in a block to cell differentiation. Nature 483: 474-478. doi: 10.1038/nature10860
|
| [66] | Uchiumi F, Larsen S, Tanuma S (2013) Transcriptional regulation of the human genes that encode DNA repair- and mitochondrial function-associated proteins. Advances in DNA Repair Rijeka, Croatia: InTech Open Access Publisher, 129-167. |
| [67] |
Gray LR, Tompkins SC, Taylor EB (2014) Regulation of pyruvate metabolism and human disease. Cell Mol Life Sci 71: 2577-2604. doi: 10.1007/s00018-013-1539-2
|
| [68] |
Behl C, Ziegler C (2014) Selected age-related disorders. Cell Aging: Molecular Mechanisms and Implications for Disease Heidelberg, Germany: Springer Briefs in Molecular Medicine, Springer Science+Business Media, 99-108. doi: 10.1007/978-3-642-45179-9_4
|
| [69] |
Bender A, Krishnan KJ, Morris CM, et al. (2006) High levels of mitochondrial DNA deletions in substantia nigra neurons in aging and Parkinson disease. Nat Genet 38: 515-517. doi: 10.1038/ng1769
|
| [70] |
Ansari A, Rahman MS, Saha SK, et al. (2017) Function of the SIRT3 mitochondrial deacetylase in cellular physiology, cancer, and neurodegenerative disease. Aging Cell 16: 4-16. doi: 10.1111/acel.12538
|
| [71] |
Salvatori I, Valle C, Ferri A, et al. (2018) SIRT3 and mitochondrial metabolism in neurodegenerative diseases. Neurochem Int 109: 184-192. doi: 10.1016/j.neuint.2017.04.012
|
| [72] |
Patel NV, Gordon MN, Connor KE, et al. (2005) Caloric restriction attenuates Abeta-deposition in Alzheimer transgenic models. Neurobiol Aging 26: 995-1000. doi: 10.1016/j.neurobiolaging.2004.09.014
|
| [73] |
Imai S, Guarente L (2010) Ten years of NAD-dependent SIR2 family deacetylases: implications for metabolic diseases. Trends Pharmacol Sci 31: 212-220. doi: 10.1016/j.tips.2010.02.003
|
| [74] |
Silva DF, Esteves AR, Oliveira CR, et al. (2017) Mitochondrial metabolism power SIRT2-dependent traffic causing Alzheimer's-disease related pathology. Mol Neurobiol 54: 4021-4040. doi: 10.1007/s12035-016-9951-x
|
| [75] |
Jung ES, Choi H, Song H, et al. (2016) p53-dependent SIRT6 expression protects Ab42-induced DNA damage. Sci Rep 6: 25628. doi: 10.1038/srep25628
|
| [76] | Blakey CA, Litt MD (2016) Epigenetic gene expression-an introduction. Epigenetic gene expression and regulation London, UK: Academic Press, Elsevier Inc, 1-19. |
| [77] |
Suvà ML, Riggi N, Bernstein BE (2013) Epigenetic reprogramming in cancer. Science 339: 1567-1570. doi: 10.1126/science.1230184
|
| [78] |
McDevitt MA (2016) Clinical applications of epigenetics. Epigenomics in health and disease San Diego, CA: Academic Press, 271-295. doi: 10.1016/B978-0-12-800140-0.00013-3
|
| [79] |
Hentze MW, Preiss T (2010) The REM phase of gene regulation. Trends Biochem Sci 35: 423-426. doi: 10.1016/j.tibs.2010.05.009
|
| [80] |
Gut P, Verdin E (2013) The nexus of chromatin regulation and intermediary metabolism. Nature 502: 489-498. doi: 10.1038/nature12752
|
| [81] |
Liu C, Vyas A, Kassab MA, et al. (2017) The role of poly ADP-ribosylation in the first wave of DNA damage response. Nucleic Acids Res 45: 8129-8141. doi: 10.1093/nar/gkx565
|
| [82] |
Bai P, Cantó C, Oudart H, et al. (2011) PARP-1 inhibition increases mitochondrial metabolism through SIRT1 activation. Cell Metab 13: 461-468. doi: 10.1016/j.cmet.2011.03.004
|
| [83] |
Baxter P, Chen Y, Xu Y, et al. (2014) Mitochondrial dysfunction induced by nuclear poly (ADP-ribose) polymerase-1: a treatable cause of cell death in stroke. Transl Stroke Res 5: 136-144. doi: 10.1007/s12975-013-0283-0
|
| [84] |
Uchiumi F, Watanabe T, Ohta R, et al. (2013) PARP1 gene expression is downregulated by knockdown of PARG gene. Oncol Rep 29: 1683-1688. doi: 10.3892/or.2013.2321
|
| [85] |
Gibson BA, Zhang Y, Jiang H, et al. (2016) Chemical genetic discovery of PARP targets reveals a role for PARP-1 in transcription elongation. Science 353: 45-50. doi: 10.1126/science.aaf7865
|
| [86] |
Uchiumi F, Fujikawa M, Miyazaki S, et al. (2013) Implication of bidirectional promoters containing duplicated GGAA motifs of mitochondrial function-associated genes. AIMS Mol Sci 1: 1-26. doi: 10.3934/molsci.2013.1.1
|
| [87] |
Desquiret-Dumas V, Gueguen N, Leman G, et al. (2013) Resveratrol induces a mitochondrial complex I dependent increase in NADH oxidation responsible for sirtuin activation in liver cells. J Biol Chem 288: 36662-36675. doi: 10.1074/jbc.M113.466490
|
| [88] | Liu J, Oberdoerffer P (2013) Metabolic modulation of chromatin: implications for DNA repair and genomic integrity. Front Genet 4: 182. |
| [89] |
Pearce EL, Poffenberger MC, Chang CH, et al. (2013) Fueling Immunity: Insights into metabolism and lymphocyte function. Science 342: 1242454. doi: 10.1126/science.1242454
|
| [90] |
Uchiumi F, Shoji K, Sasaki Y, et al. (2015) Characterization of the 5′-flanking region of the human TP53 gene and its response to the natural compound, Resveratrol. J Biochem 159: 437-447. doi: 10.1093/jb/mvv126
|
| [91] |
Di LJ, Fernandez AG, de Siervi A, et al. (2010) Transcriptional regulation of BRCA1 expression by a metabolic switch. Nat Struct Mol Biol 17: 1406-1413. doi: 10.1038/nsmb.1941
|
| [92] |
Chinnadurai G (2002) CtBP, an unconventional transcription corepressor in development and oncogenesis. Mol Cell 9: 213-224. doi: 10.1016/S1097-2765(02)00443-4
|
| [93] |
Chinnadurai G (2009) The transcription corepressor CtBP: a foe of multiple tumor suppressors. Cancer Res 69: 731-734. doi: 10.1158/0008-5472.CAN-08-3349
|
| [94] |
Shen Y, Kapfhamer D, Minnella AM, et al. (2017) Bioenergetic state regulates innate inflammatory responses through the transcriptional co-repressor CtBP. Nat Commun 8: 624. doi: 10.1038/s41467-017-00707-0
|
| [95] |
Chen YQ, Sengchanthalangsy LL, Hackett A, et al. (2000) NF-kappaB p65 (RelA) homodimer uses distinct mechanisms to recognize DNA targets. Structure 8: 419-428. doi: 10.1016/S0969-2126(00)00123-4
|
| [96] |
Yang ZF, Drumea K, Mott S, et al. (2014) GABP transcription factor (nuclear respiratory factor 2) is required for mitochondrial biogenesis. Mol Cell Biol 34: 3194-3201. doi: 10.1128/MCB.00492-12
|
| [97] |
Keckesova Z, Donaher JL, De Cock J, et al. (2017) LACTB is a tumor suppressor that modulates lipid metabolism and cell state. Nature 543: 681-686. doi: 10.1038/nature21408
|
| [98] |
Venigopal R, Jaiswal AK (1996) Nrf1 and Nrf2 positively and c-Fos and Fra1 negatively regulate the human antioxidant response element-mediated expression of NAD(P): quinone oxidoreductase1 gene. Proc Natl Acad Sci USA 93: 14960-14965. doi: 10.1073/pnas.93.25.14960
|
| [99] |
Wilson LA, Germin A, Espiritu R, et al. (2005) Ets-1 is transcriptionally up-regulated by H2O2 via an antioxidant response element. FASEB J 19: 2085-2087. doi: 10.1096/fj.05-4401fje
|
| [100] |
Wei GH, Badis G, Berger MF, et al. (2010) Genome-wide analysis of ETS-family DNA-binding in vitro and in vivo. EMBO J 29: 2147-2160. doi: 10.1038/emboj.2010.106
|
| [101] |
Houtkooper RH, Pirinen E, Auwerx J (2012) Sirtuins as regulators of metabolism and healthspan. Nat Rev Mol Cell Biol 13: 225-238. doi: 10.1038/nrm3293
|
| [102] |
Sabari BR, Zhang D, Allis CD, et al. (2017) Metabolic regulation of gene expression through histone acylations. Nat Rev Mol Cell Biol 18: 90-101. doi: 10.1038/nrm.2016.140
|
| [103] |
Limagne E, Thibaudin M, Euvrard R, et al. (2017) Sirtuin-1 activation controls tumor growth by impeding Th17 differentiation via STAT3 deacetylation. Cell Rep 19: 746-759. doi: 10.1016/j.celrep.2017.04.004
|
| [104] |
Bonkowski MS, Sinclair DA (2016) Slowing aging by design: the rise of NAD+ and sirtuin-activating compounds. Nat Rev Mol Cell Biol 17: 679-690. doi: 10.1038/nrm.2016.93
|
| [105] |
Menzies KJ, Singh K, Saleem A, et al. (2013) Sirtuin 1-mediated effects of exercise and resveratrol on mitochondrial biogenesis. J Biol Chem 288: 6968-6979. doi: 10.1074/jbc.M112.431155
|
| [106] | Sack MN, Finkel T (2014) Mitochondrial metabolism, sirtuins, and aging. Mitochondria Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press, 253-262. |
| [107] |
Gibson BA, Kraus WL (2012) New insights into the molecular and cellular functions of poly (ADP-ribose) and PARPs. Nat Rev Mol Cell Biol 13: 411-424. doi: 10.1038/nrm3376
|
| [108] |
Curtin NJ, Mukhopadhyay A, Drew Y, et al. (2012) The role of PARP in DNA repair and its therapeutic exploitation. DNA repair in cancer therapy-Molecular targets and clinical applications London, UK: Academic Press, Elsevier Inc, 55-73. doi: 10.1016/B978-0-12-384999-1.10004-6
|
| [109] |
Lord CJ, Ashworth A (2017) PARP inhibitors: Synthetic lethality in the clinic. Science 355: 1152-1158. doi: 10.1126/science.aam7344
|
| [110] |
Venkitaraman AR (2014) Cancer suppression by the chromosome custodians, BRCA1 and BRCA2. Science 343: 1470-1475. doi: 10.1126/science.1252230
|
| [111] |
Feng X, Koh DW (2013) Inhibition of poly (ADP-ribose) polymerase-1 or poly (ADP-ribose) glycohydrolase individually, but not in combination, leads to improved chemotherapeutic efficacy in HeLa cells. Int J Oncol 42: 749-756. doi: 10.3892/ijo.2012.1740
|
| [112] |
Gomes AP, Price NL, Ling AJY, et al. (2013) Declining NAD+ induces a pseudohypoxic state disrupting nuclear-mitochondrial communication during aging. Cell 155: 1624-1638. doi: 10.1016/j.cell.2013.11.037
|
| [113] |
Mouchiroud L, Houtkooper RH, Auwerx J (2013) NAD+ metabolism, a therapeutic target for age-related metabolic disease. Crit Rev Biochem Mol Biol 48: 397-408. doi: 10.3109/10409238.2013.789479
|
| [114] |
Williams PA, Harder JM, Foxworth NE, et al. (2017) Vitamin B3 modulates mitochondrial vulnerability and prevents glaucoma in aged mice. Science 355: 756-760. doi: 10.1126/science.aal0092
|
| [115] |
Sueishi Y, Nii R, Kakizaki N (2017) Resveratrol analogues like piceatannol are antioxidants as quantitatively demonstrated through the high scavenging ability against reactive oxygen species and methyl radical. Bioorg Med Chem Lett 27: 5203-5206. doi: 10.1016/j.bmcl.2017.10.045
|
| [116] |
Johnson SC, Yanos ME, Kayser EB, et al. (2013) mTOR inhibition alleviates mitochondrial disease in a mouse model of Leigh syndrome. Science 342: 1524-1528. doi: 10.1126/science.1244360
|
| [117] |
Fendt SM, Bell EL, Keibler MA, et al. (2013) Metformin decreases glucose oxidation and increases the dependency of prostate cancer cells on reductive glutamine metabolism. Cancer Res 73: 4429-4438. doi: 10.1158/0008-5472.CAN-13-0080
|
| [118] |
Yamato M, Kawano K, Yamanaka Y, et al. (2016) TEMPOL increases NAD+ and improves redox imbalance in obese mice. Redox Biol 8: 316-322. doi: 10.1016/j.redox.2016.02.007
|
| [119] |
Jackson SJT, Singletary KW, Murphy LL, et al. (2016) Phytonutrients differentially stimulate NAD(P)H: quinone oxidoreductase, inhibit proliferation, and trigger mitotic catastrophe in Hepa1c1c7 cells. J Med Food 19: 47-53. doi: 10.1089/jmf.2015.0079
|
| [120] |
Roubalová L, Dinkova-Kostova AT, Biedermann D, et al. (2017) Flavonolignan 2,3-dehydrosilydianin activates Nrf2 and upregulates NAD(P)H: quinone oxidoreductase 1 in Hepa1c1c7 cells. Fitoterapia 119: 115-120. doi: 10.1016/j.fitote.2017.04.012
|
| [121] |
Son MJ, Ryu JS, Kim JY, et al. (2017) Upregulation of mitochondrial NAD+ levels impairs the clonogenicity of SSEA1+ glioblastoma tumor-initiating cells. Exp Mol Med 49: e344. doi: 10.1038/emm.2017.74
|
| [122] |
Fang EF, Kassahun H, Croteau DL, et al. (2016) NAD+ replenishment improves lifespan and healthspan in ataxia telangiectasia model via mitophagy and DNA repair. Cell Metab 24: 566-581. doi: 10.1016/j.cmet.2016.09.004
|
| [123] | Takihara Y, Sudo D, Arakawa J, et al. (2018) Nicotinamide adenine dinucleotide (NAD+) and cell aging. New Research on Cell Aging and Death Hauppauge, NY: Nova Science Publishers, Inc, 131-158. |
| [124] |
Wartewig T, Kurgyis Z, Keppler S, et al. (2017) PD-1 is a haploinsufficient suppressor of T cell lymphomagenesis. Nature 552: 121-125. doi: 10.1038/nature24649
|
| [125] | Uchiumi F, Larsen S, Tanuma S (2016) Possible roles of a duplicated GGAA motif as a driver cis-element for cancer-associated genes. Understand Cancer – Research and Treatment Hong Kong: iConcept Press Ltd, 1-25. |
| [126] | Uchiumi F, Larsen S, Masumi A, et al. (2013) The putative implications of duplicated GGAA-motifs located in the human interferon regulated genes (ISGs). Genomics I-Humans, Animals and Plants Hong Kong: iConcept Press Ltd, 87-105. |
| [127] |
Uchiumi F, Arakawa J, Iwakoshi K, et al. (2016) Characterization of the 5′-flanking region of the human DNA helicase B (HELB) gene and its response to trans-resveratrol. Sci Rep 6: 24510. doi: 10.1038/srep24510
|
Figures(2)
Fumiaki Uchiumi, Akira Sato, Masashi Asai, Sei-ichi Tanuma. An NAD+ dependent/sensitive transcription system: Toward a novel anti-cancer therapy[J]. AIMS Molecular Science, 2020, 7(1): 12-28. doi: 10.3934/molsci.2020002
| Data-sets | Instances | Attributes | Classes | Missing Value |
| Thyroid | 215 | 5 | 3 | N |
| Parkinson | 195 | 22 | 2 | N |
| Iris | 150 | 4 | 3 | Y |
| Bupa Liver | 345 | 6 | 2 | N |
| Australian | 690 | 14 | 2 | Y |
| Breast Cancer | 699 | 9 | 2 | Y |
| Wine | 178 | 13 | 3 | N |
| WDBC | 569 | 30 | 2 | N |
DownLoad:
CSV
| Hidden Neurons | Accuracy (%) | Dataset | |||||||
| Thyroid | Parkinson | Iris | Bupa Liver | Australian | Breast Cancer | Wine | WDBC | ||
| 5 | Training | 94.19 | 88.46 | 98.31 | 78.62 | 76.27 | 94.61 | 98.89 | 97.80 |
| Testing | 93.02 | 82.05 | 97.33 | 75.36 | 76.09 | 94.38 | 95.94 | 94.83 | |
| 10 | Training | 97.67 | 89.10 | 98.33 | 80.80 | 82.25 | 94.79 | 99.26 | 98.01 |
| Testing | 93.02 | 87.17 | 92.33 | 71.01 | 74.64 | 92.41 | 98.33 | 96.83 | |
| 15 | Training | 98.49 | 93.59 | 98.33 | 81.88 | 83.88 | 95.68 | 99.29 | 97.17 |
| Testing | 94.41 | 87.18 | 91.67 | 80.65 | 74.64 | 94.41 | 94.52 | 96.56 | |
| 20 | Training | 98.84 | 92.95 | 98.33 | 81.88 | 83.88 | 99.04 | 100 | 98.19 |
| Testing | 95.35 | 89.74 | 86.67 | 68.41 | 73.82 | 97.57 | 97.25 | 96.73 | |
| 25 | Training | 98.84 | 93.59 | 98.33 | 82.25 | 87.50 | 96.05 | 100 | 97.80 |
| Testing | 95.35 | 87.18 | 83.33 | 71.01 | 86.96 | 93.82 | 95.94 | 96.73 | |
| 30 | Training | 98.84 | 96.15 | 98.33 | 82.03 | 90.22 | 96.97 | 100 | 97.80 |
| Testing | 93.02 | 87.18 | 80.00 | 70.72 | 86.96 | 93.82 | 91.77 | 96.73 | |
| 35 | Training | 98.96 | 95.51 | 98.33 | 82.97 | 91.30 | 97.13 | 100 | 98.19 |
| Testing | 95.35 | 86.87 | 90.00 | 69.66 | 89.13 | 95.08 | 94.52 | 96.56 | |
| 40 | Training | 98.84 | 96.15 | 98.33 | 80.07 | 83.88 | 97.31 | 100 | 99.12 |
| Testing | 90.70 | 92.31 | 83.33 | 71.01 | 82.61 | 92.97 | 94.38 | 96.56 | |
DownLoad:
CSV
| Dataset | Algorithm | Accuracy (%) | Hidden Neurons | |
| Training | Testing | |||
| Thyroid | ABC-ELM | 98.79 | 94.97±1.44 | 15 |
| ELM | 96.84 | 92.93±3.98 | 30 | |
| PSO-ELM | 97.92 | 94.14±3.67 | 30 | |
| IPSO-ELM | 98.33 | 94.31±2.65 | 25 | |
| E-ELM | 98.10 | 92.74±3.02 | 40 | |
| LM | 95.70 | 91.07±3.41 | 35 | |
| DAEM-ELM | 98.84 | 95.35 | 20 | |
| Parkinson | ABC-ELM | 95.12 | 89.11±3.02 | 15 |
| ELM | 92.28 | 86.15±5.79 | 40 | |
| PSO-ELM | 93.66 | 87.59±4.70 | 30 | |
| IPSO-ELM | 93.95 | 88.10±4.62 | 25 | |
| E-ELM | 94.17 | 87.08±5.70 | 35 | |
| LM | 89.24 | 82.38±4.65 | 35 | |
| DAEM-ELM | 96.15 | 91.80±2.30 | 40 | |
| Iris | ABC-ELM | 97.63 | 96.68±1.83 | 15 |
| ELM | 96.00 | 95.42±2.45 | 20 | |
| PSO-ELM | 96.38 | 95.89±1.13 | 15 | |
| IPSO-ELM | 96.76 | 96.13±1.64 | 10 | |
| E-ELM | 98.81 | 95.20±3.13 | 30 | |
| LM | 98.74 | 96.00±2.67 | 10 | |
| DAEM-ELM | 98.31 | 97.33±2.67 | 5 | |
| Bupa Liver | ABC-ELM | 77.94 | 72.83±4.21 | 15 |
| ELM | 76.58 | 71.30±5.14 | 30 | |
| PSO-ELM | 77.18 | 71.54±5.26 | 25 | |
| IPSO-ELM | 77.40 | 71.72±5.33 | 25 | |
| E-ELM | 76.26 | 71.19±5.70 | 20 | |
| LM | 74.91 | 69.37±6.04 | 35 | |
| DAEM-ELM | 81.88 | 80.65±0.41 | 15 | |
DownLoad:
CSV
| Dataset | Algorithm | Accuracy (%) | Hidden Neurons | |
| Training | Testing | |||
| Australian | ABC-ELM | 90.74 | 87.38±1.61 | 15 |
| ELM | 87.50 | 85.35±3.20 | 30 | |
| PSO-ELM | 89.37 | 86.04±2.31 | 40 | |
| IPSO-ELM | 89.65 | 86.41±2.72 | 15 | |
| E-ELM | 89.51 | 86.03±2.80 | 20 | |
| LM | 87.82 | 85.97±2.77 | 40 | |
| DAEM-ELM | 91.30 | 89.13±0.21 | 35 | |
| Breast Cancer | ABC-ELM | 98.54 | 96.97±1.09 | 10 |
| ELM | 97.42 | 96.05±1.02 | 40 | |
| PSO-ELM | 97.16 | 96.31±1.25 | 35 | |
| IPSO-ELM | 98.25 | 97.18±1.33 | 25 | |
| E-ELM | 97.88 | 96.45±1.67 | 35 | |
| LM | 96.21 | 95.96±2.24 | 40 | |
| DAEM-ELM | 99.04 | 97.54±0.35 | 20 | |
| wine | ABC-ELM | 99.97 | 98.43±1.11 | 10 |
| ELM | 99.86 | 97.98±2.08 | 25 | |
| PSO-ELM | 100 | 97.63±2.27 | 15 | |
| IPSO-ELM | 100 | 97.82±2.01 | 15 | |
| E-ELM | 100 | 98.02±1.69 | 25 | |
| LM | 99.40 | 98.05±2.55 | 30 | |
| DAEM-ELM | 99.26 | 98.33±1.33 | 10 | |
| WDBC | ABC-ELM | 98.85 | 96.82±1.23 | 10 |
| ELM | 96.43 | 96.13±1.64 | 30 | |
| PSO-ELM | 97.49 | 96.28±1.60 | 20 | |
| IPSO-ELM | 97.96 | 96.54±1.51 | 10 | |
| E-ELM | 98.03 | 96.10±1.93 | 20 | |
| LM | 96.11 | 95.17±2.22 | 30 | |
| DAEM-ELM | 98.01 | 96.83±0.42 | 10 | |
DownLoad:
CSV
| Data-sets | Instances | Attributes | Classes | Missing Value |
| Thyroid | 215 | 5 | 3 | N |
| Parkinson | 195 | 22 | 2 | N |
| Iris | 150 | 4 | 3 | Y |
| Bupa Liver | 345 | 6 | 2 | N |
| Australian | 690 | 14 | 2 | Y |
| Breast Cancer | 699 | 9 | 2 | Y |
| Wine | 178 | 13 | 3 | N |
| WDBC | 569 | 30 | 2 | N |
| Hidden Neurons | Accuracy (%) | Dataset | |||||||
| Thyroid | Parkinson | Iris | Bupa Liver | Australian | Breast Cancer | Wine | WDBC | ||
| 5 | Training | 94.19 | 88.46 | 98.31 | 78.62 | 76.27 | 94.61 | 98.89 | 97.80 |
| Testing | 93.02 | 82.05 | 97.33 | 75.36 | 76.09 | 94.38 | 95.94 | 94.83 | |
| 10 | Training | 97.67 | 89.10 | 98.33 | 80.80 | 82.25 | 94.79 | 99.26 | 98.01 |
| Testing | 93.02 | 87.17 | 92.33 | 71.01 | 74.64 | 92.41 | 98.33 | 96.83 | |
| 15 | Training | 98.49 | 93.59 | 98.33 | 81.88 | 83.88 | 95.68 | 99.29 | 97.17 |
| Testing | 94.41 | 87.18 | 91.67 | 80.65 | 74.64 | 94.41 | 94.52 | 96.56 | |
| 20 | Training | 98.84 | 92.95 | 98.33 | 81.88 | 83.88 | 99.04 | 100 | 98.19 |
| Testing | 95.35 | 89.74 | 86.67 | 68.41 | 73.82 | 97.57 | 97.25 | 96.73 | |
| 25 | Training | 98.84 | 93.59 | 98.33 | 82.25 | 87.50 | 96.05 | 100 | 97.80 |
| Testing | 95.35 | 87.18 | 83.33 | 71.01 | 86.96 | 93.82 | 95.94 | 96.73 | |
| 30 | Training | 98.84 | 96.15 | 98.33 | 82.03 | 90.22 | 96.97 | 100 | 97.80 |
| Testing | 93.02 | 87.18 | 80.00 | 70.72 | 86.96 | 93.82 | 91.77 | 96.73 | |
| 35 | Training | 98.96 | 95.51 | 98.33 | 82.97 | 91.30 | 97.13 | 100 | 98.19 |
| Testing | 95.35 | 86.87 | 90.00 | 69.66 | 89.13 | 95.08 | 94.52 | 96.56 | |
| 40 | Training | 98.84 | 96.15 | 98.33 | 80.07 | 83.88 | 97.31 | 100 | 99.12 |
| Testing | 90.70 | 92.31 | 83.33 | 71.01 | 82.61 | 92.97 | 94.38 | 96.56 | |
| Dataset | Algorithm | Accuracy (%) | Hidden Neurons | |
| Training | Testing | |||
| Thyroid | ABC-ELM | 98.79 | 94.97±1.44 | 15 |
| ELM | 96.84 | 92.93±3.98 | 30 | |
| PSO-ELM | 97.92 | 94.14±3.67 | 30 | |
| IPSO-ELM | 98.33 | 94.31±2.65 | 25 | |
| E-ELM | 98.10 | 92.74±3.02 | 40 | |
| LM | 95.70 | 91.07±3.41 | 35 | |
| DAEM-ELM | 98.84 | 95.35 | 20 | |
| Parkinson | ABC-ELM | 95.12 | 89.11±3.02 | 15 |
| ELM | 92.28 | 86.15±5.79 | 40 | |
| PSO-ELM | 93.66 | 87.59±4.70 | 30 | |
| IPSO-ELM | 93.95 | 88.10±4.62 | 25 | |
| E-ELM | 94.17 | 87.08±5.70 | 35 | |
| LM | 89.24 | 82.38±4.65 | 35 | |
| DAEM-ELM | 96.15 | 91.80±2.30 | 40 | |
| Iris | ABC-ELM | 97.63 | 96.68±1.83 | 15 |
| ELM | 96.00 | 95.42±2.45 | 20 | |
| PSO-ELM | 96.38 | 95.89±1.13 | 15 | |
| IPSO-ELM | 96.76 | 96.13±1.64 | 10 | |
| E-ELM | 98.81 | 95.20±3.13 | 30 | |
| LM | 98.74 | 96.00±2.67 | 10 | |
| DAEM-ELM | 98.31 | 97.33±2.67 | 5 | |
| Bupa Liver | ABC-ELM | 77.94 | 72.83±4.21 | 15 |
| ELM | 76.58 | 71.30±5.14 | 30 | |
| PSO-ELM | 77.18 | 71.54±5.26 | 25 | |
| IPSO-ELM | 77.40 | 71.72±5.33 | 25 | |
| E-ELM | 76.26 | 71.19±5.70 | 20 | |
| LM | 74.91 | 69.37±6.04 | 35 | |
| DAEM-ELM | 81.88 | 80.65±0.41 | 15 | |
| Dataset | Algorithm | Accuracy (%) | Hidden Neurons | |
| Training | Testing | |||
| Australian | ABC-ELM | 90.74 | 87.38±1.61 | 15 |
| ELM | 87.50 | 85.35±3.20 | 30 | |
| PSO-ELM | 89.37 | 86.04±2.31 | 40 | |
| IPSO-ELM | 89.65 | 86.41±2.72 | 15 | |
| E-ELM | 89.51 | 86.03±2.80 | 20 | |
| LM | 87.82 | 85.97±2.77 | 40 | |
| DAEM-ELM | 91.30 | 89.13±0.21 | 35 | |
| Breast Cancer | ABC-ELM | 98.54 | 96.97±1.09 | 10 |
| ELM | 97.42 | 96.05±1.02 | 40 | |
| PSO-ELM | 97.16 | 96.31±1.25 | 35 | |
| IPSO-ELM | 98.25 | 97.18±1.33 | 25 | |
| E-ELM | 97.88 | 96.45±1.67 | 35 | |
| LM | 96.21 | 95.96±2.24 | 40 | |
| DAEM-ELM | 99.04 | 97.54±0.35 | 20 | |
| wine | ABC-ELM | 99.97 | 98.43±1.11 | 10 |
| ELM | 99.86 | 97.98±2.08 | 25 | |
| PSO-ELM | 100 | 97.63±2.27 | 15 | |
| IPSO-ELM | 100 | 97.82±2.01 | 15 | |
| E-ELM | 100 | 98.02±1.69 | 25 | |
| LM | 99.40 | 98.05±2.55 | 30 | |
| DAEM-ELM | 99.26 | 98.33±1.33 | 10 | |
| WDBC | ABC-ELM | 98.85 | 96.82±1.23 | 10 |
| ELM | 96.43 | 96.13±1.64 | 30 | |
| PSO-ELM | 97.49 | 96.28±1.60 | 20 | |
| IPSO-ELM | 97.96 | 96.54±1.51 | 10 | |
| E-ELM | 98.03 | 96.10±1.93 | 20 | |
| LM | 96.11 | 95.17±2.22 | 30 | |
| DAEM-ELM | 98.01 | 96.83±0.42 | 10 | |
