Earticle

An extreme learning machine (ELM) is a recently proposed learning algorithm for a single-layer feed forward neural network. In this paper we studied the ensemble of ELM by using a bagging algorithm for facial expression recognition (FER). Facial expression analysis is widely used in the behavior interpretation of emotions, for cognitive science, and social interactions. This paper presents a method for FER based on the histogram of orientation gradient (HOG) features using an ELM ensemble. First, the HOG features were extracted from the face image by dividing it into a number of small cells. A bagging algorithm was then used to construct many different bags of training data and each of them was trained by using separate ELMs. To recognize the expression of the input face image, HOG features were fed to each trained ELM and the results were combined by using a majority voting scheme. The ELM ensemble using bagging improves the generalized capability of the network significantly. The two available datasets (JAFFE and CK+) of facial expressions were used to evaluate the performance of the proposed classification system. Even the performance of individual ELM was smaller and the ELM ensemble using a bagging algorithm improved the recognition performance significantly.

With the success of the digital economy and the rapid development of its technology, network security has received increasing attention. Intrusion detection technology has always been a focus and hotspot of research. A hybrid model that combines particle swarm optimization (PSO) and kernel extreme learning machine (KELM) is presented in this work. Continuous-valued PSO and binary PSO (BPSO) are adopted together to determine the parameter combination and the feature subset. A fitness function based on the detection rate and the number of selected features is proposed. The results show that the method can simultaneously determine the parameter values and select features. Furthermore, competitive or better accuracy can be obtained using approximately one quarter of the raw input features. Experiments proved that our method is slightly better than the genetic algorithm-based KELM model.

본 논문에서는 커널 Extreme Learning Machine을 기반으로 하여 최적화 기법들 중의 하나인 입자 군집 최적화 기법을 이용한 설계 기법을 제시한다. 제안된 Kernel Extreme Learning Machine은 기존에 사용되어지는 뉴럴 네트워크의 단점을 개선한 네트워크이다. 다시 말하면, 뉴럴 네트워크의 단점인 느린 학습속도를 개선한 네 트워크이다. 일반적으로 뉴럴 네트워크의 히든 노드들은 랜덤 초기화 후 오류 역전파 알고리즘을 이용하여 학습 한다. 이와 같은 오류 역전파 알고리즘은 매우 느린 학습속도를 보인다. 이와 같은 단점을 해결하기 위하여, Kernel Extreme Learning Machine의 히든 노드들은 랜덤 초기화 되고 학습되지 않고 출력층의 연결 하중만 학습 되어진다. 이와 같은 장점을 가진 Kernel Extreme Learning Machine의 구조를 최적화하기 위하여 입자 군집 최적 화 기법을 사용한다. 제안된 설계 방법을 적용하여 설계된 모델의 일반화 성능의 우수성을 보이기 위하여, 다수 의 머신러닝 데이터들을 이용하여 실험하고 실험을 통해 얻은 결과를 비교 평가하였다.

In this paper, we proposed the design method of Extreme Learning Machine which is optimized by using Particle Swarm Optimization Technique. Extreme Learning Machine is the improved version of the conventional neural networks which have a very slow learning speed based on the back-propagation algorithm. In the conventional neural networks, the connection weights between the input layer and the hidden layers are initialized randomly and then optimized by using the gradient decent method. The speed of the learning method for the conventional neural networks is slow. In order to overcome the drawback of the conventional neural networks, the connection weights of the hidden nodes are just initialized randomly and will not be optimized, and the only connection weights of the output nodes are learned by using least square estimation not the iterative learning method. In addition, we use Particle Swarm Optimization to optimize the proposed Extreme Learning Machine. Several machine learning bench-mark data sets are used to show the generalization performance of the proposed design method and to compare their performance with the other already studied models.

With social media growing fast, user-generated content (UGC) has become a key factor in influencing how consumers decide what to buy, especially in the travel and hospitality sector. But there are a lot of fake and extreme reviews, which are making it hard for customers to make the right choice and making competition in the market unfair. The present study focuses on YouTube, a platform with high global activity, and proposes a systematic solution that combines Natural Language Processing (NLP) and machine learning methods (e.g., VADER Sentiment Analysis, Support Vector Machine SVM, and LDA Topic Modelling) for identifying and filtering fake and extreme remarks in hotel reviews. This approach has been shown to enhance the automation and precision of review screening processes. Furthermore, it provides a theoretical foundation and practical methodologies to improve the online information ecology, thereby enhancing the quality of user decision-making.

본 논문에서는 극한 기계학습을 이용하는 하이브리드 균형 표본 유전자 알고리즘(hSBGA-ELM)을 기반으로 한 새로운 암 아류형 분류자를 제안하였다. 제안 된 암 아류형 분류자는 정확한 암 아류형 분류기 설계를 위해 공개 전체암지도 (Global Cancer Map)로부터 15063개의 유전자 발현 데이터를 사용합니다. 제안된 방법에서는 14가지(유방암, 전립선 암, 폐암, 대장 암, 림프종, 방광, 흑색 종, 자궁, 백혈병, 신장, 췌장, 난소, 중피종 및 CNS)의 암 아류형을 효율적으로 분류합니다. 제안 된 hSBGA-ELM은 유전자 선택 절차 및 암 아류형 분류를 하나의 프레임 워크로 단일화 한다. 제안 된 하이브리드 균형 표본 유전 알고리즘은 GCM 데이터베이스에서 이용 가능한 16,063 개의 유전자로부터 암 아류형 분류를 담당하는 축소된 강인 유전자 셋을 찾는다. 선택/축소된 유전자 세트는 익스트림 기계학습을 이용하여 암 아류형 분류기를 구성하는데 사용된다. 결과적으로, 크기가 축소된 강인 유전자 집합이 제안하는 암 아류형 분류기의 안정된 일반화 성능을 보장하게 한다. 제안 된 hSBGA-ELM은 암에 관여하는 것으로 예측되는 95개의 유전자를 발견하였으며 기존의 암 아류형 분류기와의 비교를 통해 제안 된 방법의 효율을 보여준다.

In this paper a novel cancer subtype's classifier based on Hybrid Samples Balanced Genetic Algorithm with Extreme Learning Machine (hSBGA-ELM) is presented. Proposed cancer subtype's classifier uses genes' expression data of 16063 genes from open Global Cancer Map (GCM) data base for accurate cancer subtype's classification. Proposed method efficiently classifies 14 subtypes of cancer (breast, prostate, lung, colorectal, lymphoma, bladder, melanoma, uterus, leukemia, renal, pancreas, ovary, mesothelioma and CNS). Proposed hSBGA-ELM unifies genes' selection procedure and cancer subtype's classification into one framework. Proposed Hybrid Samples Balanced Genetic Algorithm searches a reduced robust set of genes responsible for cancer subtype's classification from 16063 genes available in GCM data base. Selected reduced set of genes is used to build cancer subtype's classifier using Extreme Learning Machine (ELM). As a result, reduced set of robust genes guarantees stable generalization performance of the proposed cancer subtype's classifier. Proposed hSBGA-ELM discovers 95 genes probably responsible for cancer. Comparison with existing cancer subtype's classifiers clear indicates efficiency of the proposed method.

본 논문에서는 파킨슨 병 진단 및 바이오 표지자 검출을 위한 극한 기계학습을 결합하는 새로운 균형 표본 유전 알고리즘(SBGA-ELM)을 제안하였다. 접근법은 정확한 파킨슨 병 진단 및 바이오 표지자 검출을 위해 공개 파킨슨 병 데이터베이스로부터 22,283개의 유전자의 발현 데이터를 사용하며 다음의 두 가지 주요 단계를 포함하였다 : 1. 특징(유전자) 선택과 2. 분류단계이다. 특징 선택 단계에서는 제안된 균형 표본 유전 알고리즘에 기반하고 파킨스병 데이터베이스(ParkDB)의 유전자 발현 데이터를 위해 고안되었다. 제안된 제안 된 SBGA는 추가적 분석을 위해 ParkDB에서 활용 가능한 22,283개의 유전자 중에서 강인한 서브셋을 찾는다. 특징분류 단계에서는 정확한 파킨슨 병 진단을 위해 선택된 유전자 세트가 극한 기계학습의 훈련에 사용된다. 발견 된 강인한 유전자 서브세트는 안정된 일반화 성능으로 파킨슨 병 진단을 할 수 있는 ELM 분류기를 생성하게 된다. 제안된 연구에서 강인한 유전자 서브셋은 파킨슨병을 관장할 것으로 예측되는 24개의 바이오 표지자를 발견하는 데도 사용된다. 논문을 통해 발견된 강인 유전자 하위 집합은 SVM이나 PBL-McRBFN과 같은 기존의 파킨슨 병 진단 방법들을 통해 검증되었다. 실시된 두 가지 방법(SVM과 PBL-McRBFN)에 대해 모두 최대 일반화 성능을 나타내었다.

A novel Samples Balanced Genetic Algorithm combined with Extreme Learning Machine (SBGA-ELM) for Parkinson's Disease diagnosis and detecting bio-markers is presented in this paper. Proposed approach uses genes' expression data of 22,283 genes from open source ParkDB data base for accurate PD diagnosis and detecting bio-markers. Proposed SBGA-ELM includes two major steps: feature (genes) selection and classification. Feature selection procedure is based on proposed Samples Balanced Genetic Algorithm designed specifically for genes expression data from ParkDB. Proposed SBGA searches a robust subset of genes among 22,283 genes available in ParkDB for further analysis. In the "classification" step chosen set of genes is used to train an Extreme Learning Machine (ELM) classifier for an accurate PD diagnosis. Discovered robust subset of genes creates ELM classifier with stable generalization performance for PD diagnosis. In this research the robust subset of genes is also used to discover 24 bio-markers probably responsible for Parkinson's Disease. Discovered robust subset of genes was verified by using existing PD diagnosis approaches such as SVM and PBL-McRBFN. Both tested methods caused maximum generalization performance.

본 연구는 벡터자기회귀모형과 머신러닝 ELM 모형을 이용하여 오피스 임대료 예측력 이 우수한 모형을 파악하고자 한다. 본 연구에서는 오피스 임대료, 공실률, 회사채수익률, 산업생산지수와 건축착공실적을 변수로 사용하였으며 시간적 범위는 2003년 2분기부터 2014년 4분기까지로, 공간적 범위는 서울시 도심권역, 강남권역, 마포/여의도권역으로 설 정하였다. 분석결과, 지역별로 일부분 차이는 존재하였으나 대체적으로 오피스 임대료는 회사채수익률과 건축착공실적 충격에 부(-)의 반응을 나타낸 반면 공실률과 산업생산지수 충격에는 정(+)의 반응을 나타냈다. 머신러닝 ELM모형과 VAR모형의 평균제곱오차(MSE), 평균제곱근오차(RMSE) 및 평균절대백분율오차(MAPE)를 비교 분석한 결과, 머신러닝의 ELM모형의 예측력이 VAR모형보다 우수함을 실증적으로 알 수가 있었고 예측기간 초기 의 예측력이 우수하다는 것을 알수가 있었다.

In this study, office rent rate is predicted using vector autoregression(VAR) model and extreme learning machine(ELM), and a model with excellent predictive power is to be identified. The variables used in this study are office rent rate, vacancy rate, industrial production index, corporate bond yield and construction start performance. The temporal range is from the 2nd quarter of 2003 to the 4th quarter of 2014, and the spatial range is the downtown area, Gangnam area, Mapo/Yeouido area in Seoul. As a result of analysis, there were some differences by region, but generally, office rental rates showed a negative (-) response to the impact of corporate bond yield and construction start performance, while the vacancy rate and industrial production index impact showed a positive (+) response. . As a result of comparing and analyzing the root mean square error (RMSE), mean square error (MSE) and mean absolute percent error (MAPE) of ELM model and VAR model, it was empirically shown that the predictive power of the ELM model is superior to the VAR model. And the predictive power of the initial period of the prediction period was excellent.

As a widely used programming model for the purposes of processing large data sets, MapReduce (MR) becomes inevitable in data clusters or grids, e.g. a Hadoop environment. Load balancing as a key factor affecting the performance of map resource distribution, has recently gained high concerns to optimize. Current MR processes in the realization of distributed tasks to clusters use hashing with random modulo operations, which can lead to uneven data distribution and inclined loads, thereby obstruct the performance of the entire distribution system. In this paper, a virtual partition consistent hashing (VPCH) algorithm is proposed for the reduce stage of MR processes, in order to achieve such a trade-off on job allocation. Besides, experienced programmers are needed to decide the number of reducers used during the reduce phase of the MR, which makes the quality of MR scripts differ. So, an extreme learning method is employed to recommend potential number of reducer a mapped task needs. Execution time is also predicted for user to better arrange their tasks. According to the results, VPCH can lead to load balancing and our prediction model can provide fast prediction than SVM with similar accuracy maintained.

In this paper, we propose a new hybrid intelligent modeling using context clustering and Extreme Learning Machine (ELM) mechanism. It has been a sensitive issue that the ELM mechanism assigns initial parameters randomly, despite of its superior performance. The proposed approach focuses on initial parameters determination of the modeling to improve the accuracy of the ELM mechanism, through removing randomness of assignment. To accomplish it, a context clustering based on Gaussian Mixture Model (GMM), considering a relationship between input-output spaces will be adopted to a Radial Basis Function Network (RBFN) of the ELM. In addition, the proposed approach will reduce the randomness of results from the original ELM. Simulations and the results show usefulness of the proposed approach with improved performance accuracy.

We introduce a fast and effective algorithm extreme learning machine (ELM) and apply it to image denoising. GA-ELM algorithm we proposed uses genetic algorithm(GA) to decide weights and bias in the ELM. It has better global optimal characteristics than traditional optimal ELM algorithm. In this paper, we used GA-ELM to do image denosing researching work. Firstly, this paper uses training samples to train GA-ELM as the noise detector. Then, we utilize the well-trained GA-ELM to recognize noise pixels in target image. And at last, an adaptive weighted average algorithm is used to recover noise pixels recognized by GA-ELM. Experiment data shows that this algorithm has better performance than other denosing algorithm.

In this paper, a general voice activity detection (VAD) method based on pattern recognition is proposed, and a specific algorithm of endpoint detection is researched. In this method, the Extreme Learning Machine (ELM) and Genetic Algorithm (GA) optimization Support Vector Machine (SVM) is used as the training and recognition model. The simulation results indicates that ELM and GA-SVM have the same superior endpoint detection accuracy, and recognition time were similar, but the training time of ELM only up to a 1/2000 of the GA-SVM, the robustness of ELM and GA-SVM is greatly improved in noisy environment compare with the traditional VAD that depends on time-domain energy and zero crossing rate.

Due to its importance in many applications, the incomplete data mining has received increasing attention in recent years, but there has been little study of the cost-sensitive classification on incomplete data. Therefore this paper proposes the dynamic cost-sensitive extreme learning machine for classification of incomplete data based on the deep imputation network (DCELMIDC). Firstly, we propose an approach for incomplete data imputation based on the deep imputation network model, and offer the cost-sensitive extreme learning machine. Secondly, this paper introduces dynamic misclassification and test cost, and gives the chromosome coding and an evaluation method of the optimal cost. At last, on the basis of the genetic algorithm, the dynamic cost-sensitive extreme learning machine classification algorithm for mining incomplete data is given, which can search the optimal misclassification and test cost in cost spaces. The experiment results show that DCELMIDC is effective and feasible for classification of incomplete data, and can reduce the total cost.

Human activity recognition is a main research area of context-aware computing, and is widely used in many applications, such as smart home and elderly care. Smart phone-based human activity recognition is very popular by making use of the embedded inertial sensors. However, there exists the problems of misclassification activities, and how to effectively apply the model trained by known users to new users. To solve these two problems, in this paper, we proposed a novel approach, Uncertainty Sampling based posterior Probability Extreme Learning Machine (USP-ELM), by introducing two strategies: first, we transfer the actual outputs of ELM to posterior probabilities for each instances, and then use uncertainty sampling strategy for confidence level assignment to adapt the training model and improve the classification accuracy. Experimental results show that the proposed approach is more efficient, compared with the existing ELMs.

In order to lower the classification cost and improve the performance of the classifier, this paper proposes the approach of the dynamic cost-sensitive ensemble classification based on extreme learning machine for imbalanced massive data streams (DCECIMDS). Firstly, this paper gives the method of concept drifts detection by extracting the attributive characters of imbalanced massive data streams. If the change of attributive characters exceeds threshold value, the concept drift occurs. Secondly, we give Cost-sensitive extreme learning machine algorithm, and the optimal cost function is defined by the dynamic cost matrix. Build the cost-sensitive classifiers model for imbalanced massive data streams under MapReduce, and the data streams are processed in parallel. At last, the weighted cost-sensitive ensemble classifier is constructed, and the dynamic cost-sensitive ensemble classification based on extreme learning machine classification is given. The experiments demonstrate that the proposed ensemble classifier under the MapReduce framework can reduce the average misclassification cost and can make the classification results more reliable. DCECIMDS has high performance by comparing to the other classification algorithms for imbalanced data streams and can effectively deal with the concept drift.

With the large-scale application of high dimensional gene expression data which exists lots of redundant information, it may waste a lot of time in feature selection and classification. By analyzing the process of MapReduce computing paradigms on cloud platform, it is found that the feature selection which through parallel and distributed computing in MapReduce combined with extreme learning machine is appropriate for constructing a recognition method. This paper proposed a MapReduce algorithm on high gene feature for parallel and distributed selection and classification, aiming to save time resources to make a higher accuracy in training process on large scale gene datasets. Simulation experiments on gene datasets show that the running time on cloud platform is greatly shortened by the time promising the high classification accuracy.

본 논문에서는 인공 신경망의 일종인 Extreme Learning Machine의 학습 알고리즘을 기반으로 하여 노이즈에 강한 특성을 보이는 퍼지 집합 이론을 이용한 새로운 패턴 분류기를 제안 한다. 기존 인공 신경망에 비해 학습속도가 매우 빠르며, 모델의 일반화 성능이 우수하다고 알려진 Extreme Learning Machine의 학습 알고리즘을 퍼지 패턴 분류기에 적용하여 퍼지 패턴 분류기의 학습 속도와 패턴 분류 일반화 성능을 개선 한다. 제안된 퍼지패턴 분류기의 학습 속도와 일반화 성능을 평가하기 위하여, 다양한 머신 러닝 데이터 집합을 사용한다.

In this paper, we introduce a new pattern classifier which is based on the learning algorithm of Extreme Learning Machine the sort of artificial neural networks and fuzzy set theory which is well known as being robust to noise. The learning algorithm used in Extreme Learning Machine is faster than the conventional artificial neural networks. The key advantage of Extreme Learning Machine is the generalization ability for regression problem and classification problem. In order to evaluate the classification ability of the proposed pattern classifier, we make experiments with several machine learning data sets.

Online voltage stability monitoring using real-time measurements is one of the most important tasks in a smart grid system to maintain the grid stability. Loading margin is a good indicator for assessing the voltage stability level. This paper presents an Extreme Learning Machine (ELM) approach for estimation of voltage stability level under credible contingencies using real-time measurements from Phasor Measurement Units (PMUs). PMUs enable a much higher data sampling rate and provide synchronized measurements of real-time phasors of voltages and currents. Depth First (DF) algorithm is used for optimally placing the PMUs. To make the ELM approach applicable for a large scale power system problem, Mutual information (MI)-based feature selection is proposed to achieve the dimensionality reduction. MI-based feature selection reduces the number of network input features which reduces the network training time and improves the generalization capability. Voltage magnitudes and phase angles received from PMUs are fed as inputs to the ELM model. IEEE 30-bus test system is considered for demonstrating the effectiveness of the proposed methodology for estimating the voltage stability level under various loading conditions considering single line contingencies. Simulation results validate the suitability of the technique for fast and accurate online voltage stability assessment using PMU data.

본 논문에서는 Extreme Learning Machine(ELM)을 이용한 자기부상시스템 모델링 기법을 제안한다. 자기부상시스템의 모델링을 위하여 일반적으로 테일러 급수를 이용한 선형화 모델이 사용되어져 왔으나, 이런 수학적 기법의 경우 자기부상시스템의 비선형 반영에 한계가 있다는 단점을 가지고 있다. 이러한 단점을 극복하기 위해 본 논문에서는 학습시간이 빠른 특성을 가진 ELM을 이용한 자기부상시스템의 모델링 기법을 제안한다. 제안된 기법은 입력 가중치들과 은닉 바이어스들의 초기값을 무작위로 선택하고 출력 가중치들은 Moore-Penrose의 일반화된 역행렬 방법을 통하여 구해진다. 실험을 통하여 제안된 알고리즘이 자기부상시스템의 모델링에서 수학적 기법에 비해 우수한 성능을 보임을 알 수 있었다.

In this paper, a new modeling method of a magnetic levitation(Maglev) system using extreme learning machine(ELM) is proposed. The linearized methods using Taylor Series expansion has been used for modeling of a Maglev system. However, the numerical method has some drawbacks when dealing with the components with high nonlinearity of a Maglev system. To overcome this problem, we propose a new modeling method of the Maglev system with electro magnetic suspension, which is based on ELM with fast learning time than conventional neural networks. In the proposed method, the initial input weights and hidden biases of the method are usually randomly chosen, and the output weights are analytically determined by using Moore-Penrose generalized inverse. matrix Experimental results show that the proposed method can achieve better performance for modeling of Maglev system than the previous numerical method.

본 연구에서는 안정적인 물 공급과 에너지의 효율적 사용을 위한 단기 물 수요예측알고리즘 개발에 있어서, 지방 소도시 지역의 물 공급패턴에 대한 영향인자를 도출하기 위하여 기상환경인자와 과거 물 공급량에 대한 상관성 분석을 실시하였다. 그리고, 신경회로망 이론 중 ELM알고리즘을 적용한 단기 물 수요예측알고리즘을 개발하여 현장 적용성을 검토하고자 한다.