Distribution mismatch between the modeling data and the query data is a known domain adaptation issue in machine learning. To this end, in this paper, we propose a l2,1-norm based discriminative robust kernel transfer learning (DKTL) method for high-level recognition tasks. The key idea is to realize robust domain transfer by simultaneously integrating domain-class-consistency (DCC) metric based discriminative subspace learning, kernel learning in reproduced kernel Hilbert space, and representation learning between source and target domain. The DCC metric includes two properties: domain-consistency used to measure the between-domain distribution discrepancy and class-consistency used to measure the within-domain class separability. The essential objective of the proposed transfer learning method is to maximize the DCC metric, which is equivalently to minimize the domain-class-inconsistency (DCIC), such that domain distribution mismatch and class inseparability are well formulated and unified simultaneously. The merits of the proposed method include (1) the robust sparse coding selects a few valuable source data with noises (outliers) removed during knowledge transfer, and (2) the proposed DCC metric can pursue more discriminative subspaces of different domains. As a result, the maximum class-separability is also well guaranteed. Extensive experiments on a number of visual datasets demonstrate the superiority of the proposed method over other state-of-the-art domain adaptation and transfer learning methods.

Extreme learning machine (ELM) is proposed for solving a single-layer feed-forward network (SLFN) with fast learning speed and has been confirmed to be effective and efficient for pattern classification and regression in different fields. ELM originally focuses on the supervised, semi-supervised, and unsupervised learning problems, but just in the single domain. To our best knowledge, ELM with cross-domain learning capability in subspace learning has not been exploited very well. Inspired by a cognitive-based extreme learning machine technique (Cognit Comput. 6:376–390, 1; Cognit Comput. 7:263–278, 2.), this paper proposes a unified subspace transfer framework called cross-domain extreme learning machine (CdELM), which aims at learning a common (shared) subspace across domains. Three merits of the proposed CdELM are included: (1) A cross-domain subspace shared by source and target domains is achieved based on domain adaptation; (2) ELM is well exploited in the cross-domain shared subspace learning framework, and a new perspective is brought for ELM theory in heterogeneous data analysis; (3) the proposed method is a subspace learning framework and can be combined with different classifiers in recognition phase, such as ELM, SVM, nearest neighbor, etc. Experiments on our electronic nose olfaction datasets demonstrate that the proposed CdELM method significantly outperforms other compared methods.

•A unsupervised and robust subspace projection is proposed for anti-drift in E-nose.•A new domain distance concept is proposed as data distribution discrepancy metric.•The proposed subspace projection is a generalized PCA synthesis and easily solved.•The anti-drift is well manifested on two E-nose datasets of sensor drift and shift.Anti-drift is an emergent and challenging issue in sensor-related subjects. In this paper, we propose to address the time-varying drift (e.g. electronic nose drift), which is sometimes an ill-posed problem due to its uncertainty and unpredictability. Considering that drift is with different probability distribution from the regular data, a machine learning based subspace projection approach is proposed. The main idea behind is that given two data clusters with different probability distribution, we tend to find a latent projection P (i.e. a group of basis), such that the newly projected subspace of the two clusters is with similar distribution. In other words, drift is automatically removed or reduced by projecting the data onto a new common subspace. The merits are threefold: 1) the proposed subspace projection is unsupervised; without using any data label information; 2) a simple but effective domain distance is proposed to represent the mean distribution discrepancy metric; 3) the proposed anti-drift method can be easily solved by Eigen decomposition; and anti-drift is manifested with a well solved projection matrix in real application. Experiments on synthetic data and real datasets demonstrate the effectiveness and efficiency of the proposed anti-drift method in comparison to state-of-the-art methods.

•We solve the issue of gas interference in E-nose composed of MOS sensors.•The Pattern Mismatch based Interference Elimination (PMIE) method can solve this problem.•The PMIE method contains two parts: discrimination and correction.•Validity of the discrimination and correction were both proved theoretically and experimentally.•Validity of interference elimination was proved by comparing with ICA and OSC method.Metal oxide semiconductor (MOS) sensor array with cross-sensitivity to target gases is often used in electronic nose (e-nose) for monitoring indoor air quality. However, MOS sensors have their own defects of high susceptibility to some interferences which would seriously impact on the detection of target gases. Therefore it is urgent to solve the problem of interferences elimination as e-nose composed of MOS sensors cannot be used when there are interferences. A closely related method tends to discriminate the interference gases and target gases, and it depends on the type of interference gases. However, there are numerous interferences in real-world application scenario, which is impossible to be sampled in laboratory experiments. Considering that target gases detected by an e-nose can be fixed as invariant information, a novel and effective Pattern Mismatch based Interference Elimination (PMIE) method is proposed in this paper. It contains two parts: interference discrimination (i.e. pattern mismatch) and correction (i.e. interference elimination). Specifically, the principle of interference discrimination is whether a new pattern violates the rules established on the invariant target gases information (i.e. the case of interference gas appearing) or not (i.e. the case of only target gas appearing). If the current pattern of the sensor array is of interference, orthogonal signal correction algorithm (OSC) is used for interference correction. Experimental results prove that the proposed PMIE method is significantly effective for interference elimination in e-nose.

•This paper proposes to extract principal component feature using a kernel entropy component analysis technique.•A KECA plus SVM algorithm is proposed for recognition of multiple indoor air contaminants.•Experiments on discrimination of six kinds of indoor air contaminants demonstrate the effectiveness of KECA for feature extraction.Component analysis techniques for feature extraction in multi-sensor system (electronic nose) have been studied in this paper. A novel nonlinear kernel based Renyi entropy component analysis method is presented to address the feature extraction problem in sensor array and improve the odor recognition performance of E-nose. Specifically, a kernel entropy component analysis (KECA) as a nonlinear dimension reduction technique based on the Renyi entropy criterion is presented in this paper. In terms of the popular support vector machine (SVM) learning technique, a joint KECA–SVM framework is proposed as a system for nonlinear feature extraction and multi-class gases recognition in E-nose community. In particular, the comparisons with PCA, KPCA and ICA based component analysis methods that select the principal components with respect to the largest eigen-values or correlation have been fully explored. Experimental results on formaldehyde, benzene, toluene, carbon monoxide, ammonia and nitrogen dioxide demonstrate that the KECA–SVM method outperforms other methods in classification performance of E-nose. The MATLAB implementation of this work is available online at http://www.escience.cn/people/lei/index.html

•This paper proposes a new discriminant analysis framework for feature extraction and recognition.•The principle of the proposed NDA is derived mathematically.•The NDA framework is coupled with kernel PCA for classification.•The proposed KNDA is compared with state of the art e-Nose recognition methods.•The proposed KNDA shows the best performance in e-Nose experiments.Electronic nose (e-Nose) technology based on metal oxide semiconductor gas sensor array is widely studied for detection of gas components. This paper proposes a new discriminant analysis framework (NDA) for dimension reduction and e-Nose recognition. In a NDA, the between-class and the within-class Laplacian scatter matrix are designed from sample to sample, respectively, to characterize the between-class separability and the within-class compactness by seeking for discriminant matrix to simultaneously maximize the between-class Laplacian scatter and minimize the within-class Laplacian scatter. In terms of the linear separability in high dimensional kernel mapping space and the dimension reduction of principal component analysis (PCA), an effective kernel PCA plus NDA method (KNDA) is proposed for rapid detection of gas mixture components by an e-Nose. The NDA framework is derived in this paper as well as the specific implementations of the proposed KNDA method in training and recognition process. The KNDA is examined on the e-Nose datasets of six kinds of gas components, and compared with state of the art e-Nose classification methods. Experimental results demonstrate that the proposed KNDA method shows the best performance with average recognition rate and total recognition rate as 94.14% and 95.06% which leads to a promising feature extraction and multi-class recognition in e-Nose.

•This paper studies a novel classifier ensemble for multi-classification problems.•Nonlinear feature extraction method is studied using kernel trick.•An improved fusion strategy was used to integrate decisions from base classifiers.•The proposed ensemble classifier was compared with standard majority voting method.This paper presents a novel multiple classifiers system called as improved support vector machine ensemble (ISVMEN) which solves a multi-class recognition problem in electronic nose (E-nose) and aims to improve the accuracy and robustness of classification. The contributions of this paper are presented in two aspects: first, in order to improve the accuracy of base classifiers, kernel principal component analysis (KPCA) method is used for nonlinear feature extraction of E-nose data; second, in the process of establishing classifiers ensemble, a new fusion approach which conducts an effective base classifier weighted method is proposed. Experimental results show that the average classification accuracy has been improved from less than 86% to 92.58% compared with that of base classifiers. Besides, the proposed fusion method is also superior to MV fusion method (majority voting) which has 90.1% of classification accuracy. Especially, the proposed ISVMEN can obtain the best discrimination accuracy for C7H8, CO and NH3, almost 100% classification accuracy was obtained using our method. Therefore, it is easy to come to the conclusion that, in average, the proposed method is better significantly than other methods in classification and generalization performance.

•This paper proposes a rapid scheme discreteness correction for MOS gas sensors.•The scheme is to improve the reproducibility in large scale E-nose production.•The discreteness can be well corrected using the proposed scheme.•The paper shows that the discreteness has little relation with the type of gases.Metal oxide semiconductor (MOS) gas sensors have been widely reported in machine olfaction system (i.e. electronic nose/tongue) for rapid detection of gas mixture components due to their positive characteristics of cross-sensitivity, broad spectrum response and low-cost. However, the discreteness of MOS gas sensors caused by inherent sensor variability during the manufacturing process results in the failure of the batch-oriented applications of MOS gas sensors due to their weak reproducibility. Certainly, it will also cause negative influence to the development of electronic nose/tongue based on MOS gas sensors (e.g. accuracy and consistency during electronic nose/tongue detections). Therefore, the contribution of this paper is to solve the discreteness and improve the reproducibility of sensors by designing an effective and easily realized scheme for large-scale calibration. Experimental results demonstrate that the proposed scheme can effectively and rapidly realize the calibration of the sensors’ discreteness in batch of electronic noses production and the proposed scheme have also been used in industry. Besides, this paper also proves that one sensor's discreteness is constant and keeps unchanged when the sensor is exposed to different kinds of gas components.

•This paper proposes an on-line background elimination model of electronic nose.•Four key features are used for recognition which avoids the discontinuity.•A dynamical signal matrix is updated and promises the adaptive signal correction.•Study of two cases in real-world is developed in electronic nose.Metal oxide semiconductor (MOS) sensor array with some cross-sensitivities to target gases is often used in electronic nose (E-nose) combined with signal processing techniques for indoor air contaminants monitoring. However, MOS sensors have some intrinsic flaw of high susceptibility to background interference which would seriously destroy the specificity and stability of electronic nose in practical application. This paper presents an on-line counteraction of unwanted odor interference based on pattern recognition for the first time. Six kinds of target gases and four kinds of unwanted odor interferences were experimentally studied. First, two artificial intelligence learners including a multi-class least square support vector machine (learner-1) and a binary classification artificial neural network (learner-2) are developed for discrimination of unwanted odor interferences. Second, a real-time dynamically updated signal matrix is constructed for correction. Finally, an effective signal correction method was employed for E-nose data. Experimental results in the real cases studies demonstrate the effectiveness of the presented model in E-nose based on MOS gas sensors array.

Chemical sensor drift shows a chaotic behavior and unpredictability in long-term observation which makes it difficult to construct an appropriate sensor drift treatment. The main purpose of this paper is to study a new methodology for chaotic time series modeling of chemical sensor observations in embedded phase space. This method realizes a long-term prediction of sensor baseline and drift based on phase space reconstruction (PSR) and radial basis function (RBF) neural network. PSR can memory all of the properties of a chaotic attractor and clearly show the motion trace of a time series, thus PSR makes the long-term drift prediction using RBF neural network possible. Experimental observation data of three metal oxide semiconductor sensors in a year demonstrate the obvious chaotic behavior through the Lyapunov exponents. Results demonstrate that the proposed model can make long-term and accurate prediction of chemical sensor baseline and drift time series.

Electronic nose (E-nose), as an artificial olfactory system, can be used for estimation of gases concentration combined with a pattern recognition module. This paper studies the concentration estimations of indoor contaminants for air quality monitoring in dwellings using chaos based optimization artificial neural network integrated into our self-designed portable E-nose instrument. Back-propagation neural network (BPNN) has been recognized as the common pattern recognition. Considering the local optimal flaw of BPNN, this paper presents a novel chaotic sequence optimization BPNN method for improving the accuracy of E-nose prediction. Further comparison with particle swarm optimization is also employed, and maximum 26.03% and 16.4% prediction error decreased after using chaotic based optimization for formaldehyde and benzene concentration estimation. Experimental results demonstrate the superiority and efficiency of the portable E-nose instrument integrated with artificial neural network optimized by chaotic sequence based optimization algorithms in real-time monitoring of air quality in dwellings.

Computer-aided protein-coding gene prediction in uncharacterized genomic DNA sequences is one of the most important issues of biological signal processing. A modified filter method based on a statistically optimal null filter (SONF) theory is proposed for recognizing protein-coding regions. The square deviation gain (SDG) between the input and output of the model is used to identify the coding regions. The effective SDG amplification model with Class I and Class II enhancement is designed to suppress the non-coding regions. Also, an evaluation algorithm has been used to compare the modified model with most gene prediction methods currently available in terms of sensitivity, specificity and precision. The performance for identification of protein-coding regions has been evaluated at the nucleotide level using benchmark datasets and 91.4%, 96%, 93.7% were obtained for sensitivity, specificity and precision, respectively. These results suggest that the proposed model is potentially useful in gene finding field, which can help recognize protein-coding regions with higher precision and speed than present algorithms.

•This paper proposes a similarity measure based on metric fusion and joint similarity score function.•A generalized metric swarm learning that learns local sub-metrics simultaneously is formulated.•An efficient alternative optimization algorithm is given for solving the proposed GMSL in dual.Learning distance metrics for measuring the similarity between two data points in unsupervised and supervised pattern recognition has been widely studied in unconstrained face verification tasks. Motivated by the fact that enforcing single distance metric learning for verification via an empirical score threshold is not robust in uncontrolled experimental conditions, we therefore propose to obtain a metric swarm by learning local patches alike sub-metrics simultaneously that naturally formulates a generalized metric swarm learning (GMSL) model with a joint similarity score function solved by an efficient alternative optimization algorithm. Further, each sample pair is represented as a similarity vector via the well-learned metric swarm, such that the face verification task becomes a generalized SVM-alike classification problem. Therefore, the verification can be enforced in the represented metric swarm space that can well improve the robustness of verification under irregular data structure. Experiments are preliminarily conducted using several UCI benchmark datasets for solving general classification problem. Further, the face verification experiments on real-world LFW and PubFig datasets demonstrate that our proposed model outperforms several state-of-the-art metric learning methods.

Deep learning with a convolutional neural network (CNN) has been proved to be very effective in feature extraction and representation of images. For image classification problems, this work aims at exploring the capability of extreme learning machine on high-level deep features of images. Additionally, motivated by the biological learning mechanism of ELM, in this paper, an adaptive extreme learning machine (AELM) method is proposed for handling cross-task (domain) learning problems, without loss of its nature of randomization and high efficiency. The proposed AELM is an extension of ELM from single task to cross task learning, by introducing a new error term and Laplacian graph based manifold regularization term in objective function. We have discussed the nearest neighbor, support vector machines and extreme learning machines for image classification under deep convolutional activation feature representation. Specifically, we adopt 4 benchmark object recognition datasets from multiple sources with domain bias for evaluating different classifiers. The deep features of the object dataset are obtained by a well-trained CNN with five convolutional layers and three fully-connected layers on ImageNet. Experiments demonstrate that the proposed AELM is comparable and effective in single and multiple domains based recognition tasks.

•A facial beauty analysis toward attractiveness assessment application is presented.•A geometric facial beauty score function is proposed for facial aesthetic perceptron.•A semi-supervised learning with Hessian graph and random projection is proposed.•A novel geometric facial beauty (GFB) database is provided in this paper.Facial beauty analysis has been an emerging subject of multimedia and biometrics. This paper aims at exploring the essence of facial beauty from the viewpoint of geometric characteristic toward an interactive attractiveness assessment (IAA) application. As a result, a geometric facial beauty analysis method is proposed from the perspective of machine learning. Due to the troublesome and subjective beauty labeling, the accurately labeled data scarcity is caused, and result in very few labeled data. Additionally, facial beauty is related to several typical features such as texture, color, etc., which, however, can be easily deformed by make-up. For addressing these issues, a semi-supervised facial beauty analysis framework that is characterized by feeding geometric feature into the intelligent attractiveness assessment system is proposed. For experimental study, we have established a geometric facial beauty (GFB) dataset including Asian male and female faces. Moreover, an existing multi-modal beauty (M2B) database including western and eastern female faces is also tested. Experiments demonstrate the effectiveness of the proposed method. Some new perspectives on the essence of beauty and the topic of facial aesthetic are revealed. The impact of this work lies in that it will attract more researchers in related areas for beauty exploration by using intelligent algorithms. Also, the significance lies in that it should well promote the diversity of expert and intelligent systems in addressing such challenging facial aesthetic perception and rating issue.