Search

In this paper, we propose a weighted multiple classifier framework based on random projections. Similar to the mechanism of other homogeneous ensemble methods, the base classifiers in our approach are obtained by a learning algorithm on different training sets generated by projecting the original up-space training set to lower dimensional down-spaces. We then apply a Least SquarE−based method to weigh the outputs of the base classifiers so that the contribution of each classifier to the final combined prediction is different. We choose Decision Tree as the learning algorithm in the proposed framework and conduct experiments on a number of real and synthetic datasets. The experimental results indicate that our framework is better than many of the benchmark algorithms, including three homogeneous ensemble methods (Bagging, RotBoost, and Random Subspace), several well-known algorithms (Decision Tree, Random Neural Network, Linear Discriminative Analysis, K Nearest Neighbor, L2-loss Linear Support Vector Machine, and Discriminative Restricted Boltzmann Machine), and random projection-based ensembles with fixed combining rules with regard to both classification error rates and F1 scores.

Many batch learning algorithms have been introduced for offline multi-label classification (MLC) over the years. However, the increasing data volume in many applications such as social networks, sensor networks, and traffic monitoring has posed many challenges to batch MLC learning. For example, it is often expensive to re-train the model with the newly arrived samples, or it is impractical to learn on the large volume of data at once. The research on incremental learning is therefore applicable to a large volume of data and especially for data stream. In this study, we develop a Bayesian-based method for learning from multi-label data streams by taking into consideration the correlation between pairs of labels and the relationship between label and feature. In our model, not only the label correlation is learned with each arrived sample with ground truth labels but also the number of predicted labels are adjusted based on Hoeffding inequality and the label cardinality. We also extend the model to handle missing values, a problem common in many real-world data. To handle concept drift, we propose a decay mechanism focusing on the age of the arrived samples to incrementally adapt to the change of data. The experimental results show that our method is highly competitive compared to several well-known benchmark algorithms under both the stationary and concept drift settings. Please note that the published title differs from this accepted manuscript "Multi-label classification via labels correlation and one-dependence features on data stream."

In smart cities, digital image splicing measurement is very important to ensure the security and safety of city monitoring, environment data fusion, cognitive decisions, etc. However, due to images obtained from various environments of cities usually face malevolence splicing, it is hard to perform the authenticity of a legitimate image from smart cities. In this paper, a novel block Discrete Cosine Transform (BDCT) coefficients feature distribution based statistical approach is proposed to discover image forgeries for image splicing measurement. In the proposed feature, all the BDCT neighbouring modes are categorized into a number of groups following the maximum likelihood (ML) criterion to ensure the modes in the same group having similar distributions. For each group, the transition probability matrix (TPM) or the joint probability matrix (JPM) is extracted from the BDCT coefficient pairs in the image. Moreover, the proposed scheme is constructed by concatenating all the TPM/JPM features for each group. Experimental results demonstrate that our feature outperforms two state-of-the-art approaches when taking both the measurement accuracy and feature dimension into consideration.

Image classification is considered to be one of the critical tasks in hyperspectral remote sensing image processing. Recently, a convolutional neural network (CNN) has established itself as a powerful model in classification by demonstrating excellent performances. The use of a graphical model such as a conditional random field (CRF) contributes further in capturing contextual information and thus improving the classification performance. In this paper, we propose a method to classify hyperspectral images by considering both spectral and spatial information via a combined framework consisting of CNN and CRF. We use multiple spectral band groups to learn deep features using CNN, and then formulate deep CRF with CNN-based unary and pairwise potential functions to effectively extract the semantic correlations between patches consisting of 3-D data cubes. Furthermore, we introduce a deep deconvolution network that improves the final classification performance. We also introduced a new data set and experimented our proposed method on it along with several widely adopted benchmark data sets to evaluate the effectiveness of our method. By comparing our results with those from several state-of-the-art models, we show the promising potential of our method.

Death, injury, and disability from road traffic crashes continue to be a major global public health problem. Therefore, methods to reduce accident severity are of significant interest to traffic agencies and the public at large. Noisy data in the traffic accident dataset obscure the discovery of important factors and mislead conclusions. Identifying and correcting noisy values is an important goal of data cleansing and preprocessing . This paper proposes a new algorithm called NoiseCleaner to identify and correct noisy categorical attributes values in large traffic accident datasets. We evaluate our algorithm using four publicly available traffic accident datasets from Australia and United States, namely, two road crash datasets from the Queensland Government data depository (data.qld.gov.au) and two datasets from the New York's open data portal (data.ny.gov). We compare our technique with several existing state-of-the-art methods and show that our algorithm performs significantly better than the existing algorithms.

In this study, a novel framework to combine multiple classifiers in an ensemble system is introduced. Here we exploit the concept of information granule to construct granular prototypes for each class on the outputs of an ensemble of base classifiers. In the proposed method, uncertainty in the outputs of the base classifiers on training observations is captured by an interval-based representation. To predict the class label for a new observation, we first determine the distances between the output of the base classifiers for this observation and the class prototypes, then the predicted class label is obtained by choosing the label associated with the shortest distance. In the experimental study, we combine several learning algorithms to build the ensemble system and conduct experiments on the UCI, colon cancer, and selected CLEF2009 datasets. The experimental results demonstrate that the proposed framework outperforms several benchmarked algorithms including two trainable combining methods, i.e., Decision Template and Two Stages Ensemble System, AdaBoost, Random Forest, L2-loss Linear Support Vector Machine, and Decision Tree.

Good authentication performance and liveness detection are two key requirements in many authentication systems. To avoid replay attacks, a novel visual speaker authentication scheme with random prompt texts is proposed. Compared with the fixed password scenario, visual speaker authentication with random prompt texts is much more challenging because it is impossible to ask the client to pronounce every possible prompt text to be used as training samples. In order to solve this problem, a new deep convolutional neural network is proposed in this paper and it has three functional parts, namely, the lip feature network, the identity network, and the content network. In the lip feature network, a series of 3D residual units have been adopted, which can depict the static and dynamic characteristics of the lip biometrics comprehensively. By considering the distinguishing features of the identity and content authentication tasks, the identity network and the content network are designed accordingly. An end-to-end, multi-task learning scheme is proposed which can optimize the weights of all the above three networks simultaneously. Experiments have been carried out to evaluate the performance of the proposed network under both the fixed-password and the random prompt texts scenario. From the experimental results, it is shown that the proposed approach can achieve superior performance in the fixed-password scenario compared with several state-of-the-art approaches. Furthermore, it also achieves satisfactory authentication results in the random prompt texts scenario and thus it provides a reliable solution for user authentication where liveness is guaranteed.

Multi-objective problems with conflicting objectives cannot be effectively solved by aggregation-based methods. The answer to such problems is a Pareto optimal solution set. Due to the difficulty of solving multi-objective problems using multi-objective algorithms and the lack of enough expertise, researchers in different fields tend to aggregative objectives and use single-objective algorithms. This work is a seminal attempt to propose the use of multi-objective algorithms in the field of hand posture estimation. Hand posture estimation is a key step in hand gesture recognition, which is a part of an overall attempt to make human-computer interaction more like human face-to-face communication. Hand posture estimation is first formulated as a bi-objective problem. A modified version of Multi-Objective Particle Swarm Optimisation (MOPSO) is then proposed to approximate the Pareto optimal font of 50 different postures. The main motivation of integrating a new operator (called Evolutionary Population Dynamics — EPD) in MOPSO is due to the nature of hand posture estimation problems in which parameters should not be tuned in a same manner since they show varied impacts on the objectives. EPD allows randomising different parameters in a solution and provides different exploratory behaviours for the parameters of an optimisation algorithm rather than each individual solution. The MOPSO algorithm is equipped with a mechanism to randomly re-initialise poor particles around the optimal solutions in the archive. The improved MOPSO is tested on ZDT and CEC2009 test functions and compared with the standard MOPSO, NSGA-II, and MOEA/D. The results show that the proposed MOPSO (MOPSO+EPD) significantly outperforms MOPSO on the majority of test functions in terms of both convergence and coverage. MOPSO+EPD also approximates well-distributed Pareto optimal fronts for most of the postures considered in this work. The post analysis of the results is conducted to understand the relationship between the parameters and objectives of this problem (design principals) for the first time in the literature as well.

Genetic algorithms are a practical approach for finding near-optimal solutions for nondeterministic polynomial-hard problems. In this work we exploit the parallel processing capability of graphics processing units and Nvidia's CUDA programming platform to accelerate the island model genetic algorithm by modifying the evolutionary operations to fit the hardware architecture and have successfully achieved significant computational speedups.

Recent research shows that the lip feature can achieve reliable authentication performance with a good liveness detection ability. However, with the development of sophisticated human face generation methods by the deepfake technology, the talking videos can be forged with high quality and the static lip information is not reliable in such case. Meeting with such challenge, in this paper, we propose a new deep neural network structure to extract robust lip features against human and Computer-Generated (CG) imposters. Two novel network units, i.e. the feature-level Difference block (Diffblock) and the pixel-level Dynamic Response block (DRblock), are proposed to reduce the influence of the static lip information and to represent the dynamic talking habit information. Experiments on the GRID dataset have demonstrated that the proposed network can extract discriminative and robust lip features and outperform two state-of-the-art visual speaker authentication approaches in both human imposter and CG imposter scenarios.

Deep Neural Networks have achieved many successes when applying to visual, text, and speech information in various domains. The crucial reasons behind these successes are the multi-layer architecture and the in-model feature transformation of deep learning models. These design principles have inspired other sub-fields of machine learning including ensemble learning. In recent years, there are some deep homogenous ensemble models introduced with a large number of classifiers in each layer. These models, thus, require a costly computational classification. Moreover, the existing deep ensemble models use all classifiers including unnecessary ones which can reduce the predictive accuracy of the ensemble. In this study, we propose a multi-layer ensemble learning framework called MUlti-Layer heterogeneous Ensemble System (MULES) to solve the classification problem. The proposed system works with a small number of heterogeneous classifiers to obtain ensemble diversity, therefore being efficiency in resource usage. We also propose an Evolutionary Algorithm-based selection method to select the subset of suitable classifiers and features at each layer to enhance the predictive performance of MULES. The selection method uses NSGA-II algorithm to optimize two objectives concerning classification accuracy and ensemble diversity. Experiments on 33 datasets confirm that MULES is better than a number of well-known benchmark algorithms.

Biomedical data are available in many different formats, including numeric, textual reports, signals or images, and they come available from a variety of sources. Biomedical data typically suffer from incompleteness, uncertainty and vagueness, posing several challenges to perform data analysis, such high dimensionality, class imbalance or low numbers of samples [ 1 , 2 ]. Granular Computing, the term coined by Prof. L. A. Zadeh, provides a powerful tool for multiple granularity and multiple-view data analysis, which is of vital importance for understanding data driven analysis at different levels of ab- straction (granularity) [3] . It is worth stressing that human’s capabilities in effective information or ganization and efficient reasoning with complex and uncertain information is highly dependent on hierarchical Granular Computing [ 4 , 5 ]. We have been witnessing significant advances of Granular Computing in the scientific and engineering domains. Data mining based on Granular Computing in biomedical data analysis is an emerging field which crosses multiple research disciplines and in- dustry domains. As a meta-mathematical methodology, granular data mining provides a theoretical framework for biomed- ical data analytics. It helps to extract knowledge when we are provided with an insufficient data that may also contain a significant amount of unstructured, uncertain and imprecise data. Granular data mining technology has exhibited some strong capabilities and advantages in intelligent data analysis and uncertainty reasoning for biomedical data. However, de- termining how to integrate Granular Computing and data mining to combine their advantages remains an interesting and important research topic. Recent survey indicated that granular data mining research has been focused on exploring the advantages, and also the challenges, derived from collecting and mining vast amounts of available biomedical data sources. It has therefore become strongly and timely justified to develop theoretical models and practical algorithms for carrying out granular data mining for biomedical data analysis.

In this study, we introduce an ensemble system by combining homogeneous ensemble and heterogeneous ensemble into a single framework. Based on the observation that the projected data is significantly different from the original data as well as each other after using random projections, we construct the homogeneous module by applying random projections on the training data to obtain the new training sets. In the heterogeneous module, several learning algorithms will train on the new training sets to generate the base classifiers. We propose four combining algorithms based on Sum Rule and Majority Vote Rule for the proposed ensemble. Experiments on some popular datasets confirm that the proposed ensemble method is better than several well-known benchmark algorithms proposed framework has great flexibility when applied to real-world applications. The proposed framework has great flexibility when applied to real-world applications by using any techniques that make rich training data for the homogeneous module, as well as using any set of learning algorithms for the heterogeneous module.

Mixed spectral responses from different ground materials often create confusions in complex remote sensing scenes and restrict classification performance. In this regard, unmixing approaches are being successfully carried out to decompose mixed pixels into a collection of spectral signatures. In this paper, we propose a method to integrate unmixing into a deep feature learning model in order to classify hyperspectral data. We propose to generate superpixels from the abundance estimations of the underlying materials of the image obtained from an unsupervised endmember extraction algorithm called vertex component analysis (VCA). The mean abundances of the superpixels are then used as features for a deep classifier. Our proposed deep model, formulated as a joint convolutional neural network and recurrent neural network, receives significant spectral-spatial information in the data to produce better and powerful features and achieve improved classification performance than several alternative methods.

Although much effort has been spent in developing a stable algorithm for 3D building modelling from Lidar data, this topic still attracts a lot of attention in the literature. A key task of this problem is the automatic building roof segmentation. Due to the great diversity of building typology, and the noisiness and heterogeneity of point cloud data, the building roof segmentation result needs to be verified/rectified with some geometric constrains before it is used to generate the 3D building models. Otherwise, the generated building model may suffer from undesirable deformations. This paper suggests the generation of 3D building model from Lidar data in two steps. The first step is the automatic 2D building modelling and the second step is the automatic conversion of a 2D building model into 3D model. This approach allows the 2D building model to be refined before starting the 3D building model generation. Furthermore, this approach allows getting the 2D and 3D building models simultaneously. The first step of the proposed algorithm is the generation of the 2D building model. Then after enhancing and fitting the roof planes, the roof plane boundaries are converted into 3D by analysing the relationships between neighbouring planes. This is followed by the adjustment of the 3D roof vertices. Experiment indicated that the proposed algorithm is accurate and robust in generating 3D building models from Lidar data.

BACKGROUND: Common among patients with COPD is declining health-related quality of life (HRQOL). Although results of research identified some factors associated with HRQOL, resilience factors are yet to be fully investigated. METHODS: This study examined resilience and demographic factors associated with HRQOL. Participants >40 y old were recruited from community health programs and hospitals in South East Queensland. Self-administered questionnaires were used to query subjects9 HRQOL and levels of resilience. A decision tree examined the factors important to HRQOL in 159 subjects with COPD. RESULTS: Factors of importance in the HRQOL of subjects with COPD were found in 3 domains of the St George Respiratory Questionnaire. Of importance on the breathlessness domain was marital status, defensive coping, coping, number of comorbidities, relationships, decision-making, self-esteem, self-efficacy, and professional support of health and well-being. Of the symptoms domain, self-efficacy, recruitment location, anxiety/depression, decision-making, self-esteem, coping, relationships, professional support of health and well-being, and risks were important. The cough domain found recruitment location, anxiety/depression, professional support of health and well-being, coping, and defensive coping to be important for subjects9 HRQOL. CONCLUSIONS: Resilience and confounding factors were of importance in the HRQOL of subjects with COPD. Thus, consultation with a medical professional, especially at discharge, who identifies, encourages, and approves of the patient9s disease management abilities will enhance both resilience and HRQOL.

Discovering inexpensive and earth-abundant electrocatalysts to replace the scarce platinum group metal-based electrocatalysts holds the key for large-scale hydrogen fuel generation, which relies heavily on the theoretical understanding of the properties of candidate materials and their operating environment. The recent applications of the cobalt–dithiolene complex as promising electrocatalysts for the hydrogen evolution reaction have been broadened by forming low-dimensional metal–organic frameworks (MOFs) through polymerization. Using the Gibbs free energy of the adsorption of hydrogen atoms as a key descriptor, S atoms within one-dimensional MOFs are identified to be the preferred catalytic site for HERs. Our theoretical results further reveal that the activities of part S atoms can be improved by interacting with alkali metal cations from the electrolytes; specifically, the influence of cations on the performance is dependent on the electron affinity of cations. Our theoretical findings, therefore, dem...

With the development of sophisticated image editing and manipulation tools, the originality and authenticity of a digital image is usually hard to determine visually. In order to detect digital image forgeries, various kinds of digital image forensics techniques have been proposed in the last decade. Compared with active forensics approaches that require embedding additional information, passive forensics approaches are more popular due to their wider application scenario, and have attracted increasing academic and industrial research interests. Generally speaking, passive digital image forensics detects image forgeries based on the fact that there are certain intrinsic patterns in the original image left during image acquisition or storage, or specific patterns in image forgeries left during the image storage or editing. By analyzing the above patterns, the originality of an image can be authenticated. In this paper, a brief review on passive digital image forensic methods is presented in order to provide a comprehensive introduction on recent advances in this rapidly developing research area. These forensics approaches are divided into three categories based on the various kinds of traces they can be used to track—that is, traces left in image acquisition, traces left in image storage, and traces left in image editing. For each category, the forensics scenario, the underlying rationale, and state-of-the-art methodologies are elaborated. Moreover, the major limitations of the current image forensics approaches are discussed in order to point out some possible research directions or focuses in these areas. Keywords: Digital image forensics, Image-tampering detection, Multimedia security

Autonomously spiking dopaminergic neurons of the substantia nigra pars compacta (SNpc) are exquisitely specialized and suffer toxic iron-loading in Parkinson's disease (PD). However, the molecular mechanism involved remains unclear and critical to decipher for designing new PD therapeutics. The long-lasting (L-type) CaV1.3 voltage-gated calcium channel is expressed at high levels amongst nigral neurons of the SNpc, and due to its role in calcium and iron influx, could play a role in the pathogenesis of PD. Neuronal iron uptake via this route could be unregulated under the pathological setting of PD and potentiate cellular stress due to its redox activity. This Commentary will focus on the role of the CaV1.3 channels in calcium and iron uptake in the context of pharmacological targeting. Prospectively, the audacious use of artificial intelligence to design innovative CaV1.3 channel inhibitors could lead to breakthrough pharmaceuticals that attenuate calcium and iron entry to ameliorate PD pathology.

Background and Objective Cerebral microbleeds (CMB) are important biomarkers of cerebrovascular diseases and cognitive dysfunctions. Susceptibility weighted imaging (SWI) is a common MRI sequence where CMB appear as small hypointense blobs. The prevalence of CMB in the population and in each scan is low, resulting in tedious and time-consuming visual assessment. Automated detection methods would be of value but are challenged by the CMB low prevalence, the presence of mimics such as blood vessels, and the difficulty to obtain sufficient ground truth for training and testing. In this paper, synthetic CMB (sCMB) generation using an analytical model is proposed for training and testing machine learning methods. The main aim is creating perfect synthetic ground truth as similar as reals, in high number, with a high diversity of shape, volume, intensity, and location to improve training of supervised methods. Method sCMB were modelled with a random Gaussian shape and added to healthy brain locations. We compared training on our synthetic data to standard augmentation techniques. We performed a validation experiment using sCMB and report result for whole brain detection using a 10-fold cross validation design with an ensemble of 10 neural networks. Results Performance was close to state of the art (~9 false positives per scan), when random forest was trained on synthetic only and tested on real lesion. Other experiments showed that top detection performance could be achieved when training on synthetic CMB only. Our dataset is made available, including a version with 37,000 synthetic lesions, that could be used for benchmarking and training. Conclusion Our proposed synthetic microbleeds model is a powerful data augmentation approach for CMB classification with and should be considered for training automated lesion detection system from MRI SWI.

Motivation Protein–peptide interactions are one of the most important biological interactions and play crucial role in many diseases including cancer. Therefore, knowledge of these interactions provides invaluable insights into all cellular processes, functional mechanisms, and drug discovery. Protein–peptide interactions can be analyzed by studying the structures of protein–peptide complexes. However, only a small portion has known complex structures and experimental determination of protein–peptide interaction is costly and inefficient. Thus, predicting peptide-binding sites computationally will be useful to improve efficiency and cost effectiveness of experimental studies. Here, we established a machine learning method called SPRINT-Str (Structure-based prediction of protein–Peptide Residue-level Interaction) to use structural information for predicting protein–peptide binding residues. These predicted binding residues are then employed to infer the peptide-binding site by a clustering algorithm. Results SPRINT-Str achieves robust and consistent results for prediction of protein–peptide binding regions in terms of residues and sites. Matthews’ Correlation Coefficient (MCC) for 10-fold cross validation and independent test set are 0.27 and 0.293, respectively, as well as 0.775 and 0.782, respectively for area under the curve. The prediction outperforms other state-of-the-art methods, including our previously developed sequence-based method. A further spatial neighbor clustering of predicted binding residues leads to prediction of binding sites at 20–116% higher coverage than the next best method at all precision levels in the test set. The application of SPRINT-Str to protein binding with DNA, RNA and carbohydrate confirms the method‘s capability of separating peptide-binding sites from other functional sites. More importantly, similar performance in prediction of binding residues and sites is obtained when experimentally determined structures are replaced by unbound structures or quality model structures built from homologs, indicating its wide applicability. Availability and implementation http://sparks-lab.org/server/SPRINT-Str Supplementary information Supplementary data are available at Bioinformatics online.

Propose a sparse code based feature with spatiotemporal structure for lip biometrics.Achieve excellent speaker identification and authentication performance.Our feature is robust to variations caused by speaker's positions and poses. Recent research shows that lip shape and lip movement contain abundant identity-related information and can be used as a new kind of biometrics in speaker identification or authentication. In this paper, we propose a new lip feature representation for lip biometrics which is able to describe the static and dynamic characteristics of a lip sequence. The new representation captures both the physiological and behavioral aspects of the lip and is robust against variations caused by different speaker position and pose. In our approach, a lip sequence is first divided into several subsequences along the temporal dimension. For each subsequence, sparse coding (SC in short) is adopted to characterize the minutiae of the lip region and its movement in small spatiotemporal cells. Then max-pooling based on a hierarchical spatiotemporal structure is performed on the SC codes to generate the final feature for each of the subsequence. Finally, the entire lip sequence is represented by a set of features corresponding to each subsequence in it. Experiments are carried out on a dataset with 40 speakers and compared with three state-of-the-art approaches. From the experimental results, it was observed that the proposed feature achieved high identification accuracy (an accuracy of 99.96%) and very low authentication error (a Half Total Error Rate (HTER) of 0.46%), and outperformed the other approaches investigated. Moreover, even with random variations caused by different speaker position and pose, the proposed feature still provides good identification (an accuracy of 99.18%) and authentication results (a HTER of 2.34%) and has much lower performance degradation compared with the other approaches investigated. Finally, even when there is only one training sample per speaker, the proposed feature still achieves high discriminative power (an accuracy of 98.39% and HTER of 2.62%).

In a normal human brain, inter-hemispheric fissure separates the brain into the left and the right hemispheres. In this paper, we model IF as a mid-sagittal surface on the input 3D brain MR image. For this purpose, we introduce a new method to extract MSS. In the proposed method, lacunarity is used to extract an initial symmetry plane, and then, fractal dimension is calculated in order to measure similarity degree between two brain hemispheres. Inside of each axial slice, a thin-plate spline surface is constructed based on the FD and intensity values, and a local optimization is applied to fit this TPS surface to the brain data using a robust least-median-of-squares estimator. Finally, MSS is modelled as a stack of the fitted TPSs, and the optimization is applied again in order to smooth the final MSS. MSS is the output of our method. The efficiency of the proposed method is evaluated using both simulated and real MR images and is compared to the state of the art. Our studies show that the proposed method discovers significant mid-sagittal surface with respect to the increased noise level and INU existence, in clinical images and pathological samples. This superiority is reasonable because of using FD and lacunarity being noise and INU independent and optimizing by TPS working locally.

Lip image segmentation, as the first step in many lip-related tasks (e.g. automatic lipreading), is of vital significance for the subsequent procedures. Nowadays, with the increasing computational power of the mobile devices, mobile applications become more and more popular. In this paper, a new approach is proposed, which is able to segment the lip region in natural scenes and is of acceptable computational complexity to be implemented in mobile devices. Two networks including a complex teacher network and a compact student network with the same structure are employed. With the proposed remedy loss and the alternative knowledge distillation scheme, the student network can learn useful knowledge from the teacher network effectively and efficiently, and even rectify some of its segmentation errors. A dataset containing 49 people captured under natural scenes by various cellphone cameras is adopted for evaluation and the experiment results have demonstrated that the proposed student network even outperforms the teacher network with much less computational cost.

In ensemble systems, the predictions of base classifiers are aggregated by a combining algorithm (meta-classifier) to achieve better classification accuracy than using a single classifier. Experiments show that the performance of ensembles significantly depends on the choice of meta-classifier. Normally, the classifier selection method applied to an ensemble usually removes all the predictions of a classifier if this classifier is not selected in the final ensemble. Here we present an idea to only remove a subset of each classifier's prediction thereby introducing a simultaneous meta-data and meta-classifier selection method for ensemble systems. Our approach uses Cross Validation on the training set to generate meta-data as the predictions of base classifiers. We then use Ant Colony Optimization to search for the optimal subset of meta-data and meta-classifier for the data. By considering each column of meta-data, we construct the configuration including a subset of these columns and a meta-classifier. Specifically, the columns are selected according to their corresponding pheromones, and the meta-classifier is chosen at random. The classification accuracy of each configuration is computed based on Cross Validation on meta-data. Experiments on UCI datasets show the advantage of proposed method compared to several classifier and feature selection methods for ensemble systems.

Recently, multi-label classification algorithms have been increasingly required by a diversity of applications, such as text categorization, web, and social media mining. In particular, these applications often have streams of data coming continuously, and require learning and predicting done on-the-fly. In this paper, we introduce a scalable online variational inference based ensemble method for classifying multi-label data, where random projections are used to create the ensemble system. As a second-order generative method, the proposed classifier can effectively exploit the underlying structure of the data during learning. Experiments on several real-world datasets demonstrate the superior performance of our new method over several well-known methods in the literature.

Knowledge discovery is a process of finding hidden knowledge from a large volume of data that involves data mining. Data mining unveils interesting relationships among data and the results can help in making valuable predictions or recommendation in various applications. Bi-clustering is an unsupervised machine learning technique that can uncover useful information from Big data. Bi-clustering has many useful applications in various fields such as pattern classification, information retrieval, gene expression data analysis and functional annotation. The goal of bi-clustering is to detect coherent groups of data by performing clustering along the rows and columns dimension of a dataset simultaneously. Using both the rows and columns information in the data, bi-clustering usually requires the optimization of two or more conflicting objectives. In this chapter, we review some recent state-of-the-art multi-objective, evolutionary-based bi-clustering algorithms and discuss their application in data mining for multimodal and Big data.

In this paper, we aim to develop an effective combining algorithm for ensemble learning systems. The Decision Template method, one of the most popular combining algorithms for ensemble systems, does not perform well when working on certain datasets like those having imbalanced data. Moreover, point estimation by computing the average value on the outputs of base classifiers in the Decision Template method is sometimes not a good representation, especially for skewed datasets. Here we propose to search for an optimal decision template in the combining algorithm for a heterogeneous ensemble. To do this, we first generate the base classifier by training the pre-selected learning algorithms on the given training set. The meta-data of the training set is then generated via cross validation. Using the Artificial Bee Colony algorithm, we search for the optimal template that minimizes the empirical 0–1 loss function on the training set. The class label is assigned to the unlabeled sample based on the maximum of the similarity between the optimal decision template and the sample’s meta-data. Experiments conducted on the UCI datasets demonstrated the superiority of the proposed method over several benchmark algorithms.

This chapter presents a review of important issues for multimodal information processing and Big data analytics in a digital world, emphasizing the issues brought by the concept of universal machine learning intelligence and data-from-everywhere, and driven by the applications for the future and next-generation technologies. Furthermore, the chapter explains the organization of the book, describing which issues and related technologies are addressed in which chapters of the book.

This edited book will serve as a source of reference for technologies and applications for multimodality data analytics in big data environments. After an introduction, the editors organize the book into four main parts on sentiment, affect and emotion analytics for big multimodal data; unsupervised learning strategies for big multimodal data; supervised learning strategies for big multimodal data; and multimodal big data processing and applications. The book will be of value to researchers, professionals and students in engineering and computer science, particularly those engaged with image and speech processing, multimodal information processing, data science, and artificial intelligence.