Search

Dynamic routing networks, aimed at finding the best routing paths in the networks, have achieved significant improvements to neural networks in terms of accuracy and efficiency. In this paper, we see dynamic routing networks in a fresh light, formulating a routing method as a mapping from a sample space to a routing space. From the perspective of space mapping, prevalent methods of dynamic routing did not take into account how inference paths would be distributed in the routing space. Thus, we propose a novel method, termed CoDiNet, to model the relationship between a sample space and a routing space by regularizing the distribution of routing paths with the properties of consistency and diversity. Specifically, samples with similar semantics should be mapped into the same area in routing space, while those with dissimilar semantics should be mapped into different areas. Moreover, we design a customizable dynamic routing module, which can strike a balance between accuracy and efficiency. When deployed upon ResNet models, our method achieves higher performance and effectively reduces average computational cost on four widely used datasets.

Canonical correlation analysis (CCA) is a typical statistical model used to analyze the correlation components between different view representations of the same objects. When the label information is available with the data representations, CCA can be extended to its discriminative counterparts by incorporating supervision in the analysis. Although most discriminative variants of CCA have achieved improved results, nearly all of their objective functions are nonconvex, implying that optimal solutions are difficult to obtain. More important, that cross-view representations from the same sample should be consistent, that is, the cross-view semantic consistency has however not been modeled. To overcome these drawbacks, in this article, we propose a discriminant semantic correlation analysis (DSCA) model by modeling the cross-view semantic consistency for each object in the sample space rather than in the commonly used feature space. To boost the nonlinear discriminating capability of DSCA, we extend it from the Euclidean to the geodesic space by transforming the metric and incorporating both the cross-view semantic and representation correlation information and consequently obtain our final model with convex objective, namely, convex DSCA (C-DSCA). Finally, with extensive experiments and comparisons, we validate the effectiveness and superiority of the proposed method.

Till March 31st, 2021, the coronavirus disease 2019 (COVID-19) had reportedly infected more than 127 million people and caused over 2.5 million deaths worldwide. Timely diagnosis of COVID-19 is crucial for management of individual patients as well as containment of the highly contagious disease. Having realized the clinical value of non-contrast chest computed tomography (CT) for diagnosis of COVID-19, deep learning (DL) based automated methods have been proposed to aid the radiologists in reading the huge quantities of CT exams as a result of the pandemic. In this work, we address an overlooked problem for training deep convolutional neural networks for COVID-19 classification using real-world multi-source data, namely, the data source bias problem. The data source bias problem refers to the situation in which certain sources of data comprise only a single class of data, and training with such source-biased data may make the DL models learn to distinguish data sources instead of COVID-19. To overcome this problem, we propose MIx-aNd-Interpolate (MINI), a conceptually simple, easy-to-implement, efficient yet effective training strategy. The proposed MINI approach generates volumes of the absent class by combining the samples collected from different hospitals, which enlarges the sample space of the original source-biased dataset. Experimental results on a large collection of real patient data (1,221 COVID-19 and 1,520 negative CT images, and the latter consisting of 786 community acquired pneumonia and 734 non-pneumonia) from eight hospitals and health institutions show that: 1) MINI can improve COVID-19 classification performance upon the baseline (which does not deal with the source bias), and 2) MINI is superior to competing methods in terms of the extent of improvement.

A method for seismic reliability analysis based on a global damage model is proposed for a base-isolated structure. A sample space is created using the Latin Hypercube Sampling method, in which the randomness of structure and seismic ground motions is considered, and then dynamic elastic–plastic analysis is carried out for each sample. The cumulative damage factors of the upper structure and the isolated story are calculated using a global damage model of an isolated structure. The probability of progressive sideway collapse for base-isolated structure under earthquake action is obtained using the quadratic fourth-moment method, and this method is also used to analyze the evolution of the probability density of cumulative damage for structure. From aforementioned analysis, the change tendency of damage factor over time can be obtained using the global seismic damage model based on the quadratic fourth-moment method. Due to the introduction of the quadratic fourth-moment method in this paper, the impacts of skewness coefficient and kurtosis coefficient on the failure probability of structure can be taken into account, if the scheme of a structural parameter cannot be confirmed. The method provides results that are more accurate than those of the classical first-order second-moment method. A method for the seismic reliability analysis of an isolated structure from the component to the global structure is also established, in which the failure probabilities of the upper structure and the isolated story are considered with a global seismic damage model.

With the rapid development of genetic sequencing and DNA microarray technologies, a large number of gene expression data has been generated, which provides an important reference for tumor diagnosis. However, it is challenging to classify these gene expression data due to the high-dimensionality and small number of data samples. In this work, we propose an effective method to select the most discriminative genes from high-dimensional microarray data for benefiting tumor classification. In detail, each gene is regarded as a feature dimension and we build a novel computational model based on dual latent feature representation learning, referred as DLRL briefly, which can capture both the internal association of data samples and the relationship between different genes. Instead of measuring the importance of genes in original data space, we perform gene selection in the learned latent representation space which is more robust to noisy and redundant information. We first construct the affinity matrices for both samples and genes, which can represent the correlation information between data samples and genes, respectively. Then the dual latent representation learning is modelled via non-negative matrix factorization of the two affinity matrices. The low-dimensional latent representation matrix of sample space is treated as a pseudo-label matrix to guide the latent space projection of original data. Meanwhile, the sample projection matrix is unified with the latent representation matrix of gene space. An alternating algorithm is carefully designed to solve the resultant optimization problem. Extensive experiments on six commonly used publicly microarray datasets are conducted to demonstrate that the proposed method can steadily outperform other state-of-the-art methods in terms of microarray data classification. In addition, we also test the proposed model on a face image dataset as well as a digit image dataset to validate its efficacy for general unsupervised feature selection.

A quick and accurate extraction of dominant colors of background images is the basis of adaptive camouflage design. This paper proposes a Color Image Quick Fuzzy C-Means (CIQFCM) clustering algorithm based on clustering spatial mapping. First, the clustering sample space was mapped from the image pixels to the quantized color space, and several methods were adopted to compress the amount of clustering samples. Then, an improved pedigree clustering algorithm was applied to obtain the initial class centers. Finally, CIQFCM clustering algorithm was used for quick extraction of dominant colors of background image. After theoretical analysis of the effect and efficiency of the CIQFCM algorithm, several experiments were carried out to discuss the selection of proper quantization intervals and to verify the effect and efficiency of the CIQFCM algorithm. The results indicated that the value of quantization intervals should be set to 4, and the proposed algorithm could improve the clustering efficiency while maintaining the clustering effect. In addition, as the image size increased from 128 × 128 to 1024 × 1024, the efficiency improvement of CIQFCM algorithm was increased from 6.44 times to 36.42 times, which demonstrated the significant advantage of CIQFCM algorithm in dominant colors extraction of large-size images.

Image-space auxiliary features such as surface normal have significantly contributed to the recent success of Monte Carlo (MC) reconstruction networks. However, path-space features, another essential piece of light propagation, have not yet been sufficiently explored. Due to the curse of dimensionality, information flow between a regression loss and high-dimensional path-space features is sparse, leading to difficult training and inefficient usage of path-space features in a typical reconstruction framework. This paper introduces a contrastive manifold learning framework to utilize path-space features effectively. The proposed framework employs weakly-supervised learning that converts reference pixel colors to dense pseudo labels for light paths. A convolutional path-embedding network then induces a low-dimensional manifold of paths by iteratively clustering intra-class embeddings, while discriminating inter-class embeddings using gradient descent. The proposed framework facilitates path-space exploration of reconstruction networks by extracting low-dimensional yet meaningful embeddings within the features. We apply our framework to the recent image- and sample-space models and demonstrate considerable improvements, especially on the sample space. The source code is available at https://github.com/Mephisto405/WCMC.

In class-imbalance learning, Synthetic Minority Oversampling Technique (SMOTE) is a widely used technique to tackle class-imbalance problems from the data level, whereas SMOTE blindly selects neighboring minority class points when performing an interpolation among them and inevitably brings collinearity between the generated new points and the original ones. To combat these problems, we propose in this study an adaptive-weighting SMOTE method, termed as AWSMOTE. AWSMOTE applies two types of SVM-based weights into SMOTE. A kind of weight is used in variable space to combat the drawbacks of collinearity, while another weight is utilized in sample space to purposefully choose those support vectors from the minority class as the neighboring points in the interpolation. AWSMOTE is compared with SMOTE and its improved versions with six simulated datasets and 22 real-world datasets. The results demonstrate the effectiveness and advantages of the proposed approach.

Deep learning (DL) is an important enabling technology for intelligent manufacturing. The DL-based industrial image pattern recognition (DLBIIPR) plays a vital role in the improvement of product qu...

Co-clustering simultaneously performs clustering on the sample and feature dimensions of the data matrix, so it can obtain better insight into the data than traditional clustering. Adjustment learning extracts valuable information from chunklets for unsupervised cluster learning in specific scenarios, but in fact it can be easily extended to semi-supervised and supervised learning situations. In this paper, we propose a novel co-clustering framework, named co-adjustment learning for co-clustering (CALCC), and CALCC can be simultaneously used in unsupervised, semi-supervised and supervised learning situations. A novel co-adjustment learning (CAL) model is proposed to extract meaningful representations in both sample space and feature space for co-clustering. CAL can not only perform the sample projection as well as feature projection under the guidance of chunklet information, it can also transform the original data into another space with improved separability. We can obtain the row partition matrix and column partition matrix by performing the clustering process on the representations learned by the CAL model. In order to prove the availability of our framework, an unsupervised case of CALCC is introduced to make an extensive comparison with several related methods (specifically including the classic co-clustering methods and the state-of-the-art methods closely related to our work) on several image and real data sets. The experimental results show the superior performance of the CAL model in discovering discriminative representations and demonstrate the effectiveness of the CALCC framework. The proposed CALCC framework, as demonstrated in the experiments, is more effective superior to the related methods. In addition, the chunklet information can be effective to enhance the expression ability of the learned representations.

With the development of communication systems, the MIMO order of low-cost communication equipments have also increased. However, classical MIMO algorithms either have too high computational complexity or have poor performance. Some existing deep learning-based MIMO detection networks have achieved excellent results in high-order MIMO, but they have unbearable detection performance in low-order MIMO. Based on the idea of dividing the sample space, we propose MLNet. Compared with Det Net and MMNet, MLNet has made a huge leap in time-varying channels. At the same time, MLNet has four orders of magnitude less computation than existing neural network algorithms. We also propose a training method, which is not only useful for MLNet, but also improves the detection performance of DetNet.

This paper proposes an approach to predicting fatigue crack growth under mixed-mode loading based on improved Gaussian process. In terms of analyzing the theoretical background for fatigue crack growth, a corresponding finite element model is built to generate sufficient simulation data, which is utilized to obtain the key parameters (e.g., stress intensity factor) for the fatigue crack growth process. And then, a Gaussian process model is achieved to meet the condition that the stress intensity factor is a nonlinear continuous change in crack growth, especially for mixed-mode loading. Following, an idea of local sample densification method is implemented to improve Gaussian sample generation process according to the simplified model of the crack growth path. Based on the above investigation, a fatigue crack growth prediction model using the improved Gaussian process is finished, which is subsequently verified through the test data of lower bainite steel (SCM435) material. The results show that the proposed approach has better computational accuracy and efficiency than the traditional finite element method in predicting fatigue crack growth under mixed-mode loading.

In this paper, we have devised a multiobjective optimization solution framework for solving the problem of gene expression data clustering in reduced feature space. Here clustering problem is viewed from two different aspects: clustering of genes in reduced sample space or clustering of samples in reduced gene space. Three objective functions: two internal cluster validity indices and the count on the number of features are optimized simultaneously by a popular multiobjective simulated annealing based approach, namely AMOSA. Here, point symmetry based distance is used for the assignment of gene data points to different clusters. Seven publicly available benchmark gene expression data sets are used for experimental purpose. Both aspects of clustering in reduced feature space is demonstrated. The proposed gene expression clustering technique outperforms the existing nine clustering techniques. Apart from this, also some statistical and biological significant tests have been carried out to show that the proposed FSC-MOO technique is more statistically and biologically enriched

The phenomenon of flow-channeling, or the existence of preferential pathways for flow in fracture networks, is well known. Identification of the channels (“backbone”) allows for system reduction and computational efficiency in simulation of flow and transport through fracture networks. However, the purpose of machine learning techniques for backbone identification in fractured media is two-pronged system reduction for computational efficiency in simulation of flow and transport as well as physical insight into the phenomenon of flow channeling. The most critical aspect of this problem is the need to have a truly “physics-informed” technique that respects the constraint of connectivity. We present a method that views a network as a union of connected paths with each path comprising a sequence of fractures. Thus, the fundamental unit of selection becomes a sequence of fractures, classified based on attributes that characterize the sequence. In summary, this method represents a parametrization of the sample space that ensures every selected sample sub-network (which is the union of all selected sequences of fractures) always respects the constraint of connectivity, demonstrating that it is a truly physics-informed method. The algorithm has a user-defined parameter which allows control of the backbone size when using the random forest or logistic regression classifier. Even when the backbones are less than 30% in size (leading to computational savings), the backbones still capture the behavior of the breakthrough curve of the full network. Moreover, there is no need to check for path connectedness in the backbones unlike previous methods since the backbones are guaranteed to be connected.

Aiming at the problem that the test time is too long and the test efficiency is affected, an adaptive test patterns reordering method based on Gamma distribution was proposed. And a probability model based on Gamma distribution for the probability of each test pattern hitting fault was established. During the test, the circuit to be tested was added to the sample space, and the parameters of probability model were updated dynamically, and the test patterns were reordered synchronously. The experimental results showed that the reordered test patterns have higher test quality, which can reduce the test time and test cost of the faulty circuit. The algorithm is completely software-based and does not require any additional hardware overhead and is directly compatible with traditional integrated circuit testing process.

In many science and engineering fields that require computational models to predict certain physical quantities, we are often faced with the selection of the best model under the constraint that only a small sample set can be physically measured. One such example is the prediction of human perception of visual quality, where sample images live in a high dimensional space with enormous content variations. We propose a new methodology for model comparison named group maximum differentiation (gMAD) competition. Given multiple computational models, gMAD maximizes the chances of falsifying a “defender” model using the rest models as “attackers”. It exploits the sample space to find sample pairs that maximally differentiate the attackers while holding the defender fixed. Based on the results of the attacking-defending game, we introduce two measures, aggressiveness and resistance , to summarize the performance of each model at attacking other models and defending attacks from other models, respectively. We demonstrate the gMAD competition using three examples—image quality, image aesthetics, and streaming video quality-of-experience. Although these examples focus on visually discriminable quantities, the gMAD methodology can be extended to many other fields, and is especially useful when the sample space is large, the physical measurement is expensive and the cost of computational prediction is low.

Granular computing is a method of studying human intelligent information processing, which has advantage of knowledge discovery. In this paper, we convert a classification problem of sample space into a classification problem of fuzzy clustering granular space and propose a random fuzzy clustering granular hyperplane classifier (RFCGHC) from the perspective of granular computing. Most classifiers are only used to process numerical data, RFCGHC can process not only non-numerical data, such as information granules, but also numerical data. The classic granulation method is generally serial granulation, which has high time complexity. We design a parallel distributed granulation method to enhance efficiency. First, a clustering algorithm with adaptive cluster center number is proposed, where the ratio of standard deviation between categories and standard deviation within categories is as evaluation criterion. The clusters and the optimal amount of cluster centers can be achieved by the method. On the basis of these, sample set can be divided into many subsets and each sample can be granulated by these cluster centers. Then, a fuzzy clustering granular space can be formed, where fuzzy clustering granules, fuzzy clustering granular vectors, and their operators can be defined. In order to get the optimal hyperplane in the fuzzy clustering granular space to classify these samples, we design a loss function and evaluate each category with probability by fuzzy clustering granular hyperplane. In solving the loss function, genetic algorithm based on fuzzy clustering granules is adopted. Experimental results and theoretical analysis show that RFCGHC has good performance.

The consensus of multiple views on the same data will provide extra regularization, thereby improving accuracy. Based on this idea, we proposed a novel Knowledge-Supervised Learning (KSL) method for person re-identification (Re-ID), which can improve the performance without introducing extra inference cost. Firstly, we introduce isomorphic auxiliary training strategy to conduct basic multiple views that simultaneously train multiple classifier heads of the same network on the same training data. The consensus constraints aim to maximize the agreement among multiple views. To introduce this regular constraint, inspired by knowledge distillation that paired branches can be trained collaboratively through mutual imitation learning. Three novel constraints losses are proposed to distill the knowledge that needs to be transferred across different branches: similarity of predicted classification probability for cosine space constraints, distance of embedding features for euclidean space constraints, hard sample mutual mining for hard sample space constraints. From different perspectives, these losses complement each other. Experiments on four mainstream Re-ID datasets show that a standard model with KSL method trained from scratch outperforms its ImageNet pre-training results by a clear margin. With KSL method, a lightweight model without ImageNet pre-training outperforms most large models. We expect that these discoveries can attract some attention from the current de facto paradigm of "pre-training and fine-tuning" in Re-ID task to the knowledge discovery during model training.

In order to solve the problem that the number of clusters needs to be specified artificially in the partition clustering and hierarchical clustering methods, this paper proposes an adaptive clustering algorithm which can search the number of clusters independently through the self evolution pattern of samples without any prior knowledge (AdaCluster). Adacluster algorithm includes two stages: cluster number self evolution stage and label propagation stage. In the evolution stage, a small number of samples are randomly sampled from the sample space, and then the number of clusters is independently evolved according to the designed control parameters in an incremental form to obtain the initial clustering results; In the label propagation stage, the cluster labels obtained in the evolution stage are propagated to the unlabeled samples. Adacluster algorithm is completely based on self evolution pattern, and can determine the number of clusters without any prior parameters. The Adacluster algorithm is compared with other typical algorithms on a large number of datasets, and applied in image segmentation. The results show that Adacluster algorithm has excellent clustering performance and high robustness.

When using the naive Bayesian classifier (NBC), it is necessary to assume that the attributes are independent. Although the relationship between the attributes can be analyzed by the Bayesian network (BN), the network structure and parameters need to be trained. Meanwhile, each continuous attribute is given a priori probability distribution for both the NBC and the BN. If the actual probability distribution of the samples is inconsistent with the priori probability distribution, the classification result is poor. In this paper, a model-free Bayesian classifier (MFBC) is proposed to cope with the drawbacks of the NBC and the BN. The information of the sample space and the joint probability density can be obtained based on the nearest neighbor (NN) strategy. The proposed MFBC is able to cope with discrete or continuous attributes even if there is no a priori probability distribution of the attributes. It is also not necessary to build a network structure to obtain a unitive calculation framework for the attributes. Moreover, the probability distribution for limited amounts of samples is approximated using the NN method. Then, the classification label of a sample can be predicted by the MFBC. The sensitivity analysis of the MFBC is obtained using the UCI data sets. Compared with some classical and ensemble classifiers, the analysis results verify the effectiveness and convergence performance of the MFBC.

Unsupervised multi-task learning exploits the shared knowledge to improve performances by learning related tasks simultaneously. In this paper, we propose an unsupervised multi-task learning method with hierarchical data structure. It strengthens similarities between instances in the same cluster, and increases diversities of instances by utilizing instances from related clusters. Firstly, we introduce Representative Dual Features (RepDFs) that possess representative capabilities in the feature space and the sample space for each cluster concurrently. Secondly, we explore hierarchical structural similarities between clusters in related tasks from the topological perspective: 1) feature basis matrix, which learns compact representations for features in the feature space; and 2) sample refined matrix, which preserves local structures in the sample space. Thirdly, we adopt RepDFs to measure correlations between clusters and incorporate hierarchical structural similarities to conduct knowledge transfer among tasks. Experimental results on real-world data sets demonstrate the effectiveness and superiority of the proposed method over existing multi-task clustering methods.

In this paper, a heterogeneous borehole radar fusion-before-imaging method (RFBI) is proposed and validated. Different from traditional fusion methods, RFBI firstly fuses radar sample sets from heterogeneous radar systems (such as monostatic, bistatic and multi-input multi-output) and then do imaging process once in high-dimensional space to obtain imaging result with high accuracy. Specifically, according to the relative positions of transceivers, the radar sample sets to be imaged are inserted into a high-dimensional sample space for fusion, and then the fused data is converted into virtual target space according to the conversion relationship between echo space and target space in frequency-wavenumber domain. Finally, the imaging result is extracted from the target space according to the imaging conditions. In the fused high-dimensional sample space of RFBI, the data diversity and redundancy are utilized, and the impact of clutters and other interference factors on imaging is effectively reduced, which brings the benefit of high imaging accuracy to RFBI. Meanwhile, all radar sampling sets are imaged once, which enable RFBI owns the advantage of high efficiency. The comparison results with conventional fusion-imaging methods indicates and validates that RFBI is more suitable for heterogeneous radar system imaging applications.

The research literature suggests that Scratch software allows children to approach mathematical concepts and procedures (e.g., Batista & Baptista, 2014). One of its foremost qualities is the capacity to motivate, engage, and immerse players. The current chapter contributes to explore the capabilities of a learning environment that uses programming logic in a game setting. Based on challenging students (aged 8–13) to create their own games, we attempted to explore and enhance their reasoning about probabilistic ideas. Children were introduced to the block-based programming language Scratch 2.0 and used it to create their own games. The children on this research have a double role: the one of the designer and the one of the player. We explored these roles by understanding their thinking on implementing probability in their game and understanding the complex of randomness and sample space. The findings show that the idea of chance had an important role in their games and that they expressed probabilistic ideas while designing and playing their game.

Introduction: Nowadays the use of Computer-Assisted Qualitative Data Analysis Software (CAQDAS) is essential in qualitative research methodology. However, programs of this line such as ALCESTE and IRAMUTEQ have as many positive as negative points. Knowing this, it is important to analyze the usage of the aforementioned Software packages in qualitative research. Objectives: To analyze the functionalities, accessibility, and limitations of CAQDAS as main methodological processing in different qualitative surveys. Methods: Critical Comparison between two quanti-qualitative surveys previously performed by the authors of this scientific analysis. Results: The use of each software was efficient in the management of the data of non-categorical research. However, IRAMUTEQ proved to be more advantageous than ALCESTE. Discussion: The use of a CAQDAS is essential in researches with an extensive sample space of qualitative data meaning, in the current context, a great advance of this research segment, since it puts it in a level of greater assertiveness. However, the selection of which program to use varies with the affinity of the researcher in relation to them and which resources they wish to be used to generate the results, being IRAMUTEQ the software with more options for that. Conclusion: Therefore, the management of CAQDAS is one of the most promising methodological paths for qualitative research. Its benefits overcome the limitations that its algorithms have and more research on it is necessary for the execution of others to be improved.

An optimized neural network classification method based on kernel holistic learning and division (KHLD) is presented. The proposed method is based on the learned radial basis function (RBF) kernel as the research object. The kernel proposed here can be considered a subspace region consisting of the same pattern category in the training sample space. By extending the region of the sample space of the original instances, relevant information between instances can be obtained from the subspace, and the classifier’s boundary can be far from the original instances; thus, the robustness and generalization performance of the classifier are enhanced. In concrete implementation, a new pattern vector is generated within each RBF kernel according to the instance optimization and screening method to characterize KHLD. Experiments on artificial datasets and several UCI benchmark datasets show the effectiveness of our method.

Generative adversarial networks (GAN) could synthesize semantically meaningful data from standard signal distribution, which make it have considerable potential to alleviate data scarcity. In this paper, based on Evolutionary GAN, cardiac magnetic resonance images enhancement method is proposed to solve over-fitting problem caused by training convolution network with small dataset. The most optimal generator which consider the quality and diversity of generated images simultaneously from many generator mutations is chosen. Meanwhile, to expand the whole training set distribution, we combine the linear interpolation of eigenvectors to synthesize new training samples and synthesize related linear interpolation labels, which can make the discrete sample space become continuous to improve the smoothness between domains. In this paper, the effectiveness of this method is verified by classification experiments, and the influence of the proportion of synthesized samples on the classification results of cardiac magnetic resonance images is explored.

Recently, researchers begin using generated as well as transferred images to expand the training dataset of person re-identification. It expands the training dataset’s sample space, which contributes to learning more robust features for the Re-ID network. However, these works only transfer a certain kind of feature between images. There is no known attempt to transfer multiple features across multiple images. By transferring multiple features, the sample space expands further, allowing Re-ID networks to learn more robust features. In this paper, we propose a unified framework and pipeline to integrate multiple feature transfer networks. Users are also free to determine how to transfer these features. Based on the above framework and pipeline, we create a large amount of transferred images from the Market-1501 dataset and create an expanded dataset named Market1501-EX. Further more, we propose a corresponding person identification labeling method for identity loss, using the generation information provided by the dataset. We use this dataset to train a Re-ID network, and the training result in this paper rests at mid-high level of related works which also utilize generated images.

Recommender systems usually amplify the biases in the data. The model learned from historical interactions with imbalanced item distribution will amplify the imbalance by over-recommending items from the majority groups. Addressing this issue is essential for a healthy ecosystem of recommendation in the long run. Existing work applies bias control to the ranking targets (e.g., calibration, fairness, and diversity), but ignores the true reason for bias amplification and trades off the recommendation accuracy. In this work, we scrutinize the cause-effect factors for bias amplification, identifying the main reason lies in the confounding effect of imbalanced item distribution on user representation and prediction score. The existence of such confounder pushes us to go beyond merely modeling the conditional probability and embrace the causal modeling for recommendation. Towards this end, we propose a Deconfounded Recommender System (DecRS), which models the causal effect of user representation on the prediction score. The key to eliminating the impact of the confounder lies in backdoor adjustment, which is however difficult to do due to the infinite sample space of the confounder. For this challenge, we contribute an approximation operator for backdoor adjustment which can be easily plugged into most recommender models. Lastly, we devise an inference strategy to dynamically regulate backdoor adjustment according to user status. We instantiate DecRS on two representative models FM [32] and NFM [16], and conduct extensive experiments over two benchmarks to validate the superiority of our proposed DecRS.

KNN classification is an improvisational learning mode, in which they are carried out only when a test data is predicted that set a suitable K value and search the K nearest neighbors from the whole training sample space, referred them to the lazy part of KNN classification. This lazy part has been the bottleneck problem of applying KNN classification due to the complete search of K nearest neighbors. In this paper, a one-step computation is proposed to replace the lazy part of KNN classification. The one-step computation actually transforms the lazy part to a matrix computation as follows. Given a test data, training samples are first applied to fit the test data with the least squares loss function. And then, a relationship matrix is generated by weighting all training samples according to their influence on the test data. Finally, a group lasso is employed to perform sparse learning of the relationship matrix. In this way, setting K value and searching K nearest neighbors are both integrated to a unified computation. In addition, a new classification rule is proposed for improving the performance of one-step KNN classification. The proposed approach is experimentally evaluated, and demonstrated that the one-step KNN classification is efficient and promising, 13 pages. in IEEE Transactions on Knowledge and Data Engineering

The uncertainty of intelligence software makes it impossible for even developers to determine the output corresponding to each input. Therefore, it is particularly important to find a way to find intelligence software errors efficiently. Fuzzing testing technology is considered to be an effective technology for discovering software vulnerabilities, The selected test case generation strategy will affect its testing efficiency. In this article, we propose a counterexample-guided fuzzing method. According to the generated counterexamples, the scope of the sample space is continuously reduced, so as to find more counterexamples in a limited number of samples and improve the ability to uncover mistakes. In our research, The ability about revealing errors of test object has increased by 20%.

On the basis of ensuring insulation and current carrying capacity, set the electric field and temperature as constraints, the GIS/GIL co-simulation model was established with the least cost as the optimization goal. The optimization process usually requires hundreds of iterative calculations, which resulted in inefficiency of the complex and time-consuming 2D/3D co-simulation model and difficulties of meeting the needs of product development. Firstly, cosimulation models of static electric field, eddy current and dynamic fluid were established. Then using the Latin hypercube sampling method, 50 samples were extracted from the optimized sample space, and the sensitivity analysis was performed based on the results of co-simulation model. Based on the sensitivity analysis results, the simplified model was established by using the least squares fitting method. The simplified model results are verified by co-simulation model. The results showed that calculation consistency of simplified model with co-simulation model is about 98.9%, while the computational efficiency of the simplified model is 4000 times that of the co-simulation model. After optimization, the busbar cost is reduced by 26.65% on the basis of ensuring insulation and current-carrying performance.

Sea surface temperature (SST) prediction has widespread applications in the field of marine ecology, fisheries, sports and climate change studies. At present, the real-time SST forecasts are made by numerical models which are categorically based on physics-based assumptions subjected to boundary and initial conditions. They are more suited to a large spatial region than in a specific location. In this study, location-specific SST forecasts were made by combining deep learning neural networks with numerical estimators at five different locations around India for three different time horizons (daily, weekly and monthly). Firstly, forecasts were made with traditional neural networks (NNs) and then through deep learning networks. The NNs significantly improved on the results achieved by numerical forecasts which were further enhanced by the deep learning long short-term memory (LSTM) neural network over all timescales and at all the selected sites. The model was performed successfully in terms of various statistical parameters with correlation values nearing 1.0 while minimizing the errors. Additionally, a comparative study with a linear system, the autoregressive integrated moving average with exogenous input was made. The predictive skills of deep learning LSTMs are found to be more attractive than the other existing techniques (linear or other NNs) due to their ability of learning long time dependencies and extracting features from a sample space.

Most existing methods only consider the local features of the samples, and their experimental results show better performance than traditional Non-deep learning methods. However, in these methods, the global features of the sample space are usually ignored, and these ignored global features will affect the classification accuracy. To solve this problem, a novel triple capsule network framework is proposed to text classification. The training in the first stage, to obtain a basic capsule network for obtaining local features. Then, three capsule networks sharing parameters are combined spatially, and the triplet loss function is used in the second stage of training. By comparative learning, the capsule network can learn global features that can represent the spatial distance between different categories. Through comparison experiments on six datasets and ten general benchmark algorithms, the results show that our results is the first in the four datasets.

Principal component analysis (PCA) reduces dimensionality by generating uncorrelated variables and improves the interpretability of the sample space. This analysis focused on assessing the value of PCA for improving the classification accuracy of failures within current-voltage (IV) traces. Our results show that combining PCA with random forests improves classification by only ∼1% (bringing the accuracy to >99%), compared to a baseline of only random forests (without PCA) of >98%. The inclusion of PCA, however, does provide an opportunity to study an interesting representation of all of the features on a single, two-dimensional feature space. A visualization of the first two principal components (similar to IV profile but rotated) captures how the inclusion of a current differential feature causes a notable separation between failure modes due to their effect on the slope. This work continues the discussion of generating different ways of extracting information from the IV curve, which can help with failure classification - especially for failures that only exhibit marginal profile changes in IV curves.

As we all know, science and technology are developing faster and faster. Many experts and scholars have demonstrated that human diseases are related to lncRNA, but only a few associations have been confirmed, and many unknown associations need to be found. In the process of finding associations, it takes a lot of time, so finding an efficient way to predict the associations between lncRNAs and diseases is particularly important. In this paper, we propose a multi-label fusion collaborative matrix factorization (MLFCMF) approach for predicting lncRNA-disease associations (LDAs). Firstly, the lncRNA space and disease space are optimized by multi-label to enhance the intrinsic link between lncRNA and disease and to tap potential information. Multi-label learning can encode a variety of data information from the sample space. Secondly, to learn multi-label information in the data space, the fusion method is used to handle the relationship between multiple labels. More comprehensive information will be obtained by weighing the effects of different labels. The addition of Gaussian interaction profile (GIP) kernel can increase the network similarity. Finally, the lncRNA-disease associations are predicted by the method of collaborative matrix factorization. The ten-fold cross-validation method is used to evaluate the MLFCMF method, and our method finally obtains an AUC value of 0.8612. Detailed analysis of ovarian cancer, colorectal cancer, and lung cancer in the simulation experiment results. So it can be seen that our method MLFCMF is an effective model for predicting lncRNA-disease associations.

Generative adversarial network (GAN) has been widely explored for cross-modality medical image synthesis. The existing GAN models usually adversarially learn a global sample space mapping from the source-modality to the target-modality and then indiscriminately apply this mapping to all samples in the whole space for prediction. However, due to the scarcity of training samples in contrast to the complicated nature of medical image synthesis, learning a single global sample space mapping that is “optimal” to all samples is very challenging, if not intractable. To address this issue, this paper proposes sample-adaptive GAN models, which not only cater for the global sample space mapping between the source- and the target-modalities but also explore the local space around each given sample to extract its unique characteristic. Specifically, the proposed sample-adaptive GANs decompose the entire learning model into two cooperative paths. The baseline path learns a common GAN model by fitting all the training samples as usual for the global sample space mapping. The new sample-adaptive path additionally models each sample by learning its relationship with its neighboring training samples and using the target-modality features of these training samples as auxiliary information for synthesis. Enhanced by this sample-adaptive path, the proposed sample-adaptive GANs are able to flexibly adjust themselves to different samples, and therefore optimize the synthesis performance. Our models have been verified on three cross-modality MR image synthesis tasks from two public datasets, and they significantly outperform the state-of-the-art methods in comparison. Moreover, the experiment also indicates that our sample-adaptive strategy could be utilized to improve various backbone GAN models. It complements the existing GANs models and can be readily integrated when needed.

In this paper, a deep learning-based approach is applied to high dimensional, high-volume, and high-sparsity medical data to identify critical casual attributions that might affect the survival of a breast cancer patient. The Surveillance Epidemiology and End Results (SEER) breast cancer data is explored in this study. The SEER data set contains accumulated patient-level and treatment-level information, such as cancer site, cancer stage, treatment received, and cause of death. Restricted Boltzmann machines (RBMs) are proposed for dimensionality reduction in the analysis. RBM is a popular paradigm of deep learning networks and can be used to extract features from a given data set and transform data in a non-linear manner into a lower dimensional space for further modelling. In this study, a group of RBMs has been trained to sequentially transform the original data into a very low dimensional space, and then the k-means clustering is conducted in this space. Furthermore, the results obtained about the cluster membership of the data samples are mapped back to the original sample space for interpretation and insight creation. The analysis has demonstrated that essential features relating to breast cancer survival can be effectively extracted and brought forward into a much lower dimensional space formed by RBMs.

Imaging spectroscopy is a powerful new approach for observing aspects of the biological diversity of the Earth, conveying information about plant functional traits, habitat, and plant diversity itself. For decades, spectroscopic data suitable for this application have mainly been collected by aircraft. But in the next decade, global coverage from space by high-quality spectroscopic data will become available, preceded by instruments providing “global access”—not wall-to-wall coverage but data from almost anywhere in the world. For decades, scientists have experimented with and discussed optimal strategies for collecting spectroscopic data, but the next set of missions is now sufficiently well-defined that ecologists should consider how best to use the data that can now be expected. The anticipated flood of data will provide a new window on diversity, characterizing it in new ways that comprehensively sample space and change over time. Spectroscopic data will be peta-scale or larger, perhaps as much as 10 TB per day, and the data themselves will be high dimensional, requiring and allowing advanced big data techniques to be fully exploited. These data raise specific challenges such as how to characterize aggregate ecosystem characteristics, since the traits observed will change with phenology. Pixels will be fixed at ~30 m, 10–106times larger than the plants they sample; other instrument objectives are likely in the range of 10 nm spectral sampling, coverage from 400 to 2500 nm with signal to noise in the range of 250–400. Imaging spectroscopy from space represents a huge opportunity for global ecology, but many conceptual, algorithmic, and theoretical issues will challenge the users.

The steel bridge deck system, directly subjected to the vehicle load, is an important component to be considered in the optimization design of the bridges. Due to its complex structure, the design parameters are coupled with each other, and many fatigue details in the system result in time-consuming calculation during structure optimization. In view of this, a nonlinear optimization method based on the response surface methodology (RSM) is proposed in this study to simplify the design process and to reduce the amount of calculations during optimization. The optimization design of the steel bridge deck system with two-layer pavement on the top of the steel deck plate is taken as an example, the influence of eight structural parameters is considered. The Box-Behnken design is used to construct a sample space in which the eight structural parameters can be distributed evenly to reduce the calculation workload. The finite element method is used to model the mechanical responses of the steel bridge deck system. From the regression analysis by the RSM, the explicit relationships between the fatigue details and the design parameters can be obtained, based on which the nonlinear optimization design of the bridge deck system is conducted. The influence of constraint functions, objective functions, and optimization algorithms is also analyzed. The method proposed in this study is capable of considering the influence of different structural parameters and different optimization objectives according to the actual needs, which will effectively simplify the optimization design of the steel bridge deck system.

Domain adaptive person re-identification (re-ID) is a challenging task, especially when person identities in target domains are unknown. Existing methods attempt to address this challenge by transferring image styles or aligning feature distributions across domains, whereas the rich unlabeled samples in target domains are not sufficiently exploited. This paper presents a novel augmented discriminative clustering (AD-Cluster) technique that estimates and augments person clusters in target domains and enforces the discrimination ability of re-ID models with the augmented clusters. AD-Cluster is trained by iterative density-based clustering, adaptive sample augmentation, and discriminative feature learning. It learns an image generator and a feature encoder which aim to maximize the intra-cluster diversity in the sample space and minimize the intra-cluster distance in the feature space in an adversarial min-max manner. Finally, AD-Cluster increases the diversity of sample clusters and improves the discrimination capability of re-ID models greatly. Extensive experiments over Market-1501 and DukeMTMC-reID show that AD-Cluster outperforms the state-of-the-art with large margins., Comment: Accepted by CVPR'20

With the rapid growth of complex high-dimensional sparse data and the limitation of a single perspective, there is an increasing demand for new methods of feature selection from multiple perspectives. The feature selection method based on minimum regression is usually learning projection matrix, which is lack of theoretical explanation to evaluate the importance of features. Moreover, these methods cannot find the global and sparse solutions of the projection matrix. In this paper, we propose a new multi-view unsupervised feature selection method, which can learn the global and sparse solutions of the projection matrix. The new method extends the least squares regression model by adjusting the regression coefficients in least squares regression by using a set of scale factors for feature ranking. It shows that the new model can learn global and sparse solutions. In addition, the introduction of the scale factor provides a theoretical explanation for why we can use the projection matrix to sort features. In order to optimize the new model, a simple and effective algorithm for proof convergence is proposed. By classifying and clustering the data after feature selection, we can see the superiority of our algorithm.

Coreset, which is a summary of the original dataset in the form of a small weighted set in the same sample space, provides a promising approach to enable machine learning over distributed data. Although viewed as a proxy of the original dataset, each coreset is only designed to approximate the cost function of a specific machine learning problem, and thus different coresets are often required to solve different machine learning problems, increasing the communication overhead. We resolve this dilemma by developing robust coreset construction algorithms that can support a variety of machine learning problems. Motivated by empirical evidence that suitably-weighted k -clustering centers provide a robust coreset, we harden the observation by establishing theoretical conditions under which the coreset provides a guaranteed approximation for a broad range of machine learning problems, and developing both centralized and distributed algorithms to generate coresets satisfying the conditions. The robustness of the proposed algorithms is verified through extensive experiments on diverse datasets with respect to both supervised and unsupervised learning problems.

This paper presents a divide-and-conquer (DC) approach based on feature-space decomposition for classification. When large-scale data sets are present, typical approaches usually employed truncated kernel methods on the feature space or DC approaches on the sample space. However, this did not guarantee separability between classes, owing to overfitting. To overcome such problems, this paper proposes a novel DC approach on feature spaces consisting of three steps. First, we divide the feature space into several subspaces using the decomposition method proposed in this paper. Subsequently, these feature subspaces are sent into individual local classifiers for training. Finally, the outcome of local classifiers are fused together to generate the final classification results. We also propose a Cascade-TRBFKRR classifier to reweight training samples for data refinement. Experiments on large-scale data sets are carried out for performance evaluation. The results show that the error rates of the proposed DC method decreased compared with the state-of-the-art fast support vector machine solvers, e.g., reducing error rates by 10.53% and 7.53% on RCV1 and covtype data sets, respectively.

In this paper, we propose a manifold regularized correlation tracking method with augmented samples. To make better use of the unlabeled data and the manifold structure of the sample space, a manifold regularization-based correlation filter is introduced, which aims to assign similar labels to neighbor samples. Meanwhile, the regression model is learned by exploiting the block-circulant structure of matrices resulting from the augmented translated samples over multiple base samples cropped from both target and nontarget regions. Thus, the final classifier in our method is trained with positive, negative, and unlabeled base samples, which is a semisupervised learning framework. A block optimization strategy is further introduced to learn a manifold regularization-based correlation filter for efficient online tracking. Experiments on two public tracking data sets demonstrate the superior performance of our tracker compared with the state-of-the-art tracking approaches.

Identification of characteristic genes associated with specific biological processes of different cancers could provide insights into the underlying cancer genetics and cancer prognostic assessment. It is of critical importance to select such characteristic genes effectively. In this paper, a novel unsupervised characteristic gene selection method based on sample learning and sparse filtering, Sample Learning based on Deep Sparse Filtering (SLDSF), is proposed. With sample learning, the proposed SLDSF can better represent the gene expression level by the transformed sample space. Most unsupervised characteristic gene selection methods did not consider deep structures, while a multilayer structure may learn more meaningful representations than a single layer, therefore deep sparse filtering is investigated here to implement sample learning in the proposed SLDSF. Experimental studies on several microarray and RNA-Seq datasets demonstrate that the proposed SLDSF is more effective than several representative characteristic gene selection methods (e.g., RGNMF, GNMF, RPCA and PMD) for selecting cancer characteristic genes.

A novel approach in the field of classifier ensembles is proposed.The approach uses an ensemble of random decision trees.Each decision tree is trained on a different area of the input space.Areas can overlap and a good coverage of the input space is ensured.Experimental results confirm the validity of the approach. Ensemble methods have shown to be more effective than monolithic classifiers, in particular when diversity holds among their components. How to enforce diversity in classifier ensembles has received much attention from machine learning researchers, yielding a variety of different techniques and algorithms. In this paper, a novel algorithm for ensemble classifiers is proposed, in which ensemble components are trained with focus on different regions of the sample space. In so doing, diversity is mainly a consequence of the intention to limit the scope of base classifiers. The algorithm proposed in this paper shares roots with several ensemble paradigms, in particular with random forests, as it generates forests of decision trees as well. As decision trees are trained with focus on specific subsets of the sample space, the resulting ensemble is in fact a forest of local trees. Comparative experimental results highlight that, on average, these ensembles perform better than other relevant kinds of ensemble classifiers, including random forests.

Principal component analysis (PCA) and linear discriminant analysis (LDA) have been extended to be a group of classical methods in dimensionality reduction for unsupervised and supervised learning, respectively. However, compared with the PCA, the LDA loses several advantages because of the singularity of its between-class scatter, resulting in singular mapping and restriction of reduced dimension. In this paper, we propose a dimensionality reduction method by defining a full-rank between-class scatter, called reversible discriminant analysis (RDA). Based on the new defined between-class scatter matrix, our RDA obtains a nonsingular mapping. Thus, RDA can reduce the sample space to arbitrary dimension and the mapped sample can be recovered. RDA is also extended to kernel based dimensionality reduction. In addition, PCA and LDA are the special cases of our RDA. Experiments on the benchmark and real problems confirm the effectiveness of the proposed method.

This paper focuses on the specific problem of multiview learning where samples have the same feature set but different probability distributions, e.g., different viewpoints or different modalities. Since samples lying in different distributions cannot be compared directly, this paper aims to learn a latent subspace shared by multiple views assuming that the input views are generated from this latent subspace. Previous approaches usually learn the common subspace by either maximizing the empirical likelihood, or preserving the geometric structure. However, considering the complementarity between the two objectives, this paper proposes a novel approach, named low-rank discriminant embedding (LRDE), for multiview learning by taking full advantage of both sides. By further considering the duality between data points and features of multiview scene, i.e., data points can be grouped based on their distribution on features, while features can be grouped based on their distribution on the data points, LRDE not only deploys low-rank constraints on both sample level and feature level to dig out the shared factors across different views, but also preserves geometric information in both the ambient sample space and the embedding feature space by designing a novel graph structure under the framework of graph embedding. Finally, LRDE jointly optimizes low-rank representation and graph embedding in a unified framework. Comprehensive experiments in both multiview manner and pairwise manner demonstrate that LRDE performs much better than previous approaches proposed in recent literatures.

This paper presents a new general methodological approach to imbalanced learning as one of the challenging problems in pattern classification. The presented method is founded on maximization of the sample entropy. The method involves detection of distributive properties of ideally balanced regular lattice sample and acceptable transfer of these properties to an arbitrary imbalanced sample increasing its representativeness. The proposed procedure assumes undersampling applied on areas of high probability density in the sample space combined with oversampling in the areas of low density. The main achievement of this method is the increased sample class entropy which reduces the inductive learner’s tendency to favor prominent class, or cluster. In addition to class balancing, this method can be useful for function approximation, clustering, and sample dimension reduction. The high degree of generality of the method implies its applicability on data of various complexity and imbalance. The presented theoretical foundation of the method was verified on a set of proper synthetic samples. The method’s practical usability is confirmed by a comparative classification of a large set of databases including speech signal samples.