Search

Recently there has been increasing attention towards analysis dictionary learning. In analysis dictionary learning, it is an open problem to obtain the strong sparsity-promoting solutions efficiently while simultaneously avoiding the trivial solutions of the dictionary. In this paper, to obtain the strong sparsity-promoting solutions, we employ the ℓ 1∕2 norm as a regularizer. The very recent study on ℓ 1∕2 norm regularization theory in compressive sensing shows that its solutions can give sparser results than using the ℓ 1 norm. We transform a complex nonconvex optimization into a number of one-dimensional minimization problems. Then the closed-form solutions can be obtained efficiently. To avoid trivial solutions, we apply manifold optimization to update the dictionary directly on the manifold satisfying the orthonormality constraint, so that the dictionary can avoid the trivial solutions well while simultaneously capturing the intrinsic properties of the dictionary. The experiments with synthetic and real-world data verify that the proposed algorithm for analysis dictionary learning can not only obtain strong sparsity-promoting solutions efficiently, but also learn more accurate dictionary in terms of dictionary recovery and image processing than the state-of-the-art algorithms., (Copyright © 2017 Elsevier Ltd. All rights reserved.)

Dynamic off-chain routing in payment channel network (PCN)-based Internet of Things (IoT) is attracting increasing research attention. However, there are two major issues in dynamic routing in PCN-based IoT with resource-limited devices. The first issue is how to achieve high long-term transaction efficiency in PCN with dynamic channel capacities. The second issue is how to achieve a lightweight routing algorithm deployed on IoT devices while achieving high transaction efficiency, i.e., successful payment amount and success ratio. Therefore, in this paper, we propose a compact deep reinforcement learning (DRL) algorithm to learn the joint dynamic and lightweight routing policy for maximizing long-term transaction efficiency. To obtain optimal performance in dynamic routing problems for off-chain systems, a proximal policy optimization algorithm is employed to create an actor–critic learning structure for training the teacher DRL model. To obtain a compact and efficient student DRL model, an adaptive pruning technique is utilized for pruning unnecessary parameters of networks in the teacher model adaptively without affecting its learning ability. Furthermore, knowledge distillation is leveraged to improve the performance of the student network. Thus, a compact and efficient student DRL model can be developed and implemented to maximize the long-term transaction efficiency in off-chain systems on resource-limited IoT devices. The simulation results demonstrate that the proposed DRL algorithm outperforms the other baseline algorithms in PCN transaction efficiency while requiring only 10% of the computation and storage resources compared with that of the original teacher model. [ABSTRACT FROM AUTHOR]

The classification of iron ore rock plays a crucial role in assisting the mine to develop scientifically sound mining strategies. Given the limitations of traditional manual identification and chemical techniques in terms of cost and detection time, it is crucial to develop an economical, high-precision, and efficient classification method for iron ore rocks. This paper takes an iron mine in Anshan as the research center and aims to establish an iron ore rock classification model by combining the improved extreme learning machine (ELM) with near-infrared spectra (NIRS). The NIRS of iron ore rock is denoised and feature extracted using the Savitzky–Golay filter and competitive adaptive reweighted sampling (CARS) methods, respectively. In this paper, we propose the IAdaBoost algorithm by introducing the sample similarity relationship into the weight-updating strategy of the AdaBoost integration algorithm. The weighted extreme learning machine (WELM) is adopted as the base learner for the IAdaBoost algorithm. Furthermore, we employ a multi-strategy improved gray wolf optimization (IGWO) algorithm to optimize the learning rate and the number of integrated base learners in the IAdaBoost algorithm. Consequently, we propose the IGWO-IAdaBoost-WELM model. Compared to other machine learning models, our proposed model performs better and can classify iron ore rocks consistently and efficiently. Meanwhile, we also used the model to invert the remote sensing images of the iron ore mining area, and the results show that the model can achieve fast, efficient, and large-scale iron ore rock classification, which provides an important reference for mining. [Display omitted] • Combining spectroscopy with machine learning methods to classify iron ore rocks. • The machine learning structure was improved using an integrated algorithm and its parameters were optimized using a multi-strategy improved optimization algorithm. • Inversion of remote sensing images of mining areas using improved machine learning methods enables fast, efficient and large-scale iron ore rock classification, which provides an important reference for mining. [ABSTRACT FROM AUTHOR]

Deep nonnegative matrix factorization (DMF) is a particularly useful technique for learning data representation in low-dimensional space. To further obtain the complex hidden information and keep the geometrical structures of the high-dimensional data, we propose a novel deep matrix factorization model with the graph regularization (called DGsnMF). For solving the model with multi-variables, we design a forward–backward splitting scheme. After that, the convergence analysis is attached to the proposed algorithm and it is proved to converge to a critical point. Empirical experiments on benchmark datasets show that the proposed method is superior to the compared methods. [ABSTRACT FROM AUTHOR]

In recent years, there has been growing concerns on the study of dictionary learning with the nonconvex sparsity-including penalty. However, how to efficiently address the dictionary learning with the nonconvex penalty is still an open problem. In this paper, we present an efficient DC-based algorithm for dictionary learning with the nonconvex smoothly clipped absolute deviation (SCAD) penalty for strong sparsity and accurate estimation. The optimization problem we considered can be generalized as a minimization of the representation error with the SCAD penalty. The approach we proposed is based on a decomposition scheme which decomposes the whole problem into a set of subproblems with regard to single-vector factors. For handling the nonconvexity of the representation error in the subproblems, we use an alternating optimization scheme to update one factor with the other factor fixed. For tackling the nonconvexity of the SCAD penalty in the subproblems, we apply the Difference of Convex functions (DC) technology to convert the nonconvex subproblem into the resulting convex problems and thus employ DC algorithm to solve the corresponding optimization; thus the simple and straightforward solutions in the closed form can be easily derived. As verified by the numerical experiments with synthetic and real-world data, the proposed algorithm performs better than the state-of-the-art algorithms with different sparsity-including constraints. [ABSTRACT FROM AUTHOR]

In this study, we propose two analysis dictionary learning algorithms for sparse representation with analysis model. The problem is formulated with the ℓ 1 -norm regularizer and with two penalty terms on the analysis dictionary: the term of − log det ( Ω T Ω ) and the coherence penalty term. As the processing scheme, we employ a block coordinate descent framework, so that the overall problem is transformed into a set of minimizations of univariate subproblems with respect to a single-vector variable. Each subproblem is still nonsmooth, but it can be solved by a proximal operator and then the closed-form solutions can be obtained directly and explicitly. In particular, the coherence penalty, excluding excessively similar or repeated dictionary atoms, is solved at the same time as the dictionary update, thereby reducing the complexity. Furthermore, a scheme with a group of atoms is introduced in one proposed algorithm, which has a lower complexity. According to our analysis and simulation study, the main advantages of the proposed algorithms are their greater dictionary recovery ratios especially in the low-cosparsity case, and their faster running time of reaching the stable values of the dictionary recovery ratios and the recovery cosparsity compared with state-of-the-art algorithms. In addition, one proposed algorithm performs well in image denoising and in noise cancellation. [ABSTRACT FROM AUTHOR]

[Display omitted] • Data augmentation based on spectral difference and content difference. • Spectral subtraction based on maximum or minimum values. • Augmentation of all data or training data based on spectral and content differences. • Model construction based on spectral difference and content difference. • Prediction for content of new samples through ensemble averaging. Soil analysis makes for developing precision agriculture and monitoring land quality, while the models available for spectroscopy-based chemometrics are constrained by limited samples from small areas. The paper proposed sample expansion and model construction based on spectral difference and content difference, realizing data augmentation and deep learning applied to original samples with limited numbers. The spectral subtraction based on maximum or minimum values exploited the maximum or minimum values to acquire the spectral difference and content difference, which provided a new data form for model construction. Keeping enhanced samples whose spectral difference and content difference were all zero was useful for improving model performance. Augmentation of all data or training data based on maximum or minimum values-based spectral subtraction, which sorted the contents and made them the maximum or minimum values in sequence, achieved sample expansion by the spectral difference and content difference. The model utilized the random vector functional link (RVFL) network, extreme learning machine (ELM), and one-dimensional convolutional neural network (1D CNN), which could predict the content of new samples through ensemble averaging when predicting content difference. The experimental result showed the model of the spectral subtraction based on maximum or minimum values had a similar performance to that of the original samples. Augmentation of all data improved model performance by only RVFL and ELM. Augmentation of training data verified 1D CNN was better than RVFL and ELM. The paper implements a new data augmentation method and applies CNN to original samples with inadequate numbers, which lays the foundation for an improved model and applying spectral preprocessing. [ABSTRACT FROM AUTHOR]

[Display omitted] • A method for rapid coal proximate analysis is proposed. • The method combines spectroscopy with deep learning. • A deep learning model named DR_TELM is proposed, which includes dilated convolution, multi-level residual connection and TELM. • The method proposed achieves high precision and fast speed. Proximate analysis of coal is of profound significance for understanding coal quality and promoting rational utilization of coal resources. Traditional coal proximate analysis mainly uses chemical analysis methods, which have the disadvantages of slow speed and high cost. This paper proposed an approach combining reflectance spectroscopy with deep learning (DL) for rapid proximate analysis of coal. First, 80 sets of coal spectral data are enhanced by data augmentation, outlier detection, and dimensional transformation to improve the number and quality of samples. Then, an analytical model combining dilated convolution, multi-level residual connection, and a two-hidden-layer extreme learning machine (TELM), named DR_TELM, was proposed. The model extracted effective features from coal spectral data by a convolutional neural network (CNN) and utilized TELM as a regressor to achieve feature identification and content prediction. The experimental results showed that DR_TELM achieved coefficients of determination (R2) of 0.981, 0.989, 0.990, 0.985, 0.989 and root mean square errors (RMSE) of 0.533, 1.833, 1.111, 1.808, 0.723 for the content prediction of moisture, ash, volatile matter, fixed carbon and higher heating value (HHV), respectively. And while ensuring high accuracy, the test time is only 0.034 s. It is fully demonstrated that DR_TELM can rapidly and accurately analyze coal. [ABSTRACT FROM AUTHOR]

Multi-view learning aims to obtain more comprehensive understanding than single-view learning by observing objects from different views. However, most existing multi-view learning algorithms are still facing problems in obtaining enough discriminative information from the multi-view data: (1) most models cannot fully exploit consistent and complementary information simultaneously; (2) existing group sparsity based multi-view learning methods cannot extract the most relevant and sparest features. This paper proposes the efficient group non-convex sparsity regularized partially shared dictionary learning for multi-view learning, which employs the partially shared dictionary learning model to excavate both consistency and complementarity simultaneously from the multi-view data, and utilizes the generalized group non-convex sparsity for more discriminative and sparser representations beyond the convex ℓ 2 , 1 norm. To solve the non-convex optimization problem, we derive the generalized optimization framework for different group non-convex sparsity regularizers based on the proximal splitting method. Corresponding proximal operators for structured sparse coding in the framework are derived to form algorithms for different group non-convex sparsity regularizers, i.e., the ℓ 2 , p (0 < p < 1) norm and the ℓ 2 , l o g regularizer. In experiments, we conduct multi-view clustering in seven real-world multi-view datasets, and performances validate the effectiveness of both group information and non-convexity. Furthermore, results show that appropriate coefficient sharing ratios can help to exploit consistent information while keeping complementary information from multi-view data, thus helping to improve clustering performances. In addition, the convergence performances show that the proposed algorithm can obtain the best clustering performances among compared algorithms and can converge efficiently and stably with reasonable running time costs. [ABSTRACT FROM AUTHOR]