Representative Algorithms Research Articles

Efficient Sparse Representation for Learning With High-Dimensional Data.

Due to the capability of effectively learning intrinsic structures from high-dimensional data, techniques based on sparse representation have begun to display an impressive impact on several fields, such as image processing, computer vision, and pattern recognition. Learning sparse representations isoften computationally expensive due to the iterative computations needed to solve convex optimization problems in which the number of iterations is unknown before convergence. Moreover, most sparse representation algorithms focus only on determining the final sparse representation results and ignore the changes in the sparsity ratio (SR) during iterative computations. In this article, two algorithms are proposed to learn sparse representations based on locality-constrained linear representation learning with probabilistic simplex constraints. Specifically, the first algorithm, called approximated local linear representation (ALLR), obtains a closed-form solution from individual locality-constrained sparse representations. The second algorithm, called ALLR with symmetric constraints (ALLRSC), further obtains a symmetric sparse representation result with a limited number of computations; notably, the sparsity and convergence of sparse representations can be guaranteed based on theoretical analysis. The steady decline in the SR during iterative computations is a critical factor in practical applications. Experimental results based on public datasets demonstrate that the proposed algorithms perform better than several state-of-the-art algorithms for learning with high-dimensional data.

Kernel recursive least squares dictionary learning algorithm

An online dictionary learning algorithm for kernel sparse representation is developed in the current paper. In this framework, the input signal nonlinearly mapped into the feature space is sparsely represented based on a virtual dictionary in the same space. At any instant, the dictionary is updated in two steps. In the first step, the input signal samples are sparsely represented in the feature space, using the dictionary that has been updated based on the previous data. In the second step, the dictionary is updated. In this paper, a novel recursive dictionary update algorithm is derived, based on the recursive least squares (RLS) approach. This algorithm gradually updates the dictionary, upon receiving one or a mini-batch of training samples. An efficient implementation of the algorithm is also formulated. Experimental results over four datasets in different fields show the superior performance of the proposed algorithm in comparison with its counterparts. In particular, the classification accuracy obtained by the dictionaries trained using the proposed algorithm gradually approaches that of the dictionaries trained in batch mode. Moreover, in spite of lower computational complexity, the proposed algorithm overdoes all existing online kernel dictionary learning algorithms.

Automatic graph representation algorithm for heterogeneous catalysis

One of the most appealing aspects of machine learning for material design is its high throughput exploration of chemical spaces, but to reach the ceiling of machine learning-aided exploration, more than current model architectures and processing algorithms are required. New architectures such as graph neural networks have seen significant research investments recently. For heterogeneous catalysis, defining substrate intramolecular bonds and adsorbate/substrate intermolecular bonds is a time-consuming and challenging process. Before applying a model, dataset pre-processing, node/bond descriptor design, and specific model constraints have to be considered. In this work, a framework designed to solve these issues is presented in the form of an automatic graph representation algorithm (AGRA) tool to extract the local chemical environment of metallic surface adsorption sites. This tool is able to gather multiple adsorption geometry datasets composed of different systems and combine them into a single model. To show AGRA’s excellent transferability and reduced computational cost compared to other graph representation methods, it was applied to five different catalytic reaction datasets and benchmarked against the Open Catalyst Projects graph representation method. The two oxygen reduction reaction (ORR) datasets with O/OH adsorbates obtained 0.053 eV root-mean-square deviation (RMSD) when combined together, whereas the three carbon dioxide reduction reaction datasets with CHO/CO/COOH obtained an average performance of 0.088 eV RMSD. To further display the algorithm’s versatility and extrapolation ability, a model was trained on a subset combination of all five datasets with an RMSD of 0.105 eV. This universal model was then used to predict a wide range of adsorption energies and an entirely new ORR catalyst system, which was then verified through density functional theory calculations.

Deterministic and Probabilistic Approaches to Model and Update Dynamic Systems

Updating simulation models, based on flight test or other data, or creating a model using only experimental data is a necessary process that can be costly and laborious. Simulation updates are necessary for high-fidelity simulations used in analysis, control development, handling qualities prediction, fleet training, etc. This paper presents methods to update aircraft models based on deterministic and probabilistic approaches. The deterministic approach is based on a collection of signal processing, statistical analysis, and maximum likelihood estimators embodied in the Algorithms to Update Simulation Parameters with Experimental Data tool. Mathematically sound and statistically rigorous techniques are used to calculate updates of current model parameters and suggest new terms capturing unmodeled dynamics, thereby improving correlation with observed vehicle response. The probabilistic approach is based on generalized polynomial chaos theory included in the Algorithms for Uncertainty Representation and Analysis (AURA) tool. The AURA tool allows the user to update probabilistic models based on experimental data, formulated as a Bayesian inference problem. This paper will present the methodology for both deterministic and probabilistic approaches and provide examples of their application to different aircraft designs, including tilt-rotor and rotary-wing unmanned air system vehicles. The results for both deterministic and probabilistic modeling approaches demonstrate excellent results. The probabilistic approach is novel in its expressive power to propagate and update probabilistic models and provides users with an intuitive and powerful modeling approach.

CORDIC KSVD based Online Dictionary Learning for Speech Enhancement on ASIC/FPGA Platforms

Background: The enhancement of real-world speech signals is still a challenging task to eliminate noises, namely reverberation, background, street, and babble noises. Recently learned methods like dictionary learning have become increasingly popular and showed promising results in speech enhancement. The K-means Singular Value Decomposition (KSVD) algorithm is best suited for dictionary learning among many sparse representation algorithms. Moreover, the orthogonal matching pursuit (OMP) based algorithm used for signal recovery is given. The orthogonal matching pursuit (OMP) based algorithm for signal recovery gives the best enhancement results. On the other hand, FPGAs and ASICs are widely used to accelerate speech enhancement applications. FPGAs are commonly used in healthcare and consumer applications, where speech enhancement plays a crucial role. Methods: This paper proposes a modified KSVD algorithm that can easily be implemented onto hardware platforms like FPGAs and ASICS. Instead of using the double-precision arithmetic for the singular value decomposition part of the KSVD algorithm, we proposed to use CORDIC (Coordinate Rotation Digital Computer) based QR decomposition and QR-based singular value decomposition in dictionary learning. Results: The proposed KSVD algorithm is optimal with the CORDIC algorithm that can reduce by 7-8 times the processing time. Conclusion: The finding indicates that the proposed work is best suited to FPGA or ASIC platforms.

Sparse representation scheme with enhanced medium pixel intensity for face recognition

AbstractSparse representation is an effective data classification algorithm that depends on the known training samples to categorise the test sample. It has been widely used in various image classification tasks. Sparseness in sparse representation means that only a few of instances selected from all training samples can effectively convey the essential class‐specific information of the test sample, which is very important for classification. For deformable images such as human faces, pixels at the same location of different images of the same subject usually have different intensities. Therefore, extracting features and correctly classifying such deformable objects is very hard. Moreover, the lighting, attitude and occlusion cause more difficulty. Considering the problems and challenges listed above, a novel image representation and classification algorithm is proposed. First, the authors’ algorithm generates virtual samples by a non‐linear variation method. This method can effectively extract the low‐frequency information of space‐domain features of the original image, which is very useful for representing deformable objects. The combination of the original and virtual samples is more beneficial to improve the classification performance and robustness of the algorithm. Thereby, the authors’ algorithm calculates the expression coefficients of the original and virtual samples separately using the sparse representation principle and obtains the final score by a designed efficient score fusion scheme. The weighting coefficients in the score fusion scheme are set entirely automatically. Finally, the algorithm classifies the samples based on the final scores. The experimental results show that our method performs better classification than conventional sparse representation algorithms.

A High-Order Proximity-Incorporated Nonnegative Matrix Factorization-Based Community Detector

Community describes the functional mechanism of an undirected network, making community detection a fundamental tool for graph representation learning-related applications like social circle discovery. To date, a Symmetric and Nonnegative Matrix Factorization (SNMF) model has been frequently adopted to address this issue owing to its high interpretability and scalability. However, most existing SNMF-based community detectors neglect the high-order proximity in an undirected network, thus suffering from accuracy loss caused by incomplete information. Motivated by this discovery, this paper proposes a <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">H</u> igh- <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">O</u> rder <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">P</u> roximity-incorporated <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">N</u> onnegative <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">M</u> atrix <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F</u> actorization (HOP-NMF)-based community detector with the following three-fold ideas: a) adopting a weighted pointwise mutual information-based approach to measure the high-order proximity among nodes in a network; b) leveraging an iterative network enhancement scheme to encode the captured high-order proximity into the network to effectively enhance the its information; and c) implementing a capacity-enlarged and graph-regularized factorization algorithm for highly-accurate representation to the enhanced network. With the above design, an HOP-NMF model is able to achieve highly-accurate community detection results. Theoretical proof is rigorously conducted to validate its convergence ability. Extensively empirical studies on eight real networks from industrial applications demonstrate that an HOP-NMF-based community detector significantly outperforms sophisticated and state-of-the-art community detectors in detection accuracy.

An intervention study of VR sports games on the perceptual and motor abilities of children with incomplete intellectual development

Abstract Childhood is a critical stage for the development of perceptual and motor abilities, and strengthening the training of children with incomplete mental development at this stage will affect the development of motor skills during their growth. In this paper, we take VR technology as a starting point to build a perceptual model and introduce a convolutional sparse representation algorithm. First, a sparse representation with few non-zero elements is found to optimize a function consisting of a data fidelity term and a sparse induced penalty function. Then, the sum of the convolution of the filter and the convolution sparse feature map, i.e., the convolution operation, is computed to generate the translation invariants. Then the convolutional sparse coding method is introduced to the traditional unsupervised problem by calculating the minimization objective function and solving it in an iterative manner alternatively. Finally, the constituents of the signal are analyzed and the discrete equivalence of the convolution is derived based on the Fourier transform to derive the intervening variables. The experimental results showed that the mean value of the post-test of motor ability compared with the pre-test increased by 4.6 through an eight-week VR sports game training intervention study test on different children with incomplete mental development. Therefore, it is of great theoretical and practical significance to understand the characteristics of perceptual and motor abilities of children with incomplete intellectual development and to develop corresponding programs for VR sports game training according to their characteristics.

A survey on geometric shape representation of objects based on medial axis transform

&lt;div&gt;&lt;p&gt;Geometric shape representation algorithms are key technologies in the fields of computer graphics and geometric modeling. The Medial Axis Transform (MAT) is an important geometric model description tool that provides a simplified representation of complex geometric shapes while ensuring accurate descriptions of geometric shape and topology. Therefore, it can meet the requirements of many modern research fields, including geometric modeling, pattern recognition, model segmentation, model deformation, physical simulation, path planning, and more. This paper first introduces the basic concept of the medial axis transform, including the definition of the medial axis transform and the concept of medial axis primitives. It then describes the extraction algorithms for the medial axis transform, specific research on the medial axis transform in computer vision and computer graphics, potential applications of the medial axis transform, and medial axis transform datasets. Finally, the disadvantages and advantages of the medial axis transform are discussed, and some suggestions on possible future research directions are presented.&lt;/p&gt;&lt;/div&gt;

Sparse representation of acoustic emission signals and its application in pipeline leak location

The acoustic detection of pipeline leaks is strongly influenced by the associated interference and noise. In this paper, acoustic emission signals are analyzed using a sparse representation method and the main components associated with the leakage are extracted. Dictionary learning is performed using training samples composed of the leakage signals and noise signals. The measured signal is sparsely decomposed on the composite dictionary, allowing the main leakage components to be estimated. The effects of dictionary dimensionality, redundancy, sparsity constraints, and other parameters on the performance of the sparse representation algorithm are investigated. Finally, the leak location in the pipeline is determined through cross-correlation analysis of the reconstructed acoustic signals. Experimental results show that the proposed sparse representation method effectively improves the signal-to-noise ratio of the acoustic emission signals, and correspondingly improves the accuracy and reliability of pipeline leak location compared with traditional localization methods.

Chromatic Differentiation of Functional Mappings of the Composition of Nucleic Acids

Color visualization of the DNA of diverse living beings can help in the exploration of the issue of chromatic differentiation of functional mappings of the nucleotide composition of DNA molecules. By “chromatic differentiation”, we mean the coloring of these mappings. Algorithms for coloring genetic representations improve the perception of complex genetic information using color. Methodologically, to build the chromatic differentiation of functional mappings of the nucleotide composition of DNA, we employed the system of nucleotide Walsh functions and the Chaos Game Representation (CGR) algorithm. The authors compared these two approaches and proposed a modified CGR algorithm. The work presents various algorithms of chromatic differentiation based on the nucleotide Walsh functions at a specific location of the fragment in the nucleotide chain and on the frequencies of those fragments. The results of the analysis provide examples of chromatic differentiation in a variety of parametric spaces. The paper describes various approaches to coloring and video animation of DNA molecules in their chromatically differentiated spans of physicochemical parameters.

Cooperative Decision-Making for Mixed Traffic at an Unsignalized Intersection Based on Multi-Agent Reinforcement Learning

Despite rapid advances in vehicle intelligence and connectivity, there is still a significant period in mixed traffic where connected, automated vehicles and human-driven vehicles coexist. The behavioral uncertainty of human-driven vehicles makes decision-making a challenging task in an unsignalized intersection scenario. In this paper, a decentralized multi-agent proximal policy optimization (MAPPO) based on an attention representations algorithm (Attn-MAPPO) was developed to make joint decisions at an intersection to avoid collisions and cross the intersection effectively. To implement this framework, by exploiting the shared information, the system was modeled as a model-free, fully cooperative, multi-agent system. The vehicle employed an attention module to extract the most valuable information from its neighbors. Based on the observation and traffic rules, a joint policy was identified to work more cooperatively based on the trajectory prediction of all the vehicles. To facilitate the collaboration between the vehicles, a weighted reward assignment scheme was proposed to focus more on the vehicles approaching intersections. The results presented the advantages of the Attn-MAPPO framework and validated the effectiveness of the designed reward function. Ultimately, the comparative experiments were conducted to demonstrate that the proposed approach was more adaptive and generalized than the heuristic rule-based model, which revealed its great potential for reinforcement learning in the decision-making of autonomous driving.

Incipient detection of bearing fault using impulse feature enhanced weighted sparse representation

The bearing fault impact impulses induced by the contact between components with drawback, is difficult to be detected at sprouting stage due to the interference of background noise, harmonics, random shocks, etc. In this paper, an impulse feature enhanced weighted sparse representation (IFEWSR) algorithm is proposed to accurately detect the weak bearing fault impact feature from incipient stage condition monitoring (CM) signal. Firstly, a modified fault period estimation method is presented to improve the robustness and reduce the computational complexity of recently proposed algorithms. Secondly, a novel weighting strategy on wavelet coefficients, indicated by the period-assisted corelated kurtosis of envelope spectrum (CKSES), is presented to denote the contribution of subband signals for sparse representation calculation framework. In addition, the mean normalized energy-weight deviation (MNEWD) rule is proposed to evaluate the performance of the weighting algorithm on subband signals which is blank at present. Thirdly, a novel fault feature enhancement technique is developed to better capture the bearing fault feature information. The effectiveness and superiority of the proposed method are proved by simulation and experiments. Results show that the proposed IFEWSR method provides higher accuracy for incipient fault feature extraction and outperforms other state-of-the-art methods.

Variational Graph Autoencoder with Adversarial Mutual Information Learning for Network Representation Learning

With the success of Graph Neural Network (GNN) in network data, some GNN-based representation learning methods for networks have emerged recently. Variational Graph Autoencoder (VGAE) is a basic GNN framework for network representation. Its purpose is to well preserve the topology and node attribute information of the network to learn node representation, but it only reconstructs network topology, and does not consider the reconstruction of node features. This strategy will make node representation can not well reserve node features information, impairing the ability of the VGAE method to learn higher quality representations. To solve this problem, we arise a new network representation method to improve the VGAE method for well retaining both node features and network structure information. The method utilizes adversarial mutual information learning to maximize the mutual information (MI) of node features and node representations during the encoding process of the variational autoencoder, which forces the variational encoder to get the representation containing the most informative node features. The method consists of three parts: a variational graph autoencoder includes a variational encoder (MI generator (G)) and a decoder, a positive MI sample module (maximizing MI module), and an MI discriminator (D). Furthermore, we explain why maximizing MI between node features and node representation can reconstruct node attributes. Finally, we conduct experiments on seven public representative datasets for nodes classification, nodes clustering, and graph visualization tasks. Experimental results demonstrate that the proposed algorithm significantly outperforms current popular network representation algorithms on these tasks. The best improvement is 17.13% than the VGAE method.

Learning Distributed Representations and Deep Embedded Clustering of Texts

Instructors face significant time and effort constraints when grading students’ assessments on a large scale. Clustering similar assessments is a unique and effective technique that has the potential to significantly reduce the workload of instructors in online and large-scale learning environments. By grouping together similar assessments, marking one assessment in a cluster can be scaled to other similar assessments, allowing for a more efficient and streamlined grading process. To address this issue, this paper focuses on text assessments and proposes a method for reducing the workload of instructors by clustering similar assessments. The proposed method involves the use of distributed representation to transform texts into vectors, and contrastive learning to improve the representation that distinguishes the differences among similar texts. The paper presents a general framework for clustering similar texts that includes label representation, K-means, and self-organization map algorithms, with the objective of improving clustering performance using Accuracy (ACC) and Normalized Mutual Information (NMI) metrics. The proposed framework is evaluated experimentally using two real datasets. The results show that self-organization maps and K-means algorithms with Pre-trained language models outperform label representation algorithms for different datasets.

ReBADD-SE: Multi-objective molecular optimisation using SELFIES fragment and off-policy self-critical sequence training

The discovery of drugs to selectively remove disease-related cells is challenging in computer-aided drug design. Many studies have proposed multi-objective molecular generation methods and demonstrated their superiority using the public benchmark dataset for kinase inhibitor generation tasks. However, the dataset does not contain many molecules that violate Lipinski’s rule of five. Thus, it remains unclear whether existing methods are effective in generating molecules violating the rule, such as navitoclax. To address this, we analysed the limitations of existing methods and propose a multi-objective molecular generation method with a novel parsing algorithm for molecular string representation and a modified reinforcement learning method for the efficient training of multi-objective molecular optimisation. The proposed model had success rates of 84% in GSK3b+JNK3 inhibitor generation and 99% in Bcl-2 family inhibitor generation tasks.

Graph Representation Learning Meets Computer Vision: A Survey

A graph structure is a powerful mathematical abstraction, which can not only represent information about individuals but also capture the interactions between individuals for reasoning. Geometric modeling and relational inference based on graph data is a long-standing topic of interest in the computer vision community. In this article, we provide a systematic review of graph representation learning and its applications in computer vision. First, we sort out the evolution of representation learning on graphs, categorizing them into the nonneural network and neural network methods based on the way the nodes are encoded. Specifically, nonneural network methods, such as graph embedding and probabilistic graphical models, are introduced, and neural network methods, such as graph recurrent neural networks, graph convolutional networks, and variants of graph neural networks, are also presented. Then, we organize the applications of graph representation algorithms in various vision tasks (such as image classification, semantic segmentation, object detection, and tracking) for review and reference, and the typical graph construction approaches in computer vision are also summarized. Finally, on the background of biology and brain inspiration, we discuss the existing challenges and future directions of graph representation learning and computer vision.

PDA-PRGCN: identification of Piwi-interacting RNA-disease associations through subgraph projection and residual scaling-based feature augmentation

BackgroundEmerging evidences show that Piwi-interacting RNAs (piRNAs) play a pivotal role in numerous complex human diseases. Identifying potential piRNA-disease associations (PDAs) is crucial for understanding disease pathogenesis at molecular level. Compared to the biological wet experiments, the computational methods provide a cost-effective strategy. However, few computational methods have been developed so far.ResultsHere, we proposed an end-to-end model, referred to as PDA-PRGCN (PDA prediction using subgraph Projection and Residual scaling-based feature augmentation through Graph Convolutional Network). Specifically, starting with the known piRNA-disease associations represented as a graph, we applied subgraph projection to construct piRNA-piRNA and disease-disease subgraphs for the first time, followed by a residual scaling-based feature augmentation algorithm for node initial representation. Then, we adopted graph convolutional network (GCN) to learn and identify potential PDAs as a link prediction task on the constructed heterogeneous graph. Comprehensive experiments, including the performance comparison of individual components in PDA-PRGCN, indicated the significant improvement of integrating subgraph projection, node feature augmentation and dual-loss mechanism into GCN for PDA prediction. Compared with state-of-the-art approaches, PDA-PRGCN gave more accurate and robust predictions. Finally, the case studies further corroborated that PDA-PRGCN can reliably detect PDAs.ConclusionPDA-PRGCN provides a powerful method for PDA prediction, which can also serve as a screening tool for studies of complex diseases.

AN ALGEBRAIC DECOMPOSITION OF HEMITROPIC PSEUDOTENSORS IN N DIMENSIONS AND APPLICATIONS TO MICROPOLAR CONTINUUM THEORIES

The paper is devoted to problems related to algebraic decompositions and coordinate representations of tensors with constant components, hemitropic tensors and pseudotensors. Such objects of tensor algebra are interesting from the micropolar continuum mechanics viewpoint. Algebraic properties and coordinate representations of tensors and pseudotensors with constant components are discussed. The base examples of tensors with constant components usable in continuum mechanics are given. An algebraic algorithm for coordinate representations of tensors and pseudotensors with constant components proposed by prof. G.B. Gurevich is highlighted and employed.The notions of fully isotropic, conventionally isotropic, unconventionally isotropic, semi-isotropic (demitropic, hemitropic, chiral) fourth rank tensors and pseudotensors are proposed. The coordinate representations of a Cartesian hemitropic fourth rank tensor in three-dimensional space are obtained in terms of the Kronecker delta and metric tensor products. Based on an unconventional definition of a hemitropic fourth rank tensor, general coordinate representations in dimensions in terms of the Kronecker deltas and metric tensors are given. A comparison of an arbitrary hemitropic fourth rank tensor and a tensor with constant components is carried out. A general form of the elastic potential of a linear anisotropic micropolar elastic continuum is obtained by the pseudotensor technique. Obtained anisotropic micropolar potential is reduced to a hemitropic one by proposed coordinate representations of fourth rank tensors. Coordinate representations for constitutive tensors and pseudotensors usable in mathematical modeling of linear hemitropic micropolarelastic continua are obtained as a modification of pseudotensors with constant components in three-dimensional space.

PhiKitA: Phishing Kit Attacks Dataset for Phishing Websites Identification

Recent studies have shown that phishers are using phishing kits to deploy phishing attacks faster, easier and more massive. Detecting phishing kits in deployed websites might help to detect phishing campaigns earlier. To the best of our knowledge, there are no datasets providing a set of phishing kits that are used in websites that were attacked by phishing. In this work, we propose PhiKitA, a novel dataset that contains phishing kits and also phishing websites generated using these kits. We have applied MD5 hashes, fingerprints, and graph representation DOM algorithms to obtain baseline results in PhiKitA in three experiments: familiarity analysis of phishing kit samples, phishing website detection and identifying the source of a phishing website. In the familiarity analysis, we find evidence of different types of phishing kits and a small phishing campaign. In the binary classification problem for phishing detection, the graph representation algorithm achieved an accuracy of 92.50%, showing that the phishing kit data contain useful information to classify phishing. Finally, the MD5 hash representation obtained a 39.54% F1 score, which means that this algorithm does not extract enough information to distinguish phishing websites and their phishing kit sources properly.