A Novel Tensor Decomposition-Based Efficient Detector for Low-Altitude Aerial Objects With Knowledge Distillation Scheme

Unmanned aerial vehicles(UAVs) have gained significant attention in practical applications, especially the low-altitude aerial(LAA) object detection imposes stringent requirements on recognition accuracy and computational resources. In this paper, the LAA images-oriented tensor decomposition and knowledge distillation-based network(TDKD-Net) is proposed,where the TT-format TD(tensor decomposition) and equalweighted response-based KD(knowledge distillation) methods are designed to minimize redundant parameters while ensuring comparable performance. Moreover, some robust network structures are developed, including the small object detection head and the dual-domain attention mechanism, which enable the model to leverage the learned knowledge from small-scale targets and selectively focus on salient features. Considering the imbalance of bounding box regression samples and the inaccuracy of regression geometric factors, the focal and efficient IoU(intersection of union) loss with optimal transport assignment(F-EIoU-OTA)mechanism is proposed to improve the detection accuracy. The proposed TDKD-Net is comprehensively evaluated through extensive experiments, and the results have demonstrated the effectiveness and superiority of the developed methods in comparison to other advanced detection algorithms, which also present high generalization and strong robustness. As a resource-efficient precise network, the complex detection of small and occluded LAA objects is also well addressed by TDKD-Net, which provides useful insights on handling imbalanced issues and realizing domain adaptation.

Block Incremental Dense Tucker Decomposition with Application to Spatial and Temporal Analysis of Air Quality Data

How can we efficiently store and mine dynamically generated dense tensors for modeling the behavior of multidimensional dynamic data?Much of the multidimensional dynamic data in the real world is generated in the form of time-growing tensors.For example,air quality tensor data consists of multiple sensory values gathered from wide locations for a long time.Such data,accumulated over time,is redundant and consumes a lot ofmemory in its raw form.We need a way to efficiently store dynamically generated tensor data that increase over time and to model their behavior on demand between arbitrary time blocks.To this end,we propose a Block IncrementalDense Tucker Decomposition(BID-Tucker)method for efficient storage and on-demand modeling ofmultidimensional spatiotemporal data.Assuming that tensors come in unit blocks where only the time domain changes,our proposed BID-Tucker first slices the blocks into matrices and decomposes them via singular value decomposition(SVD).The SVDs of the time×space sliced matrices are stored instead of the raw tensor blocks to save space.When modeling from data is required at particular time blocks,the SVDs of corresponding time blocks are retrieved and incremented to be used for Tucker decomposition.The factor matrices and core tensor of the decomposed results can then be used for further data analysis.We compared our proposed BID-Tucker with D-Tucker,which our method extends,and vanilla Tucker decomposition.We show that our BID-Tucker is faster than both D-Tucker and vanilla Tucker decomposition and uses less memory for storage with a comparable reconstruction error.We applied our proposed BID-Tucker to model the spatial and temporal trends of air quality data collected in South Korea from 2018 to 2022.We were able to model the spatial and temporal air quality trends.We were also able to verify unusual events,such as chronic ozone alerts and large fire events.

Schur Forms and Normal-Nilpotent Decompositions

Real and complex Schur forms have been receiving increasing attention from the fluid mechanics community recently,especially related to vortices and turbulence.Several decompositions of the velocity gradient tensor,such as the triple decomposition of motion(TDM)and normal-nilpotent decomposition(NND),have been proposed to analyze the local motions of fluid elements.However,due to the existence of different types and non-uniqueness of Schur forms,as well as various possible definitions of NNDs,confusion has spread widely and is harming the research.This work aims to clean up this confusion.To this end,the complex and real Schur forms are derived constructively from the very basics,with special consideration for their non-uniqueness.Conditions of uniqueness are proposed.After a general discussion of normality and nilpotency,a complex NND and several real NNDs as well as normal-nonnormal decompositions are constructed,with a brief comparison of complex and real decompositions.Based on that,several confusing points are clarified,such as the distinction between NND and TDM,and the intrinsic gap between complex and real NNDs.Besides,the author proposes to extend the real block Schur form and its corresponding NNDs for the complex eigenvalue case to the real eigenvalue case.But their justification is left to further investigations.

An Efficient Randomized Fixed-Precision Algorithm for Tensor Singular Value Decomposition

The existing randomized algorithms need an initial estimation of the tubal rank to compute a tensor singular value decomposition.This paper proposes a new randomized fixed-precision algorithm which for a given third-order tensor and a prescribed approximation error bound,it automatically finds the tubal rank and corresponding low tubal rank approximation.The algorithm is based on the random projection technique and equipped with the power iteration method for achieving better accuracy.We conduct simulations on synthetic and real-world datasets to show the efficiency and performance of the proposed algorithm.

Research on infrared dim and small target detection algorithm based on low-rank tensor recovery

In order to rapidly and accurately detect infrared small and dim targets in the infrared image of complex scene collected by virtual prototyping of space-based downward-looking multiband detection,an improved detection algorithm of infrared small and dim target is proposed in this paper.Firstly,the original infrared images are changed into a new infrared patch tensor mode through data reconstruction.Then,the infrared small and dim target detection problems are converted to low-rank tensor recovery problems based on tensor nuclear norm in accordance with patch tensor characteristics,and inverse variance weighted entropy is defined for self-adaptive adjustment of sparseness.Finally,the low-rank tensor recovery problem with noise is solved by alternating the direction method to obtain the sparse target image,and the final small target is worked out by a simple partitioning algorithm.The test results in various spacebased downward-looking complex scenes show that such method can restrain complex background well by virtue of rapid arithmetic speed with high detection probability and low false alarm rate.It is a kind of infrared small and dim target detection method with good performance.

A Novel Multichannel Audio Signal Compression Method Based on Tensor Representation and Decomposition

Multichannel audio signal is more difficult to be compressed than mono and stereo ones.A novel multichannel audio signal compression method based on tensor representation and decomposition is proposed in this paper.The multichannel audio is represented with 3-order tensor space and is decomposed into core tensor with three factor matrices in the way of channel,time and frequency.Only the truncated core tensor is transmitted which will be multiplied by the pre-trained factor matrices to reconstruct the original tensor space.Objective and subjective experiments have been done to show a very noticeable compression capability with an acceptable output quality.The novelty of the proposed compression method is that it enables both high compression capability and backward compatibility with limited signal distortion to the hearing.

A multispectral image compression and encryption algorithm based on tensor decomposition and chaos

A multi spectral image compression and encryption algorithm that combines Karhunen-Loeve(KL) transform,tensor decomposition and chaos is proposed for solving the security problem of multi-spectral image compression and transmission.Firstly,in order to eliminate residual spatial redundancy and most of the spectral redundancy,the image is performed by KL transform.Secondly,to further eliminate spatial redundancy and reduce block effects in the compression process,two-dimensional discrete 9/7 wavelet transform is performed,and then Arnold transform and encryption processing on the transformed coefficients are performed.Subsequently,the tensor is decomposed to keep its intrinsic structure intact and eliminate residual space redundancy.Finally,differential pulse filters are used to encode the coefficients,and Tent mapping is used to implement confusion diffusion encryption on the code stream.The experimental results show that the method has high signal-to-noise ratio,fast calculation speed,and large key space,and it is sensitive to keys and plaintexts with a positive effect in spectrum assurance at the same time.

Key Exchange Protocol Based on Tensor Decomposition Problem

The hardness of tensor decomposition problem has many achievements, but limited applications in cryptography, and the tensor decomposition problem has been considered to have the potential to resist quantum computing. In this paper, we firstly proposed a new variant of tensor decomposition problem, then two one-way functions are proposed based on the hard problem. Secondly we propose a key exchange protocol based on the one-way functions, then the security analysis, efficiency, recommended parameters and etc. are also given. The analyses show that our scheme has the following characteristics: easy to implement in software and hardware, security can be reduced to hard problems, and it has the potential to resist quantum computing.Besides the new key exchange can be as an alternative comparing with other classical key protocols.

Relations between cubic equation, stress tensor decomposition, and von Mises yield criterion

Inspired by Cardano＇s method for solving cubic scalar equations, the addi- tive decomposition of spherical/deviatoric tensor （DSDT） is revisited from a new view- point. This decomposition simplifies the cubic tensor equation, decouples the spher- ical/deviatoric strain energy density, and lays the foundation for the von Mises yield criterion. Besides, it is verified that under the precondition of energy decoupling and the simplest form, the DSDT is the only possible form of the additive decomposition with physical meanings.

Multi-Aspect Incremental Tensor Decomposition Based on Distributed In-Memory Big Data Systems

Purpose:We propose In Par Ten2,a multi-aspect parallel factor analysis three-dimensional tensor decomposition algorithm based on the Apache Spark framework.The proposed method reduces re-decomposition cost and can handle large tensors.Design/methodology/approach:Considering that tensor addition increases the size of a given tensor along all axes,the proposed method decomposes incoming tensors using existing decomposition results without generating sub-tensors.Additionally,In Par Ten2 avoids the calculation of Khari–Rao products and minimizes shuffling by using the Apache Spark platform.Findings:The performance of In Par Ten2 is evaluated by comparing its execution time and accuracy with those of existing distributed tensor decomposition methods on various datasets.The results confirm that In Par Ten2 can process large tensors and reduce the re-calculation cost of tensor decomposition.Consequently,the proposed method is faster than existing tensor decomposition algorithms and can significantly reduce re-decomposition cost.Research limitations:There are several Hadoop-based distributed tensor decomposition algorithms as well as MATLAB-based decomposition methods.However,the former require longer iteration time,and therefore their execution time cannot be compared with that of Spark-based algorithms,whereas the latter run on a single machine,thus limiting their ability to handle large data.Practical implications:The proposed algorithm can reduce re-decomposition cost when tensors are added to a given tensor by decomposing them based on existing decomposition results without re-decomposing the entire tensor.Originality/value:The proposed method can handle large tensors and is fast within the limited-memory framework of Apache Spark.Moreover,In Par Ten2 can handle static as well as incremental tensor decomposition.

Recommender Systems Based on Tensor Decomposition

Recommender system is an effective tool to solve the problems of information overload.The traditional recommender systems,especially the collaborative filtering ones,only consider the two factors of users and items.While social networks contain abundant social information,such as tags,places and times.Researches show that the social information has a great impact on recommendation results.Tags not only describe the characteristics of items,but also reflect the interests and characteristics of users.Since the traditional recommender systems cannot parse multi-dimensional information,in this paper,a tensor decomposition model based on tag regularization is proposed which incorporates social information to benefit recommender systems.The original Singular Value Decomposition(SVD)model is optimized by mining the co-occurrence and mutual exclusion of tags,and their features are constrained by the relationship between tags.Experiments on real dataset show that the proposed algorithm achieves superior performance to existing algorithms.

Parallel Active Subspace Decomposition for Tensor Robust Principal Component Analysis

Tensor robust principal component analysis has received a substantial amount of attention in various fields.Most existing methods,normally relying on tensor nuclear norm minimization,need to pay an expensive computational cost due to multiple singular value decompositions at each iteration.To overcome the drawback,we propose a scalable and efficient method,named parallel active subspace decomposition,which divides the unfolding along each mode of the tensor into a columnwise orthonormal matrix(active subspace)and another small-size matrix in parallel.Such a transformation leads to a nonconvex optimization problem in which the scale of nuclear norm minimization is generally much smaller than that in the original problem.We solve the optimization problem by an alternating direction method of multipliers and show that the iterates can be convergent within the given stopping criterion and the convergent solution is close to the global optimum solution within the prescribed bound.Experimental results are given to demonstrate that the performance of the proposed model is better than the state-of-the-art methods.

TdBrnn:An Approach to Learning Users’Intention to Legal Consultation with Normalized Tensor Decomposition and Bi-LSTM

With the development of Internet technology and the enhancement of people’s concept of the rule of law,online legal consultation has become an important means for the general public to conduct legal consultation.However,different people have different language expressions and legal professional backgrounds.This phenomenon may lead to the phenomenon of different descriptions of the same legal consultation.How to accurately understand the true intentions behind different users’legal consulting statements is an important issue that needs to be solved urgently in the field of legal consulting services.Traditional intent understanding algorithms rely heavily on the lexical and semantic information between the original data,and are not scalable,and often require taxing manual annotation work.This article proposes a new approach TdBrnn which is based on the normalized tensor decomposition method and Bi-LSTM to learn users’intention to legal consulting.First,we present the users’legal consulting statements as a tensor.And then we use the normalized tensor decomposition layer proposed by this article to extract the tensor elements and structural information of the original tensor which can best represent users’intention of legal consultation,namely the core tensor.The core tensor relies less on the lexical and semantic information of the original users’legal consulting statements data,it reduces the dimension of the original tensor,and greatly reduces the computational complexity of the subsequent Bi-LSTM algorithm.Furthermore,we use a large number of core tensors obtained by the tensor decomposition layer with users’legal consulting statements tensors as inputs to continuously train Bi-LSTM,and finally derive the users’legal consultation intention classification model which can comprehensively understand the user’s legal consultation intention.Experiments show that our method has faster convergence speed and higher accuracy than traditional recurrent neural networks.

Efficient tensor decomposition method for noncircular source in colocated coprime MIMO radar

An effective method via tensor decomposition is proposed to deal with the joint direction-of-departure(DOD)and direction-of-arrival(DOA)estimation of noncircular sources in colocated coprime MIMO radar.By decomposing the transmitter and receiver into two sparse subarrays,noncircular property of source can be used to construct new extended received signal model for two sparse subarrays.The new received model can double the virtual array aperture due to the elliptic covariance of imping sources is nonzero.To further exploit the multidimensional structure of the noncircular received model,we stack the subarray output and its conjugation according to mode-1 unfolding and mode-2 unfolding of a third-order tensor,respectively.Thus,the corresponding extended tensor model consisted of noncircular information for DOA and DOD can be obtained.Then,the higher-order singular value decomposition technique is utilized to estimate the accurate signal subspace and angular parameter can be automatically paired via the rotational invariance relationship.Specifically,the ambiguous angle can be eliminated and the true targets can be achieved with the aid of the coprime property.Furthermore,a closed-form expression for the deterministic CRB under the NC sources scenario is also derived.Simulation results verify the superiority of the proposed estimator.

Detection of T-wave Alternans in ECG Signals Using FRFT and Tensor Decomposition

T-wave alternans(TWA)refers to the periodic beat-to-beat variation in the amplitude of T-wave in the electrocardiogram(ECG)signal in an ABAB-pattern.TWA has been proven to be a very important indicator of malignant arrhythmia risk stratification.A new method to detect TWA by combining fractional Fourier transform(FRFT)and tensor decomposition is proposed.First,the T-wave vector is extracted from the ECG of each heartbeat,and its FRFT amplitudes at multiple orders are arranged to form a T-wave matrix.Then,a third-order tensor is composed of T-wave matrices of several consecutive heart beats.After tensor decomposition,projection matrices are obtained in three dimensions.The complexity of the projection matrix is measured by Shannon entropy to obtain feature vector to detect the presence of TWA.Results show that the sensitivity,specificity,and accuracy of the algorithm on the MIT-BIH database are 91.16%,94.25%,and 92.68%,respectively.This method effectively utilizes the fractional domain information of ECG,and shows the promising potential of the FRFT in ECG signal processing.

Community Evolution Based on Tensor Decomposition

With the high dimensionality of data, the method of tensor decomposition has attracted much attention in the field of data research and analysis. The tensor decomposition is well reflected in the study of high-dimensional data. The existing research uses the results of tensor decomposition to conduct community discovery. Based on the existing research, this paper presents a method to study community evolution using the results of tensor decomposition. The feature matrix obtained by the tensor decomposition algorithm was analyzed, and the real-time activity of the community with the feature matrix with time slice direction was studied to obtain the event process of community evolution. Experimental results in real data sets show that this method can well analyze dynamic events in the dataset and community evolution events.

关于Focal Loss的Tensor Train多项式分类器

在模式分类领域,多项式分类器因其复杂决策边界能力而得到广泛研究。利用TensorTrain分解形式来表示多项式分类器,可有效克服维数灾难。针对多项式分类器在训练过程中遇到的训练集分布不平衡问题,文章使用FocalLoss重塑了标准交叉损失,以降低分配给易分类样本的损失的权重,并在被广泛使用的图像分类数据集MNIST上验证了分类器的有效性。

On Polar Decomposition of Tensors with Einstein Product and a Novel Iterative Parametric Method

This study aims to investigate the polar decomposition of tensors with the Einstein product for thefirst time.The polar decomposition of tensors can be computed using the singular value decomposition of the tensors with the Einstein product.In the following,some iterative methods forfinding the polar decomposi-tion of matrices have been developed into iterative methods to compute the polar decomposition of tensors.Then,we propose a novel parametric iterative method tofind the polar decomposition of tensors.Under the obtained conditions,we prove that the proposed parametric method has the order of convergence four.In every iteration of the proposed method,only four Einstein products are required,while other iterative methods need to calculate multiple Einstein products and one tensor inversion in each iteration.Thus,the new method is superior in terms of efficiency index.Finally,the numerical comparisons performed among several well-known methods,show that the proposed method is remarkably efficient and accurate.

Fast PARAFAC decomposition with application to polarization sensitive array parameter estimations

In tensor theory, the parallel factorization （PARAFAC）decomposition expresses a tensor as the sum of a set of rank-1tensors. By carrying out this numerical decomposition, mixedsources can be separated or unknown system parameters can beidentified, which is the so-called blind source separation or blindidentification. In this paper we propose a numerical PARAFACdecomposition algorithm. Compared to traditional algorithms, wespeed up the decomposition in several aspects, i.e., search di-rection by extrapolation, suboptimal step size by Gauss-Newtonapproximation, and linear search by n steps. The algorithm is ap-plied to polarization sensitive array parameter estimation to showits usefulness. Simulations verify the correctness and performanceof the proposed numerical techniques.

Randomized Generalized Singular Value Decomposition

The generalized singular value decomposition(GSVD)of two matrices with the same number of columns is a very useful tool in many practical applications.However,the GSVD may suffer from heavy computational time and memory requirement when the scale of the matrices is quite large.In this paper,we use random projections to capture the most of the action of the matrices and propose randomized algorithms for computing a low-rank approximation of the GSVD.Serval error bounds of the approximation are also presented for the proposed randomized algorithms.Finally,some experimental results show that the proposed randomized algorithms can achieve a good accuracy with less computational cost and storage requirement.