Original Vector Research Articles

Biological Image Processing Algorithm Based on Attention Mechanism and Convolutional Neural Network

This paper proposes a model algorithm based on convolutional neural network combined with attention mechanism to realize fast and accurate identification of biological image. Firstly, deformable convolution is used to extract features in the horizontal and vertical directions, respectively. Secondly, attention modules are used to capture remote dependencies in one spatial direction, while accurate position information is retained in another spatial direction, so that information in both vertical and horizontal directions can be retained; after a series of transformations, the attention vector is obtained and multiplied back to the original feature vector as a weight factor. The experimental results show that the proposed algorithm can effectively improve the image quality, improve the image clarity, avoid color distortion, and achieve good results in both synthetic and real low-illumination images, and the subjective and objective evaluation indicators are better than the contrast algorithm.

Исследование устойчивости к аномальным наблюдениям модификаций метода главных компонент

The paper considers the problem of reducing multidimensional correlated indicators. One of the approaches to solving this problem is based on the method of principal components, which makes it possible to compactly describe the vector with correlated coordinates (components) using the principal components vector with uncorrelated coordinates of much smaller dimension, while retaining most of the information about correlation structure of the original vector. On simulated and real data, several modifications of the principal components method were compared differing in the method of evaluating correlation matrix of the observation vector. The work objective is to demonstrate advantages of the robust modifications of the principal components method in cases, where data contained the abnormal values. To compare the considered modifications on the model data, metric was introduced that measured the difference between estimated and true eigenvalues of the initial data correlation matrix. This metric behavior depending on the probability distribution of observations was studied by computer simulation. As the distributions, multivariate distributions with the off-diagonal correlation matrices simulating a polluted sample were selected. Next, a sample of 13 correlated socioeconomic indicators for 85 countries was considered, where 46 abnormal values were identified. The considered modifications of the principal components method chose the same optimal number of principal components equal to three. However, the real data compression quality, which was defined as the share of the initial indicators total variance described by the first three principal components, turned out to be significantly higher for the robust modifications of the principal components method. Results obtained on these real data are in good agreement with conclusions of the computer simulation

An Improved In-Motion Coarse Alignment Method for SINS/GPS Integration with Initial Velocity Error Suppression.

The integrated system with the strapdown inertial navigation system (SINS) and the global positioning system (GPS) is the most popular navigation mode. It has been used in many navigation fields. Before the integrated system works properly, it must determine the initial attitude for SINS. In SINS/GPS-integrated systems, the navigational velocity can be used to carry out the initial alignment when the system is installed in the in-motion vehicle. However, the initial velocity errors are not considered in the current popular in-motion alignment methods for SINS/GPS integration. It is well-known that the initial velocity errors must exist when the initial velocity is obtained from the GPS outputs. In this paper, an improved method was proposed to solve this problem. By analyzing the original observation vectors in the in-motion coarse alignment method, an average operation was used to construct the intermediate vectors, and the new observation vector can be calculated by subtracting the intermediate vector from the original observation vector. Then, the initial velocity errors can be eliminated from the newly constructed observation vector. Thus, the interferences of the initial velocity errors for the initial alignment process can be suppressed. The simulation and field tests are designed to verify the performance of the proposed method. The tests results showed that the proposed method can obtain the higher accurate results than the current methods when the initial velocity is considered. Additionally, the results of the proposed method were similar to the current methods when the initial velocity errors were not considered. This shows that the initial velocity errors were eliminated effectively by the proposed method, and the alignment accuracy were not decreased.

Smooth low-rank representation with a Grassmann manifold for tensor completion

Utilizing low-rank representation, recent methods have efficiently estimated the low-rank tensor for tensor completion (TC). However, owing to the identifiability issue, these methods inevitably preserve some outliers of the sparse tensor as elements of the low-rank tensor, which leads to an imprecise estimation of the low-rank tensor. In this paper, we propose a smooth low-rank representation with a Grassmann manifold (SLRR-GM) model to tackle the identifiability issue. Specifically, we first smooth the low-frequency and high-frequency discontinuities of the low-rank tensor to sufficiently remove the preserved outliers of the sparse tensor. Accordingly, a smooth low-rank representation model is built to precisely estimate the low-rank tensor. Second, we smooth the discontinuities through a small number of the feature vectors that characterize the original vectors of the low-rank tensor. This uses the Grassmann manifold to effectively reduce the computational complexity of smoothing the discontinuities. Third, we integrate the smooth low-rank representation model with the Grassmann manifold to propose SLRR-GM. Based on the alternating direction method of multipliers (ADMM), an algorithm is proposed to solve the SLRR-GM model. We also investigate the complexity and convergence to validate the algorithmic effectiveness. Experiments in different scenarios demonstrate that the proposed method outperforms several state-of-the-art methods in terms of the quantitative and visual aspects with the acceptable processing time.

Orbital Angular Momentum Radio-Wave-Based 2-D Target Angle Estimation With Mutual Coupling

Vortex electromagnetic waves are conveniently produced by a uniform circular array (UCA). The mutual coupling matrix (MCM) of UCAs presents rotational symmetry, which can be employed to cut down the parameters in the circuit model. The impedance matrix also does not rely on a specific port. This work uses rotational symmetry to expand the MCM into the orbital angular momentum (OAM) mode domain. The OAM-domain phase sequence is derived by the discrete Fourier transform (DFT) of the first row of the MCM. The connection between the number of OAM modes and the excitation of elements is described. Furthermore, an improved propagator method (PM) for OAM-based decoupled 2-D angle estimation is proposed. This method employs the original steering vector of OAM waves to process the receiving signal, which avoids errors caused by the Bessel function approximation. The estimation error is also reduced by redefining the propagator matrix. In terms of computational complexity, the proposed method avoids spectral search-and-element level eigenvalue decomposition (EVD). The results demonstrate that the coupling is compensated effectively by the OAM-mode domain MCM. The error of the proposed method is less than 1° under the condition of SNR = 20 dB and 1000 snapshots. The computational complexity is reduced by 1–2 orders of magnitude compared to the origin PM. The method is also suitable for 1-D estimation scenarios.

Well-posedness of the deterministic transport equation with singular velocity field perturbed along fractional Brownian paths

In this article we prove path-by-path uniqueness in the sense of Davie [25] and Shaposhnikov [46] for SDE's driven by a fractional Brownian motion with a sufficiently small Hurst parameter H∈(0,12), uniformly in the initial conditions, where the drift vector field is allowed to be merely bounded and measurable.Using this result, we construct weak unique regular solutions in Wlock,p([0,1]×Rd), p>d of the classical transport and continuity equations with singular velocity fields perturbed along fractional Brownian paths.The latter results provide a systematic way of producing examples of singular velocity fields, which cannot be treated by the regularity theory of DiPerna-Lions [28], Ambrosio [2] or Crippa-De Lellis [23].Our approach is based on a priori estimates at the level of flows generated by a sequence of mollified vector fields, converging to the original vector field, and which are uniform with respect to the mollification parameter. In addition, we use a compactness criterion based on Malliavin calculus from [24] as well as a supremum estimate in time of moments of the derivative of the flow of SDE solutions.

Efficient global sensitivity analysis for high-dimensional outputs combining data-driven probability models and dimensionality reduction

• GSA for applications involving complex numerical models and high-dimensional outputs is examined • Two different, recently developed, GSA formulations are combined • Principal component analysis (PCA) is firstly adopted as a dimensionality reduction technique. • Relevant latent output GSA statistics are then calculated by extending probability model-based GSA. • Extension pertains to estimation of necessary covariance statistics This paper examines the efficient variance-based global sensitivity analysis (GSA), quantified by estimating first-/higher-order and total-effect Sobol’ indices, for applications involving complex numerical models and high-dimensional outputs. Two different, recently developed, techniques are combined to address the associated challenges. Principal component analysis (PCA) is first considered as a dimensionality reduction technique. The GSA for the original output vector is then formulated by calculating variance and covariance statistics for the low-dimensional latent output space identified by PCA. These statistics are efficiently approximated by extending recent work on data-driven, probability model-based GSA (PM-GSA). The extension, constituting the main novel contribution of this work, pertains to the estimation of covariance statistics beyond the variance statistics examined in the original PM-GSA formulation. Specifically, a Gaussian mixture model (GMM) is developed to approximate the joint probability density function between some subset of the input vector, and each latent output, or each pair of latent outputs. The GMM is then utilized to estimate the aforementioned statistics. Results across two natural hazards engineering examples show that the dimensionality reduction and transformation of output space established through PCA do not impact the overall accuracy of the PM-GSA, and that the proposed implementation accommodates highly-efficient GSA estimates.

On the Compressive Power of Boolean Threshold Autoencoders.

An autoencoder is a layered neural network whose structure can be viewed as consisting of an encoder, which compresses an input vector to a lower dimensional vector, and a decoder, which transforms the low-dimensional vector back to the original input vector (or one that is very similar). In this article, we explore the compressive power of autoencoders that are Boolean threshold networks by studying the numbers of nodes and layers that are required to ensure that each vector in a given set of distinct input binary vectors is transformed back to its original. We show that for any set of n distinct vectors there exists a seven-layer autoencoder with the optimal compression ratio, (i.e., the size of the middle layer is logarithmic in n ), but that there is a set of n vectors for which there is no three-layer autoencoder with a middle layer of logarithmic size. In addition, we present a kind of tradeoff: if the compression ratio is allowed to be considerably larger than the optimal, then there is a five-layer autoencoder. We also study the numbers of nodes and layers required only for encoding, and the results suggest that the decoding part is the bottleneck of autoencoding. For example, there always is a three-layer Boolean threshold encoder that compresses n vectors into a dimension that is twice the logarithm of n .

Evolutionary Computation in Action: Hyperdimensional Deep Embedding Spaces of Gigapixel Pathology Images

One of the main obstacles of adopting digital pathology is the challenge of efficient processing of hyperdimensional digitized biopsy samples, called whole slide images (WSIs). Exploiting deep learning and introducing compact WSI representations are urgently needed to accelerate image analysis and facilitate the visualization and interpretability of pathology results in a postpandemic world. In this paper, we introduce a new evolutionary approach for WSI representation based on large-scale multi-objective optimization (LSMOP) of deep embeddings. We start with patch-based sampling to feed KimiaNet , a histopathology-specialized deep network, and to extract a multitude of feature vectors. Coarse multi-objective feature selection uses the reduced search space strategy guided by the classification accuracy and the number of features. In the second stage, the frequent features histogram (FFH), a novel WSI representation, is constructed by multiple runs of coarse LSMOP. Fine evolutionary feature selection is then applied to find a compact (short-length) feature vector based on the FFH and contributes to a more robust deep-learning approach to digital pathology supported by the stochastic power of evolutionary algorithms. We validate the proposed schemes using The Cancer Genome Atlas (TCGA) images in terms of WSI representation, classification accuracy, and feature quality. Furthermore, a novel decision space for multicriteria decision making in the LSMOP field is introduced. Finally, a patch-level visualization approach is proposed to increase the interpretability of deep features. The proposed evolutionary algorithm finds a very compact feature vector to represent a WSI (almost 14,000 times smaller than the original feature vectors) with 8% higher accuracy compared to the codes provided by the state-of-the-art methods.

Distributed estimation for multi‐agent systems with relative measurements and quantized communication: A feedback quadratic framework

SummaryIn this paper, the feedback quadratic distributed state estimation problem is investigated for a class of linear discrete time‐varying multi‐agent systems subject to non‐Gaussian noises. For the multi‐agent systems, each agent has access to not only the local measurements but the relative states to its adjacent agents as well. Due to the bandwidth constraints of the digital communication networks, the signals are quantized and then exchanged between two adjacent agents, where the probabilistic uniform quantizations are taken into account. The purpose of the addressed problem is to design a novel quadratic distributed estimator for each agent based on the local information and the predictions received from the neighbors. In particular, an output injection term is introduced to handle the unstable systems. By means of the state/measurement augmentation approach, the underlying system is transformed into an augmented one, which aggregates the original vector and its second‐order Kronecker power. Accordingly, a distributed estimator is constructed such that an upper bound is ensured for the estimation error covariance and the suboptimal estimator parameters, which minimize such an upper bound, are subsequently calculated in terms of the solutions to certain matrix difference equations. Finally, an illustrative example is provided to verify the effectiveness and superiority of the proposed quadratic estimation scheme.

АВТОКОДИРОВЩИК ЭЛЕКТРИЧЕСКОЙ АКТИВНОСТИ ГОЛОВНОГО МОЗГА ЧЕЛОВЕКА

The authors identify hidden parameters of the function describing the electrical activity of the human brain, obtained using electroencephalography (EEG), with the help of an artificial neural network and deep machine learning. The compression of applied information, necessary to reduce the dimensionality of the feature space of the data in order to obtain a model of an artificial neural network– an autoencoder is formulated. The novelty of the general solution and the theoretical aspects and problems of existing compression methods are described. An experimental study is carried out, which consists in obtaining an autoencoder model using applied data EEG sequences containing visual evoked potentials. The compression problem is solved by decreasing the dimensionality of the multidimensional vector associated with the sample. The autoencoder encodes the original multi-dimensional vector into a vector of smaller dimensionality. Using deep machine learning, a coding function is found such that reverse decoding into the original vector can be performed. As a result of the empirical selection of the vector dimensionality, the best experimental model of the autoencoder was chosen, which compresses the feature space of dimensionality equal to 1260 (in the initial sense EEG signals of duration 0,2 s) to a 24-dimensional space, with the possibility of the reconstruction of the initial signal with losses of not more than 10 %.

Stability of Compressed Recursive Least Squares with Forgetting Factor Algorithm

To identify the unknown sparse time-varying parameters of the stochastic dynamic system, we integrate compressive sensing theory with the traditional recursive least squares with forgetting factor (FFLS) algorithm, and propose a compressed adaptive filtering algorithm. Our algorithm is designed to first compress the original high-dimensional sparse regression vector by using the sensing matrix, and then apply the FFLS algorithm to estimate the compressed parameters. Subsequently, the original high-dimensional sparse parameters can be well recovered by a reconstruction technique. We introduce an excitation condition on the compressed stochastic regressors, under which the stability of the proposed algorithm (i.e., the upper bound of the estimation error) is established without assuming independence, stationarity or ergodicity of the system signals. The effectiveness of our theoretical results is demonstrated by a numerical example, which also shows that our proposed algorithm has better performance than both the compressed least mean squares algorithm and the uncompressed FFLS algorithm for tracking high-dimensional sparse parameters.

Identification and dynamics prediction of a plane vortex structure based on a mathematical model of a point vortices system

The aim of the article is developing and analyse an algorithmic method for solution finding of one inverse problem of 2d vortex fluid dynamics. It is identification and prediction of the flow structure evolution of the based on the data on fluid velocity vectors in a set of reference points. Theoretical analysis of convergence and adequacy of the method is difficult due to the ill-posedness typical of inverse problems, these issues studied experimentally. Methods. The proposed method uses a mathematical model of a point vortex dynamics system for identification and prediction flow structures. The parameters of the model system are found by minimising the functional that evaluates the closeness of the original and model vectors fields at the reference points. The prediction of the vortex structure dynamics is based on the solution of the Cauchy problem for a system of ordinary differential equations with the parameters found in the first stage. Results. As a result of the calculations, we found it out: the algorithm converges to the desired minimum from a wide range of initial approximations; the algorithm converges in all cases when the identified structure consists of sufficiently distant vortices; the forecast of the development of the current gives good results with a steady flow; if the above conditions are violated, the part of successful calculations decreases, false identification and an erroneous forecast may occur; with the convergence of the method, the coordinates and circulation of the eddies of the model system are close to the characteristics of the eddies of the test configurations; the structures of the streamlines of the flows are topologically equivalent; convergence depends more on location than on the number of vectors used for identification. Conclusion. An algorithm for solving the problem of identifying and the evolution forecast of a 2d vortex flow structure is proposed when the fluid velocity vectors in a finite set of reference points are known. The method showed its high efficiency when using from 40 to 200 reference points. The results of the study make it possible to recommend the proposed algorithm for identifying flat vortex structures, which consist of vortices separated from each other.

Classify Human Activities Based On Deepforest

One important aspect of human behavior understanding is the recognition and monitoring of daily activities. An accurate activity recognition system can improve the quality of life in many key areas. The multi-metric feature extraction and DeepForest classifier designed in this paper effectively had solved the problems of incomprehensive feature extraction and insufficient classifier accuracy.First, this paper extracts a total of 450 dimensional feature vectors using mean, variance, maximum, skewness, minimum, kurtosis, regression of independent variables and sample entropy as feature indicators, so that the features of the original vectors can be presented comprehensively. The dimensionality of the feature set is too large, so PCA is used to reduce the dimensionality of the high-dimensional feature vector. After studying the relationship between the dimensionality reduction and the cumulative contribution rate, it is concluded that the cumulative contribution rate reaches 90% when the dimensionality is 15, which can retain the original features better. Then DeepForest was used as the classifier, and the reduced-dimensional feature set was used as the sample set to divide the training set and test set by 3:1. The model was tested on the test set after training, and the classification accuracy of the test set was 98.202%. Six classifiers, GaussianNB, SVM, K-NN, XGBoost, RandomForest and DecisionTree, were selected as the control experimental group, of which only the RandomForest model reached 97%, while the rest of the control models did not achieve 95% effect, indicating that DeepForest was more accurate in classifying human activities.Next, the generalisation ability of the DeepForest classifier was assessed using Monte Carlo Cross Validation (MCCV), K-fold and its confusion matrix. The MCCV validation used 50% of the data as the training set and 30% of the data as the test set, and set the number of splits to 10, resulting in a mean accuracy of 98.227%. The size of K in the K-fold validation was determined to be 8 based on the number of people conducting the human activity experiment, and the final mean accuracy value obtained was 97.790%. The combined confusion matrix from the K-fold validation (which aggregates the confusion matrix for each classification result) was calculated, and the results showed that the highest accuracy reached 100% for A9 and A14 classifications, where A5 and A10, A15 and A10, A12 and A5 were more likely to be confused, with the highest error rate for A5 classification, which was 4.167%, and the rest of the activity classifications were better.

Revisiting Multi-Codebook Quantization.

Multi-Codebook Quantization (MCQ) is a generalized version of existing codebook-based quantizations for Approximate Nearest Neighbor (ANN) search. Specifically, MCQ picks one codeword for each sub-codebook independently and takes the sum of picked codewords to approximate the original vector. The objective function involves no constraints, therefore, MCQ theoretically has the potential to achieve the best performance because solutions of other codebook-based quantization methods are all covered by MCQ's solution space under the same codebook size setting. However, finding the optimal solution to MCQ is proved to be NP-hard due to its encoding process, i.e., converting an input vector to a binary code. To tackle this, researchers apply constraints to it to find near-optimal solutions or employ heuristic algorithms that are still time-consuming for encoding. Different from previous approaches, this paper takes the first attempt to find a deep solution to MCQ. The encoding network is designed to be as simple as possible, so the very complex encoding problem becomes simply a feed-forward. Compared with other methods on three datasets, our method shows state-of-the-art performance. Notably, our method is 11×-38× faster than heuristic algorithms for encoding, which makes it more practical for the real scenery of large-scale retrieval. Our code is publicly available: https://github.com/DeepMCQ/DeepQ.

DETECTION OF ANOMALOUS EVENTS BASED ON DEEP LEARNING-BILSTM

Video anomaly detection in smart cities is a critical errand in computer vision that plays an imperative role in intelligent surveillance and public security but is challenging due to its differing, complex, and rare event in real-time surveillance situations. Different deep learning models utilize a critical amount of training data without generalization capabilities and with long time complexity. In this work, and to overcome these problems, an algorithm for reducing the size of the extracted features have been suggested, and this was done by combining every 15 video frames to generate the new features vectors which will be fed into our classifier model, the values of new features vectors represent the summation of the values of original features vectors got from Resnet50. Finally, the new feature vectors are fed into our classifier model to detect the abnormality. We conducted comprehensive tests on a variety of anomaly detection benchmark datasets to verify the proposed framework's functionality in complex surveillance scenarios. The Numerical results were carried out on the UCF-Crime dataset, with the proposed approach achieving Area Under Curve (AUC) scores of 93.61% on the database's test set.

Fixed point theorems and periodic problems for nonlinear Hill’s equation

Two main results of fixed point theory in infinite dimensional space are Schauder’s theorem and the contraction mapping principle. Krasnoselskii combined them into one fixed point result. In this paper, we continue the study of extensions of these theorems investigating a convex modular in a original vector space, not in modular space and without Delta _{2} condition, to provide certain extensions of Banach contraction principle and Krasnoselskii fixed point theorem. We applied that theorem to solve the nonlinear periodic problem of Hill’s equation.

Animal Fiber Recognition Based on Feature Fusion of the Maximum Inter-Class Variance

Abstract Cashmere and wool are common raw materials in the textile industry. The clothes made of cashmere are popular because of the excellent comfort. A system that can quickly and automatically classify the two will improve the efficiency of fiber recognition in the textile industry. We propose a classification method of cashmere and wool fibers based on feature fusion using the maximum inter-class variance. First, the fiber target area is obtained by the preprocessing algorithm. Second, the features of sub-images are extracted through the algorithm of the Discrete Wavelet Transform. It is linearly fused by introducing the weight in the approximate and detailed features. The maximum separability of the feature data can be achieved by the maximum inter-class variance. Finally, different classifiers are used to evaluate the performance of the proposed method. The support vector machine classifier can achieve the highest recognition rate, with an accuracy of 95.20%. The experimental results show that the recognition rate of the fused feature vectors is improved by 6.73% compared to the original feature vectors describing the image. It verifies that the proposed method provides an effective solution for the automatic recognition of cashmere and wool.

The Optimized γ-Globin Lentiviral Vector GGHI-mB-3D Leads to Nearly Therapeutic HbF Levels In Vitro in CD34+ Cells from Sickle Cell Disease Patients.

We have previously demonstrated that both the original γ-globin lentiviral vector (LV) GGHI and the optimized GGHI-mB-3D LV, carrying the novel regulatory elements of the 3D HPFH-1 enhancer and the 3' β-globin UTR, can significantly increase HbF production in thalassemic CD34+ cells and ameliorate the disease phenotype in vitro. In the present study, we investigated whether the GGHI-mB-3D vector can also exhibit an equally therapeutic effect, following the transduction of sickle cell disease (SCD) CD34+ cells at MOI 100, leading to HbF increase coupled with HbS decrease, and thus, to phenotype improvement in vitro. We show that GGHI-mB-3D LV can lead to high and potentially therapeutic HbF levels, reaching a mean 2-fold increase to a mean value of VCN/cell of 1.0 and a mean transduction efficiency of 55%. Furthermore, this increase was accompanied by a significant 1.6-fold HbS decrease, a beneficial therapeutic feature for SCD. In summary, our data demonstrate the efficacy of the optimized γ-globin lentiviral vector to improve the SCD phenotype in vitro, and highlights its potential use in future clinical SCD trials.

A trajectory and force dual-incremental robot skill learning and generalization framework using improved dynamical movement primitives and adaptive neural network control

Due to changes in the environment and errors that occurred during skill initialization, the robot's operational skills should be modified to adapt to new tasks. As such, skills learned by the methods with fixed features, such as the classical Dynamical Movement Primitive (DMP), are difficult to use when the using cases are significantly different from the demonstrations. In this work, we propose an incremental robot skill learning and generalization framework including an incremental DMP (IDMP) for robot trajectory learning and an adaptive neural network (NN) control method, which are incrementally updated to enable robots to adapt to new cases. IDMP uses multi-mapping feature vectors to rebuild the forcing function of DMP, which are extended based on the original feature vector. In order to maintain the original skills and represent skill changes in a new task, the new feature vector consists of three parts with different usages. Therefore, the trajectories are gradually changed by expanding the feature and weight vectors, and all transition states are also easily recovered. Then, an adaptive NN controller with performance constraints is proposed to compensate dynamics errors and changed trajectories after using the IDMP. The new controller is also incrementally updated and can accumulate and reuse the learned knowledge to improve the learning efficiency. Compared with other methods, the proposed framework achieves higher tracking accuracy, realizes incremental skill learning and modification, achieves multiple stylistic skills, and is used for obstacle avoidance with different heights, which are verified in three comparative experiments.