High-dimensional Feature Vectors Research Articles

Direction finding for coherent sources with deep hybrid neural networks

ABSTRACT In this paper, an intelligent direction finding (DF) architecture based on hybrid neural network is proposed. Firstly, the array outputs of DF system are multiplied by the corresponding ideal array manifold and a high-dimensional preprocessed feature vector is derived. In the next, the preprocessed feature is straightly fed into the auto-encoder. The auto-encoder outputs are then used as the input of the cascaded deep residual neural network in order to extract spatial spectral of sources. Finally, the direction-of-arrival (DOA) results are derived according to the locations of the normalised spectral peaks. There are two advantages with our proposed network architecture. One is the DF estimation accuracy is independent on the array aperture size when the number of elements is fixed. And the other is that the proposed network structure and processing procedures can be directly generalised into unknown scenarios but having higher angle resolution capacity at the same time. Finally, extensive numerical experiments illustrate the correctness and potential advantages of the proposed deep hybrid neural network.

Improved Deep Hashing with Scalable Interblock for Tourist Image Retrieval

Tourist image retrieval has attracted increasing attention from researchers. Mainly, supervised deep hash methods have significantly boosted the retrieval performance, which takes hand-crafted features as inputs and maps the high-dimensional binary feature vector to reduce feature-searching complexity. However, their performance depends on the supervised labels, but few labeled temporal and discriminative information is available in tourist images. This paper proposes an improved deep hash to learn enhanced hash codes for tourist image retrieval. It jointly determines image representations and hash functions with deep neural networks and simultaneously enhances the discriminative capability of tourist image hash codes with refined semantics of the accompanying relationship. Furthermore, we have tuned the CNN to implement end-to-end training hash mapping, calculating the semantic distance between two samples of the obtained binary codes. Experiments on various datasets demonstrate the superiority of the proposed approach compared to state-of-the-art shallow and deep hashing techniques.

Urban Morphological Feature Extraction and Multi-Dimensional Similarity Analysis Based on Deep Learning Approaches

The study of urban morphology contributes to the evolution of cities and sustainable development. Urban morphological feature extraction and similarity analysis represents a practical framework in many studies to interpret and introduce the current built environment to aid in proposing novel designs. In conventional methods, morphological features are represented based on qualitative descriptions, symbolical interpretation, or manually selected indicators. However, these methods could cause subjective bias and limit the generalizability. This study proposes a hybrid data-driven approach to support quantitative morphological descriptions and multi-dimensional similarity analysis for urban design decision-making and to further morphology-related studies using information abundance via a deep-learning approach. We constructed a dataset of 3817 residential plots with geometrical and related infrastructure information. A deep convolutional neural network, GoogLeNet, was implemented with the plots’ figure–ground images, by quantifying the morphological features into 2048-dimensional feature vectors. We conducted a similarity analysis of the plots by calculating the Euclidean distance between the high-dimensional feature vectors. Then, a comparison study was performed by retrieving cases based on the plot shape and plots with buildings separately. The proposed method considers the overall characteristics of the urban morphology and social infrastructure situations for similarity analysis. This method is flexible and effective. The proposed framework indicates the feasibility and potential of integrating task-oriented information to introduce custom and adequate references via deep learning methods, which could support decision making and association studies on morphology with urban consequences. This work could serve as a basis for further typo-morphology studies and other morphology-related ecological, social, and economic studies for sustainable built environments.

Land Cover Classification using Machine Learning Approaches from High Resolution Images

Land cover classification has become an interesting research area in the field of remote sensing. Machine learning techniques have shown great success for various application in the domain of land cover classification. This paper focuses on the classification of land covers obtained from high resolution images using two well-known classification methods by integrating with object-based segmentation technique. First, graph-based minimal spanning tree segmentation was applied to segment the original image pixels into objects. The segmented objects were then used to obtained spectral, spatial and texture features which were then combined to form a single high dimensional feature vector. These features were then used to train and test the artificial neural network (ANN) and support vector machine (SVM). The proposed method was evaluated on a dataset consisting of high resolution multi-spectral images with four classes (tea area, other trees, roads and builds, bare land). The experiments showed that ANN was more accuracy as it scored average accuracy of 82.60% while SVM produced 73.66%. Moreover, when postprocessing using majority analysis was applied, the average accuracy improved to 86.18%.

Land Cover Classification using Machine Learning Approaches from High Resolution Images

Land cover classification has become an interesting research area in the field of remote sensing. Machine learning techniques have shown great success for various application in the domain of land cover classification. This paper focuses on the classification of land covers obtained from high resolution images using two well-known classification methods by integrating with object-based segmentation technique. First, graph-based minimal spanning tree segmentation was applied to segment the original image pixels into objects. The segmented objects were then used to obtained spectral, spatial and texture features which were then combined to form a single high dimensional feature vector. These features were then used to train and test the artificial neural network (ANN) and support vector machine (SVM). The proposed method was evaluated on a dataset consisting of high resolution multi-spectral images with four classes (tea area, other trees, roads and builds, bare land). The experiments showed that ANN was more accuracy as it scored average accuracy of 82.60% while SVM produced 73.66%. Moreover, when postprocessing using majority analysis was applied, the average accuracy improved to 86.18%.

An Optimal Feature Selection Method for Automatic Face Retrieval Using Enhanced Grasshopper Optimization Algorithm

Background: Retrieval of facial images based on its contents is one of the main areas of research. However, images contain high dimensional feature vectors and it is a challenging task to select the relevant features due to the variations available in the images of similar objects. Therefore, the selection of relevant features is an important step to make the facial retrieval system computationally efficient and more accurate. Objective: The main aim of this paper is to design and develop an efficient feature selection method for obtaining relevant and non-redundant features from the face images so that the accuracy and computational cost of a face retrieval system can be improved. Methods: The proposed feature selection method uses a new enhanced grasshopper optimization algorithm to obtain the significant features from the high dimensional features vector of face images. The proposed algorithm modifies the target vector by considering more than one best solution which maintain the elitism property and save the search from local optimum. Furthermore, it has been utilized to select the prominent features from the high dimensional facial features vector. Results: The performance of the proposed feature selection method has been tested on Oracle Research Laboratory face database. The proposed method eliminates 89% features which are minimum among the other methods and increases the accuracy of face retrieval system to 96.5%. Conclusion: The enhanced grasshopper optimization algorithm-based feature selection method for face retrieval system outperforms the existing methods in terms of accuracy and computational cost.

Counterfactual Explanations for Oblique Decision Trees:Exact, Efficient Algorithms

We consider counterfactual explanations, the problem of minimally adjusting features in a source input instance so that it is classified as a target class under a given classifier. This has become a topic of recent interest as a way to query a trained model and suggest possible actions to overturn its decision. Mathematically, the problem is formally equivalent to that of finding adversarial examples, which also has attracted significant attention recently. Most work on either counterfactual explanations or adversarial examples has focused on differentiable classifiers, such as neural nets. We focus on classification trees, both axis-aligned and oblique (having hyperplane splits). Although here the counterfactual optimization problem is nonconvex and nondifferentiable, we show that an exact solution can be computed very efficiently, even with high-dimensional feature vectors and with both continuous and categorical features, and demonstrate it in different datasets and settings. The results are particularly relevant for finance, medicine or legal applications, where interpretability and counterfactual explanations are particularly important.

Global Oceanic Eddy Identification: A Deep Learning Method From Argo Profiles and Altimetry Data

The inadequate spatial resolution of altimeter results in low identification efficiency of oceanic eddies, especially for small-scale eddies. It is well known that eddies can not only induce sea surface signal but more importantly have typical vertical structure characteristics. However, although the vertical structure characteristics are usually used for statistical analysis, they are seldom considered in the process of eddy recognition. This study is devoted to identifying eddies from the perspective of their vertical signal derived from the 18-year Argo data. Due to the irregular and noisy profile pattern, the direct identification of eddy core from Argo profile is deemed to be a challenge. With the popularity of artificial intelligence, a new hybrid method that combines the advantages of convolutional neural network (CNN) with extreme gradient boosting (XGBoost) is proposed to extract the representative vertical feature and identify eddy from a profile. First, CNN is employed as a feature extractor to automatically obtain vertical features from the input profile at the bottom of the network. Second, the obtained high-dimensional feature vectors are inputted into the XGBoost model, combined with other profile features for classifying profiles that are outside altimeter-identified eddies (Alt eddy). Finally, extensive experiments are implemented to demonstrate the efficiency of the proposed method. The results show that the classification accuracy of CNN-XGBoost model can reach 98%, and about 36% eddies are recaptured. These eddies, dubbed CNN-XGB eddies, are benchmarked against Alt eddies for the vertical structure and geographical distribution, demonstrating a similar or even stronger vertical signal and a prominent eddy belt in the tropical ocean. Within the proposed theory framework, there are various potentials to obtain a better outlook for eddy identification and in situ float observations.

Supervised Learning for Nonsequential Data: A Canonical Polyadic Decomposition Approach.

Efficient modeling of feature interactions underpins supervised learning for nonsequential tasks, characterized by a lack of inherent ordering of features (variables). The brute force approach of learning a parameter for each interaction of every order comes at an exponential computational and memory cost (curse of dimensionality). To alleviate this issue, it has been proposed to implicitly represent the model parameters as a tensor, the order of which is equal to the number of features; for efficiency, it can be further factorized into a compact tensor train (TT) format. However, both TT and other tensor networks (TNs), such as tensor ring and hierarchical Tucker, are sensitive to the ordering of their indices (and hence to the features). To establish the desired invariance to feature ordering, we propose to represent the weight tensor through the canonical polyadic (CP) decomposition (CPD) and introduce the associated inference and learning algorithms, including suitable regularization and initialization schemes. It is demonstrated that the proposed CP-based predictor significantly outperforms other TN-based predictors on sparse data while exhibiting comparable performance on dense nonsequential tasks. Furthermore, for enhanced expressiveness, we generalize the framework to allow feature mapping to arbitrarily high-dimensional feature vectors. In conjunction with feature vector normalization, this is shown to yield dramatic improvements in performance for dense nonsequential tasks, matching models such as fully connected neural networks.

Exploring human-nature interactions in national parks with social media photographs and computer vision.

Understanding the activities and preferences of visitors is crucial for managing protected areas and planning conservation strategies. Conservation culturomics promotes the use of user-generated online content in conservation science. Geotagged social media content is a unique source of in situ information on human presence and activities in nature. Photographs posted on social media platforms are a promising source of information, but analyzing large volumes of photographs manually remains laborious. We examined the application of state-of-the-art computer-vision methods to studying human-nature interactions. We used semantic clustering, scene classification, and object detection to automatically analyze photographs taken in Finnish national parks by domestic and international visitors. Our results showed that human-nature interactions can be extracted from user-generated photographs with computer vision. The different methods complemented each other by revealing broad visual themes related to level of the data set, landscape photogeneity, and human activities. Geotagged photographs revealed distinct regional profiles for national parks (e.g., preferences in landscapes and activities), which are potentially useful in park management. Photographic content differed between domestic and international visitors, which indicates differences in activities and preferences. Information extracted automatically from photographs can help identify preferences among diverse visitor groups, which can be used to create profiles of national parks for conservation marketing and to support conservation strategies that rely on public acceptance. The application of computer-vision methods to automatic content analysis of photographs should be explored further in conservation culturomics, particularly in combination with rich metadata available on social media platforms.

Basis exchange and learning algorithms for extracting collinear patterns

ABSTRACT Understanding large data sets is one of the most important and challenging problems in the modern days. Exploration of genetic data sets composed of high dimensional feature vectors can be treated as a leading example in this context. A better understanding of large, multivariate data sets can be achieved through exploration and extraction of their structure. Collinear patterns can be an important part of a given data set structure. Collinear (flat) pattern exists in a given set of feature vectors when many of these vectors are located on (or near) some plane in the feature space. Discovered flat patterns can reflect various types of interaction in an explored data set. The presented paper compares basis exchange algorithms with learning algorithms in the task of flat patterns extraction.

A feature-level fusion based improved multimodal biometric recognition system using ear and profile face

Ear based identity recognition subject to uncontrolled conditions such as illumination changes, pose variation, low contrast, partial occlusion and noise, is an active research area in the field of biometrics. Meanwhile, multimodal biometrics is becoming increasingly popular and offers improved performance due to the use of multiple sources of information. In this paper, a multimodal biometrics system is proposed based on the ear and profile face that not only alleviates the short-comings of ear biometrics but also improves the overall recognition rate. The ear and profile face modalities are first represented individually using the combination of two efficient local feature descriptors namely, local phase quantization (LPQ) and local directional patterns (LDP). These histogram-based local descriptors are then combined into a high-dimensional feature vector that preserves complementary information in both frequency and spatial domains. The PCA along with the z-score normalization technique is independently applied on each feature vector and the resultant reduced feature vectors are combined at the feature level. The kernel discriminative common vector (KDCV) approach is finally exploited over the combined feature set to derive more discriminative and non-linear features for the identification of individuals using kNN classifier. The effectiveness of the proposed model has been verified with the deep features derived from three popular pre-trained CNN models such as AlexNet, VGG16 and GoogleNet. Experimental results on two benchmark databases clearly show that the proposed approach achieves better performance than individual modality and other state-of-the-art methods.

Dynamic Mode Decomposition based salient edge/region features for content based image retrieval

Considering the gap between low-level image features and high-level retrieval concept, this paper investigates the effect of incorporating visual saliency based features for content-based image retrieval(CBIR).Visual saliency plays an important role in human perception due to its capability to focus the attention on the point of interest, i.e. an intended target. This selection based processing can be well explored in localized CBIR systems, since in context of CBIR the users will be interested only in certain parts of the image. The proposed methodology uses Dynamic Mode Decomposition framework to extract the saliency map which highlights the part of the image that grabs human attention. Then, based on the saliency map, an efficient salient edge detection model is introduced. Visual saliency based features (salient region, edge) are then combined with texture and color features to formulate the high dimensional feature vector for image retrieval. State-of-the-art learning based CBIR models demands for user feed back to model the retrieval concept. In contrast with these models, proposed CBIR system does not require any user interaction, since it uses perceptual level features for the retrieval task. Performance of the proposed CBIR system is evaluated and confirmed on images from Wang’s dataset using benchmark evaluation metrics like precision and recall. Experimental results reveals that incorporation of saliency features can represent human perception well and produces good retrieval performance.

Deep multi-sequence multi-grained cascade forest for tobacco drying condition identification

The term, “drying condition” refers to the actual dehydration capacity of a rotary dryer in the tobacco drying process which is directly related to the drying effect. However, identifying different drying conditions relies heavily on the judgment of field engineers who have rich domain knowledge and practical experiences. In this study, we proposed an improved multi-sequence multi-grained cascade forest model, MSgcForest, to classify and identify different drying conditions. An improved multi-sequence multi-grained feature scanning mechanism is added to perform spacial and sequential feature extraction from raw production-related data, which transforms the input features into high-dimensional feature vectors and increases the discriminative power of the drying condition features. Comparison with existing models indicates that the proposed MSgcForest outperforms the other alternatives even for small-scale training data. In particular, this method successfully transforms the fuzzy artificial judgment of the drying condition into a data-driven identification with high precision, which provides a promising prospect for identifying working conditions in industrial processes.

Improving Hate Speech Detection of Urdu Tweets Using Sentiment Analysis

Sentiment Analysis is a technique that is being used abundantly nowadays for customer reviews analysis, popularity analysis of electoral candidates, hate speech detection and similar applications. Sentiment analysis on tweets encounters challenges such as highly skewed classes, high dimensional feature vectors and highly sparse data. In this study, we have analyzed the improvement achieved by successively addressing these problems in order to determine their severity for sentiment analysis of tweets. Firstly, we prepared a comprehensive data set consisting of Urdu Tweets for sentiment analysis-based hate speech detection. To improve the performance of the sentiment classifier, we employed dynamic stop words filtering, Variable Global Feature Selection Scheme (VGFSS) and Synthetic Minority Optimization Technique (SMOTE) to handle the sparsity, dimensionality and class imbalance problems respectively. We used two machine learning algorithms i.e., Support Vector Machines (SVM) and Multinomial Naïve Bayes' (MNB) for investigating performance in our experiments. Our results show that addressing class skew along with alleviating the high dimensionality problem brings about the maximum improvement in the overall performance of the sentiment analysis-based hate speech detection.

An AVMD Method Based on Energy Ratio and Deep Belief Network for Fault Identification of Automation Transmission Device

Considering that the vibration signals of gears and bearings in the automatic transmission device are complex and the fault features are difficult to extract. This paper proposes a method for extracting fault features of transmission device using adaptive variational modal decomposition (AVMD), and uses deep belief network (DBN) for pattern recognition. The vibration signal is decomposed by AVMD using energy ratio method. The intrinsic mode function (IMF) with abundant fault information is obtained. By calculating the energy entropy of each IMF component and form a high-dimensional feature vector as the input of DBN to establish an early fault identification model. The early fault data of the PHM2009 transmission device experimental platform was selected for identification and analysis. The identification results show that AVMD can extract the weak features of transmission device fault signals more accurately than empirical mode decomposition (EMD). Moreover, DBN has a higher fault identification accuracy rate than support vector machine (SVM), probabilistic neural network (PNN), back propagation neural network (BP) and Kohonen self-organizing competition neural network.

Flow Regimes Identification-based Multidomain Features for Gas–Liquid Two-Phase Flow in Horizontal Pipe

Accurate flow regimes identification is of great significance for the deep understanding of the flow structure and ensuring the safe and economic operation of actual production in the multiphase flow process. For the flow regimes identification of the gas-liquid two-phase flow in the horizontal pipe, a multidomain features processing scheme is proposed. The water holdup data are acquired by conductance rings sensor, and processed by the improved empirical wavelet transform (EWT) to obtain the Fourier spectrum in the frequency domain and decomposition components in the multiscale domain. The obtained multidomain feature information is quantified to get the high-dimensional feature vectors of different flow regimes. To make the features more representative and realize the visualization of the structural relationship between the flow regime samples, the isometric feature mapping (Isomap) method in manifold learning is used to synthesize the feature structure relationships and map them to low-dimensional space. The obtained low-dimensional feature vectors of training samples are input into the support vector machine (SVM) to acquire the flow regimes identification model. Compared with other flow regimes identification schemes that feature reduction uses different manifold learning methods, or feature extraction is not in multidomain, or data preprocessing uses traditional wavelet transform, the proposed scheme can better retain the global structural relationship of the flow regimes, has the characteristics of strong stability, and the identification rate of the gas-liquid two-phase flow regimes in the horizontal pipe is over 95%.

Performance Boosting of Scale and Rotation Invariant Human Activity Recognition (HAR) with LSTM Networks Using Low Dimensional 3D Posture Data in Egocentric Coordinates

Human activity recognition (HAR) has been an active area in computer vision with a broad range of applications, such as education, security surveillance, and healthcare. HAR is a general time series classification problem. LSTMs are widely used for time series classification tasks. However, they work well with high-dimensional feature vectors, which reduce the processing speed of LSTM in real-time applications. Therefore, dimension reduction is required to create low-dimensional feature space. As it is experimented in previous study, LSTM with dimension reduction yielded the worst performance among other classifiers, which are not deep learning methods. Therefore, in this paper, a novel scale and rotation invariant human activity recognition system, which can also work in low dimensional feature space is presented. For this purpose, Kinect depth sensor is employed to obtain skeleton joints. Since angles are used, proposed system is already scale invariant. In order to provide rotation invariance, body relative direction in egocentric coordinates is calculated. The 3D vector between right hip and left hip is used to get the horizontal axis and its cross product with the vertical axis of global coordinate system assumed to be the depth axis of the proposed local coordinate system. Instead of using 3D joint angles, 8 number of limbs and their corresponding 3D angles with X, Y, and Z axes of the proposed coordinate system are compressed with several dimension reduction methods such as averaging filter, Haar wavelet transform (HWT), and discrete cosine transform (DCT) and employed as the feature vector. Finally, extracted features are trained and tested with LSTM (long short-term memory) network, which is an artificial recurrent neural network (RNN) architecture. Experimental and benchmarking results indicate that proposed framework boosts the performance of LSTM by approximately 30% accuracy in low-dimensional feature space.

Improved coral reefs optimization with adaptive $$\beta $$-hill climbing for feature selection

For any classification problem, the dimension of the feature vector used for classification has great importance. This is because, in a high-dimensional feature vector, it is found that some are non-informative or even redundant as they do not contribute to the learning process of the classifier. Rather, they may be the reason for low classification accuracy and high training time of the learning model. To address this issue, researchers apply various feature selection (FS) methods as found in the literature. In recent years, meta-heuristic algorithms have been proven to be effective in solving FS problems. The Coral Reefs Optimizer (CRO) which is a cellular type evolutionary algorithms has good tuning between its exploration and exploitation ability. This has motivated us to present an improved version of CRO with the inclusion of adaptive $$\beta $$ -hill climbing to increase the exploitation ability of CRO. The proposed method is assessed on 18 standard UCI-datasets by means of three distinct classifiers, KNN, Random Forest and Naive Bayes classifiers. It is also analyzed with 10 state-of-the-art meta-heuristics FS procedure, and the outputs show an excellent performance of the proposed FS method reaching better results than the previous methods considered here for comparison. The source code of this work is publicly available at https://github.com/ahmed-shameem/Projects .

Fingerprint liveness detection through fusion of pores perspiration and texture features

Spoofing attacks on the fingerprint scanners become major and serious concern with the growing development and use of biometric technologies. Level 1 and Level 2 features; which are said to be unique and most commonly used features in fingerprint verification systems, are easily get spoofed. Moreover, single feature based specifically designed spoof detection methods are not performed well on different fingerprint scanners and spoofing materials. This paper proposed the fusion of pores perspiration and texture features in static software based approach to identify live and fake fingerprints. The pores perspiration activity is quantified by computing the ridge signal energy and gray level distributions around the detected pores. These pore characteristics are statically determined instead of dynamic measurement. Autoencoder neural network is used to reduce the high dimensional feature vector and learn its low dimensional hidden representation unsupervisedly. The binary classification in two classes: live and spoof is performed by the supervisedly trained softmax classifier. The performance of the classifier is evaluated in terms of Average Classification Error (ACE) and misclassification rates: FerrLive and FerrFake. The experimental results carried out on LivDet 2013 and LivDet 2015 databases show the improvement of the classifier performance in comparison to the state-of-the-art methods.