Robust face recognition via sparse representation

Jointly constrained group sparse connectivity representation improves early diagnosis of Alzheimer's disease on routinely acquired T1-weighted imaging-based brain network

Background

Radiomics-based morphological brain networks (radMBN) constructed from routinely acquired structural MRI (sMRI) data have gained attention in Alzheimer's disease (AD). However, the radMBN suffers from limited characterization of AD because sMRI only characterizes anatomical changes and is not a direct measure of neuronal pathology or brain activity.

Purpose

To establish a group sparse representation of the radMBN under a joint constraint of group-level white matter fiber connectivity and individual-level sMRI regional similarity (JCGS-radMBN).

Methods

Two publicly available datasets were adopted, including 120 subjects from ADNI with both T1-weighted image (T1WI) and diffusion MRI (dMRI) for JCGS-radMBN construction, 818 subjects from ADNI and 200 subjects solely with T1WI from AIBL for validation in early AD diagnosis. Specifically, the JCGS-radMBN was conducted by jointly estimating non-zero connections among subjects, with the regularization term constrained by group-level white matter fiber connectivity and individual-level sMRI regional similarity. Then, a triplet graph convolutional network was adopted for early AD diagnosis. The discriminative brain connections were identified using a two-sample t-test, and the neurobiological interpretation was validated by correlating the discriminative brain connections with cognitive scores.

Results

The JCGS-radMBN exhibited superior classification performance over five brain network construction methods. For the typical NC vs. AD classification, the JCGS-radMBN increased by 1-30% in accuracy over the alternatives on ADNI and AIBL. The discriminative brain connections exhibited a strong connectivity to hippocampus, parahippocampal gyrus, and basal ganglia, and had significant correlation with MMSE scores.

Conclusion

The proposed JCGS-radMBN facilitated the AD characterization of brain network established on routinely acquired imaging modality of sMRI.

Supplementary Information

The online version of this article (10.1007/s13755-023-00269-0) contains supplementary material, which is available to authorized users.

On-the-move heterogeneous face recognition in frequency and spatial domain using sparse representation

Heterogeneity of a probe image is one of the most complex challenges faced by researchers and implementers of current surveillance systems. This is due to existence of multiple cameras working in different spectral ranges in a single surveillance setup. This paper proposes two different approaches including spatial sparse representations (SSR) and frequency sparse representation (FSR) to recognize on-the-move heterogeneous face images with database of single sample per person (SSPP). SCface database, with five visual and two Infrared (IR) cameras, is taken as a benchmark for experiments, which is further confirmed using CASIA NIR-VIS 2.0 face database with 17580 visual and IR images. Similarity, comparison is performed for different scenarios such as, variation of distances from a camera and variation in sizes of face images and various visual and infrared (IR) modalities. Least square minimization based approach for finding the solution is used to match face images as it makes the recognition process simpler. A side by side comparison of both the proposed approaches with the state-of-the-art, classical, principal component analysis (PCA), kernel fisher analysis (KFA) and coupled kernel embedding (CKE) methods, along with modern low-rank preserving projection via graph regularized reconstruction (LRPP-GRR) method, is also presented. Experimental results suggest that the proposed approaches achieve superior performance.

Typicality-Aware Adaptive Similarity Matrix for Unsupervised Learning

Graph-based clustering approaches, especially the family of spectral clustering, have been widely used in machine learning areas. The alternatives usually engage a similarity matrix that is constructed in advance or learned from a probabilistic perspective. However, unreasonable similarity matrix construction inevitably leads to performance degradation, and the sum-to-one probability constraints may make the approaches sensitive to noisy scenarios. To address these issues, the notion of typicality-aware adaptive similarity matrix learning is presented in this study. The typicality (possibility) rather than the probability of each sample being a neighbor of other samples is measured and adaptively learned. By introducing a robust balance term, the similarity between any pairs of samples is only related to the distance between them, yet it is not affected by other samples. Therefore, the impact caused by the noisy data or outliers can be alleviated, and meanwhile, the neighborhood structures can be well captured according to the joint distance between samples and their spectral embeddings. Moreover, the generated similarity matrix has block diagonal properties that are beneficial to correct clustering. Interestingly, the results optimized by the typicality-aware adaptive similarity matrix learning share the common essence with the Gaussian kernel function, and the latter can be directly derived from the former. Extensive experiments on synthetic and well-known benchmark datasets demonstrate the superiority of the proposed idea when comparing with some state-of-the-art methods.

SERS microscopy as a tool for comprehensive biochemical characterization in complex samples

Surface enhanced Raman scattering (SERS) spectra of biomaterials such as cells or tissues can be used to obtain biochemical information from nanoscopic volumes in these heterogeneous samples. This tutorial review discusses the factors that determine the outcome of a SERS experiment in complex bioorganic samples. They are related to the SERS process itself, the possibility to selectively probe certain regions or constituents of a sample, and the retrieval of the vibrational information in order to identify molecules and their interaction. After introducing basic aspects of SERS experiments in the context of biocompatible environments, spectroscopy in typical microscopic settings is exemplified, including the possibilities to combine SERS with other linear and non-linear microscopic tools, and to exploit approaches that improve lateral and temporal resolution. In particular the great variation of data in a SERS experiment calls for robust data analysis tools. Approaches will be introduced that have been originally developed in the field of bioinformatics for the application to omics data and that show specific potential in the analysis of SERS data. They include the use of simulated data and machine learning tools that can yield chemical information beyond achieving spectral classification.

Vein pattern visualisation for biometric identification with cGAN on a New Zealand dataset

Forensic identification using vein patterns in standard colour images presents significant challenges due to their low visibility. Recent efforts have employed various computational techniques, including artificial neural networks and optical vein disclosure, to enhance vein pattern detection. However, these methods still face limitations in reliability when compared to Near-Infrared (NIR) reference images. One of the biggest challenges of the studies is the limited number of available datasets that have synchronised colour and NIR images from body limbs. This paper introduces a new dataset comprising 602 pairs of synchronised NIR and RGB forearm images from a diverse population, ethically approved and collected in Auckland, New Zealand. Using this dataset, we also propose a conditional Generative Adversarial Networks (cGANs) model to translate RGB images into their NIR equivalents. Our evaluations focus on matching accuracy, vein length measurements, and contrast quality, demonstrating that the translated vein patterns closely resemble their NIR counterparts. This advancement offers promising implications for forensic identification techniques.

Accelerated Stochastic Variance Reduction Gradient Algorithms for Robust Subspace Clustering

Robust face clustering enjoys a wide range of applications for gate passes, surveillance systems and security analysis in embedded sensors. Nevertheless, existing algorithms have limitations in finding accurate clusters when data contain noise (e.g., occluded face clustering and recognition). It is known that in subspace clustering, the ℓ1- and ℓ2-norm regularizers can improve subspace preservation and connectivity, respectively, and the elastic net regularizer (i.e., the mixture of the ℓ1- and ℓ2-norms) provides a balance between the two properties. However, existing deterministic methods have high per iteration computational complexities, making them inapplicable to large-scale problems. To address this issue, this paper proposes the first accelerated stochastic variance reduction gradient (RASVRG) algorithm for robust subspace clustering. We also introduce a new momentum acceleration technique for the RASVRG algorithm. As a result of the involvement of this momentum, the RASVRG algorithm achieves both the best oracle complexity and the fastest convergence rate, and it reaches higher efficiency in practice for both strongly convex and not strongly convex models. Various experimental results show that the RASVRG algorithm outperformed existing state-of-the-art methods with elastic net and ℓ1-norm regularizers in terms of accuracy in most cases. As demonstrated on real-world face datasets with different manually added levels of pixel corruption and occlusion situations, the RASVRG algorithm achieved much better performance in terms of accuracy and robustness.

An elastic competitive and discriminative collaborative representation method for image classification

Collaborative representation-based (CR) methods have become prevalent for pattern classification tasks, achieving formidable performance. Theoretically, we expect the learned class-specific representation of the correct class to be discriminative against others, with the representation of the correct class contributing dominantly in CR. However, most existing CR methods focus on improving discrimination while having a limited impact on enhancing the representation contribution of the correct category. In this work, we propose a novel CR approach for image classification called the elastic competitive and discriminative collaborative representation-based classifier (ECDCRC) to simultaneously strengthen representation contribution and discrimination of the correct class. The ECDCRC objective function penalizes two key terms by fully incorporating label information. The competitive term integrates the nearest subspace representation with corresponding elastic factors into the model, allowing each class to have varying competition intensities based on similarity with the query sample. This enhances the representation contribution of the correct class in CR. To further improve discrimination, the discriminative term introduces an elastic factor as a weight in the model to represent the gap between the query sample and the representation of each class. Moreover, instead of focusing on representation coefficients, the designed ECDCRC weights associated with representation components directly relate to the representation of each class, enabling more direct and precise discrimination improvement. Concurrently, sparsity is also enhanced through the two terms, further boosting model performance. Additionally, we propose a robust ECDCRC (R-ECDCRC) to handle image classification with noise. Extensive experiments on seven public databases demonstrate the proposed method's superior performance over related state-of-the-art CR methods.

Automated detection of myocardial infarction based on an improved state refinement module for LSTM/GRU

Myocardial infarction (MI) is a common cardiovascular disease caused by the blockages of coronary arteries. The visual inspection of electrocardiogram (ECG) is the main diagnosis pattern, while it is taxing and time-consuming. Motivated from state refinement module for long short term memory (SRM-LSTM), we proposed two improved state refinement frameworks based on LSTM and gated recurrent unit (GRU) called ISRM-LSTM and ISRM-GRU. Both are capable of adaptively refining current states of sample points in ECG with a message passing mechanism than existing LSTM. To evaluate the validity, both are installed into convolutional network architecture and standard LSTM, GRU and Residual networks are employed as control groups across the Physikalisch-Technische Bundesanstalt database. Empirical results confirm noticeable performance improvements than control groups and several existing algorithms with an accuracy of 99.1%. To our knowledge, both modules are the first attempt to consider the interaction characteristics into deep network and improve interpretability exhibiting considerable potentials on lightweight devices thanks to only utilization of three channel ECGs.

Positive Competitive Networks for Sparse Reconstruction

We propose and analyze a continuous-time firing-rate neural network, the positive firing-rate competitive network (PFCN), to tackle sparse reconstruction problems with non-negativity constraints. These problems, which involve approximating a given input stimulus from a dictionary using a set of sparse (active) neurons, play a key role in a wide range of domains, including, for example, neuroscience, signal processing, and machine learning. First, by leveraging the theory of proximal operators, we relate the equilibria of a family of continuous-time firing-rate neural networks to the optimal solutions of sparse reconstruction problems. Then we prove that the PFCN is a positive system and give rigorous conditions for the convergence to the equilibrium. Specifically, we show that the convergence depends only on a property of the dictionary and is linear-exponential in the sense that initially, the convergence rate is at worst linear and then, after a transient, becomes exponential. We also prove a number of technical results to assess the contractivity properties of the neural dynamics of interest. Our analysis leverages contraction theory to characterize the behavior of a family of firing-rate competitive networks for sparse reconstruction with and without non-negativity constraints. Finally, we validate the effectiveness of our approach via a numerical example.

Joint Metric Learning-Based Class-Specific Representation for Image Set Classification

With the rapid advances in digital imaging and communication technologies, recently image set classification has attracted significant attention and has been widely used in many real-world scenarios. As an effective technology, the class-specific representation theory-based methods have demonstrated their superior performances. However, this type of methods either only uses one gallery set to measure the gallery-to-probe set distance or ignores the inner connection between different metrics, leading to the learned distance metric lacking robustness, and is sensitive to the size of image sets. In this article, we propose a novel joint metric learning-based class-specific representation framework (JMLC), which can jointly learn the related and unrelated metrics. By iteratively modeling probe set and related or unrelated gallery sets as affine hull, we reconstruct this hull sparsely or collaboratively over another image set. With the obtained representation coefficients, the combined metric between the query set and the gallery set can then be calculated. In addition, we also derive the kernel extension of JMLC and propose two new unrelated set constituting strategies. Specifically, kernelized JMLC (KJMLC) embeds the gallery sets and probe sets into the high-dimensional Hilbert space, and in the kernel space, the data become approximately linear separable. Extensive experiments on seven benchmark databases show the superiority of the proposed methods to the state-of-the-art image set classifiers.

Knockoffs-SPR: Clean Sample Selection in Learning With Noisy Labels

A noisy training set usually leads to the degradation of the generalization and robustness of neural networks. In this article, we propose a novel theoretically guaranteed clean sample selection framework for learning with noisy labels. Specifically, we first present a Scalable Penalized Regression (SPR) method, to model the linear relation between network features and one-hot labels. In SPR, the clean data are identified by the zero mean-shift parameters solved in the regression model. We theoretically show that SPR can recover clean data under some conditions. Under general scenarios, the conditions may be no longer satisfied; and some noisy data are falsely selected as clean data. To solve this problem, we propose a data-adaptive method for Scalable Penalized Regression with Knockoff filters (Knockoffs-SPR), which is provable to control the False-Selection-Rate (FSR) in the selected clean data. To improve the efficiency, we further present a split algorithm that divides the whole training set into small pieces that can be solved in parallel to make the framework scalable to large datasets. While Knockoffs-SPR can be regarded as a sample selection module for a standard supervised training pipeline, we further combine it with a semi-supervised algorithm to exploit the support of noisy data as unlabeled data. Experimental results on several benchmark datasets and real-world noisy datasets show the effectiveness of our framework and validate the theoretical results of Knockoffs-SPR.

A transferable in-silico augmented ischemic model for virtual myocardial perfusion imaging and myocardial infarction detection

This paper proposes an innovative approach to generate a generalized myocardial ischemia database by modeling the virtual electrophysiology of the heart and the 12-lead electrocardiography projected by the in-silico model can serve as a ready-to-use database for automatic myocardial infarction/ischemia (MI) localization and classification. Although the virtual heart can be created by an established technique combining the cell model with personalized heart geometry to observe the spatial propagation of depolarization and repolarization waves, we developed a strategy based on the clinical pathophysiology of MI to generate a heterogeneous database with a generic heart while maintaining clinical relevance and reduced computational complexity. First, the virtual heart is simplified into 11 regions that match the types and locations, which can be diagnosed by 12-lead ECG; the major arteries were divided into 3-5 segments from the upstream to the downstream based on the general anatomy. Second, the stenosis or infarction of the major or minor coronary artery branches can cause different perfusion drops and infarct sizes. We simulated the ischemic sites in different branches of the arteries by meandering the infarction location to elaborate on possible ECG representations, which alters the infraction's size and changes the transmembrane potential (TMP) of the myocytes associated with different levels of perfusion drop. A total of 8190 different case combinations of cardiac potentials with ischemia and MI were simulated, and the corresponding ECGs were generated by forward calculations. Finally, we trained and validated our in-silico database with a sparse representation classification (SRC) and tested the transferability of the model on the real-world Physikalisch Technische Bundesanstalt (PTB) database. The overall accuracies for localizing the MI region on the PTB data achieved 0.86, which is only 2% drop compared to that derived from the simulated database (0.88). In summary, we have shown a proof-of-concept for transferring an in-silico model to real-world database to compensate for insufficient data.

Fuzzy Graph Subspace Convolutional Network

Graph convolutional networks (GCNs) are a popular approach to learn the feature embedding of graph-structured data, which has shown to be highly effective as well as efficient in performing node classification in an inductive way. However, with massive nongraph-organized data existing in application scenarios nowadays, it is critical to exploit the relationships behind the given groups of data, which makes better use of GCN and broadens the application field. In this article, we propose the f uzzy g raph s ubspace c onvolutional n etwork (FGSCN) to provide a brand-new paradigm for feature embedding and node classification with graph convolution (GC) when given an arbitrary collection of data. The FGSCN performs GC on the f uzzy s ubspace ( F -space), which simultaneously learns from the underlying subspace information in the low-dimensional space as well as its neighborliness information in the high-dimensional space. In particular, we construct the fuzzy homogenous graph GF on the F -space by fusing the homogenous graph of neighborliness GN and homogenous graph of subspace GS (defined by the affinity matrix of the low-rank representation). Here, it is proven that the GC on F -space will propagate both the local and global information through fuzzy set theory. We evaluated FGSCN on 15 unique datasets with different tasks (e.g., feature embedding, visual recognition, etc.). The experimental results showed that the proposed FGSCN has significant superiority compared with current state-of-the-art methods.

A low-cost machine learning framework for predicting drug-drug interactions based on fusion of multiple features and a parameter self-tuning strategy

Poly-drug therapy is now recognized as a crucial treatment, and the analysis of drug-drug interactions (DDIs) offers substantial theoretical support and guidance for its implementation. Predicting potential DDIs using intelligent algorithms is an emerging approach in pharmacological research. However, the existing supervised models and deep learning-based techniques still have several limitations. This paper proposes a novel DDI analysis and prediction framework called the Multi-View Semi-supervised Graph-based (MVSG) framework, which provides a comprehensive judgment by integrating multiple DDI features and functions without any time-consuming training process. Unlike conventional approaches, MVSG can search for the most suitable similarity (or distance) measurement among DDI data and construct graph structures for each feature. By employing a parameter self-tuning strategy, MVSG fuses multiple graphs according to the contributions of features' information. The actual anticancer drug data are extracted from the authoritative public database for evaluating the effectiveness of our framework, including 904 drugs, 7730 DDI records and 19 types of drug interactions. Validation results indicate that the prediction is more accurate when multiple features are adopted by our framework. In comparison to conventional machine learning techniques, MVSG can achieve higher performance even with less labeled data and without a training process. Finally, MVSG is employed to narrow down the search for potential valuable combinations.

Discriminative Regression With Adaptive Graph Diffusion

In this article, we propose a new linear regression (LR)-based multiclass classification method, called discriminative regression with adaptive graph diffusion (DRAGD). Different from existing graph embedding-based LR methods, DRAGD introduces a new graph learning and embedding term, which explores the high-order structure information between four tuples, rather than conventional sample pairs to learn an intrinsic graph. Moreover, DRAGD provides a new way to simultaneously capture the local geometric structure and representation structure of data in one term. To enhance the discriminability of the transformation matrix, a retargeted learning approach is introduced. As a result of combining the above-mentioned techniques, DRAGD can flexibly explore more unsupervised information underlying the data and the label information to obtain the most discriminative transformation matrix for multiclass classification tasks. Experimental results on six well-known real-world databases and a synthetic database demonstrate that DRAGD is superior to the state-of-the-art LR methods.

Vigilance estimation using truncated l1 distance kernel-based sparse representation regression with physiological signals

BACKGROUND

With a large number of accidents caused by the decline in the vigilance of operators, finding effective automatic vigilance monitoring methods is a work of great significance in recent years. Based on physiological signals and machine learning algorithms, researchers have opened up a path for objective vigilance estimation.

METHODS

Sparse representation (SR)-based recognition algorithms with excellent performance and simple models are very promising approaches in this field. This paper aims to study the adaptability and performance improvement of truncated l1 distance (TL1) kernel on SR-based algorithm in the context of physiological signal vigilance estimation. Compared with the traditional radial basis function (RBF), the TL1 kernel has good adaptiveness to nonlinearity and is suitable for the discrimination of complex physiological signals. A recognition framework based on TL1 and SR theory is proposed. Firstly, the inseparable physiological features are mapped to the reproducing kernel Kreĭn space through the infinite-dimensional projection of the TL1 kernel. Then the obtained kernel matrix is converted into the symmetric positive definite matrix according to the eigenspectrum approaches. Finally, the final prediction result is obtained through the sparse representation regression process.

RESULTS

We verified the performance of the proposed framework on the popular SEED-VIG dataset containing physiological signals (electroencephalogram and electrooculogram) associated with vigilance. In the experimental results, the TL1 kernel is superior to the RBF kernel in both performance and kernel parameter stability.

CONCLUSIONS

This demonstrates the effectiveness of the TL1 kernel in distinguishing physiological signals and the excellent vigilance estimation capability of the proposed framework. Moreover, the contribution of our research motivates the development of physiological signal recognition based on kernel methods.

Discriminative Dictionary Pair Learning With Scale-Constrained Structured Representation for Image Classification

The dictionary pair learning (DPL) model aims to design a synthesis dictionary and an analysis dictionary to accomplish the goal of rapid sample encoding. In this article, we propose a novel structured representation learning algorithm based on the DPL for image classification. It is referred to as discriminative DPL with scale-constrained structured representation (DPL-SCSR). The proposed DPL-SCSR utilizes the binary label matrix of dictionary atoms to project the representation into the corresponding label space of the training samples. By imposing a non-negative constraint, the learned representation adaptively approximates a block-diagonal structure. This innovative transformation is also capable of controlling the scale of the block-diagonal representation by enforcing the sum of within-class coefficients of each sample to 1, which means that the dictionary atoms of each class compete to represent the samples from the same class. This implies that the requirement of similarity preservation is considered from the perspective of the constraint on the sum of coefficients. More importantly, the DPL-SCSR does not need to design a classifier in the representation space as the label matrix of the dictionary can also be used as an efficient linear classifier. Finally, the DPL-SCSR imposes the l2,p -norm on the analysis dictionary to make the process of feature extraction more interpretable. The DPL-SCSR seamlessly incorporates the scale-constrained structured representation learning, within-class similarity preservation of representation, and the linear classifier into one regularization term, which dramatically reduces the complexity of training and parameter tuning. The experimental results on several popular image classification datasets show that our DPL-SCSR can deliver superior performance compared with the state-of-the-art (SOTA) dictionary learning methods. The MATLAB code of this article is available at https://github.com/chenzhe207/DPL-SCSR.

Flash-Based Computing-in-Memory Architecture to Implement High-Precision Sparse Coding

To address the concerns with power consumption and processing efficiency in big-size data processing, sparse coding in computing-in-memory (CIM) architectures is gaining much more attention. Here, a novel Flash-based CIM architecture is proposed to implement large-scale sparse coding, wherein various matrix weight training algorithms are verified. Then, with further optimizations of mapping methods and initialization conditions, the variation-sensitive training (VST) algorithm is designed to enhance the processing efficiency and accuracy of the applications of image reconstructions. Based on the comprehensive characterizations observed when considering the impacts of array variations, the experiment demonstrated that the trained dictionary could successfully reconstruct the images in a 55 nm flash memory array based on the proposed architecture, irrespective of current variations. The results indicate the feasibility of using Flash-based CIM architectures to implement high-precision sparse coding in a wide range of applications.

Adversarially Learning Occlusions by Backpropagation for Face Recognition

With the accomplishment of deep neural networks, face recognition methods have achieved great success in research and are now being applied at a human level. However, existing face recognition models fail to achieve state-of-the-art performance in recognizing occluded face images, which are common scenarios captured in the real world. One of the potential reasons for this is the lack of large-scale training datasets, requiring labour-intensive and costly labelling of the occlusions. To resolve these issues, we propose an Adversarially Learning Occlusions by Backpropagation (ALOB) model, a simple yet powerful double-network framework used to mitigate manual labelling by contrastively learning the corrupted features against personal identity labels, thereby maximizing the loss. To investigate the performance of the proposed method, we compared our model to the existing state-of-the-art methods, which function under the supervision of occlusion learning, in various experiments. Extensive experimentation on LFW, AR, MFR2, and other synthetic masked or occluded datasets confirmed the effectiveness of the proposed model in occluded face recognition by sustaining better results in terms of masked face recognition and general face recognition. For the AR datasets, the ALOB model outperformed other advanced methods by obtaining a 100% recognition rate for images with sunglasses (protocols 1 and 2). We also achieved the highest accuracies of 94.87%, 92.05%, 78.93%, and 71.57% TAR@FAR = 1 × 10-3 in LFW-OCC-2.0 and LFW-OCC-3.0, respectively. Furthermore, the proposed method generalizes well in terms of FR and MFR, yielding superior results in three datasets, LFW, LFW-Masked, and MFR2, and producing accuracies of 98.77%, 97.62%, and 93.76%, respectively.

Learning an Invariant and Equivariant Network for Weakly Supervised Object Detection

Weakly Supervised Object Detection (WSOD) is of increasing importance in the community of computer vision as its extensive applications and low manual cost. Most of the advanced WSOD approaches build upon an indefinite and quality-agnostic framework, leading to unstable and incomplete object detectors. This paper attributes these issues to the process of inconsistent learning for object variations and the unawareness of localization quality and constructs a novel end-to-end Invariant and Equivariant Network (IENet). It is implemented with a flexible multi-branch online refinement, to be naturally more comprehensive-perceptive against various objects. Specifically, IENet first performs label propagation from the predicted instances to their transformed ones in a progressive manner, achieving affine-invariant learning. Meanwhile, IENet also naturally utilizes rotation-equivariant learning as a pretext task and derives an instance-level rotation-equivariant branch to be aware of the localization quality. With affine-invariance learning and rotation-equivariant learning, IENet urges consistent and holistic feature learning for WSOD without additional annotations. On the challenging datasets of both natural scenes and aerial scenes, we substantially boost WSOD to new state-of-the-art performance. The codes have been released at: https://github.com/XiaoxFeng/IENet.

Incremental Learning for Online Data Using QR Factorization on Convolutional Neural Networks

Catastrophic forgetting, which means a rapid forgetting of learned representations while learning new data/samples, is one of the main problems of deep neural networks. In this paper, we propose a novel incremental learning framework that can address the forgetting problem by learning new incoming data in an online manner. We develop a new incremental learning framework that can learn extra data or new classes with less catastrophic forgetting. We adopt the hippocampal memory process to the deep neural networks by defining the effective maximum of neural activation and its boundary to represent a feature distribution. In addition, we incorporate incremental QR factorization into the deep neural networks to learn new data with both existing labels and new labels with less forgetting. The QR factorization can provide the accurate subspace prior, and incremental QR factorization can reasonably express the collaboration between new data with both existing classes and new class with less forgetting. In our framework, a set of appropriate features (i.e., nodes) provides improved representation for each class. We apply our method to the convolutional neural network (CNN) for learning Cifar-100 and Cifar-10 datasets. The experimental results show that the proposed method efficiently alleviates the stability and plasticity dilemma in the deep neural networks by providing the performance stability of a trained network while effectively learning unseen data and additional new classes.

Motor imagery classification using sparse representations: an exploratory study

The non-stationary nature of the EEG signal poses challenges for the classification of motor imagery. sparse representation classification (SRC) appears as an alternative for classification of untrained conditions and, therefore, useful in motor imagery. Empirical mode decomposition (EMD) deals with signals of this nature and appears at the rear of the classification, supporting the generation of features. In this work we evaluate the combination of these methods in a multiclass classification problem, comparing them with a conventional method in order to determine if their performance is regular. For comparison with SRC we use multilayer perceptron (MLP). We also evaluated a hybrid approach for classification of sparse representations with MLP (RSMLP). For comparison with EMD we used filtering by frequency bands. Feature selection methods were used to select the most significant ones, specifically Random Forest and Particle Swarm Optimization. Finally, we used data augmentation to get a more voluminous base. Regarding the first dataset, we observed that the classifiers that use sparse representation have results equivalent to each other, but they outperform the conventional MLP model. SRC and SRMLP achieve an average accuracy of [Formula: see text] and [Formula: see text] respectively while the MLP is [Formula: see text], representing a gain between [Formula: see text] and [Formula: see text]. The use of EMD in relation to other feature processing techniques is not superior. However, EMD does not influence negatively, there is an opportunity for improvement. Finally, the use of data augmentation proved to be important to obtain relevant results. In the second dataset, we did not observe the same results. Models based on sparse representation (SRC, SRMLP, etc.) have on average a performance close to other conventional models, but without surpassing them. The best sparse models achieve an average accuracy of [Formula: see text] among the subjects in the base, while other model reach [Formula: see text]. The improvement of self-adaptive mechanisms that respond efficiently to the user's context is a good way to achieve improvements in motor imagery applications. However, other scenarios should be investigated, since the advantage of these methods was not proven in all datasets studied. There is still room for improvement, such as optimizing the dictionary of sparse representation in the context of motor imagery. Investing efforts in synthetically increasing the training base has also proved important to reduce the costs of this group of applications.

Balance guided incomplete multi-view spectral clustering

There is a large volume of incomplete multi-view data in the real-world. How to partition these incomplete multi-view data is an urgent realistic problem since almost all of the conventional multi-view clustering methods are inapplicable to cases with missing views. In this paper, a novel graph learning-based incomplete multi-view clustering (IMVC) method is proposed to address this issue. Different from existing works, our method aims at learning a common consensus graph from all incomplete views and obtaining a clustering indicator matrix in a unified framework. To achieve a stable clustering result, a relaxed spectral clustering model is introduced to obtain a probability consensus representation with all positive elements that reflect the data clustering result. Considering the different contributions of views to the clustering task, a weighted multi-view learning mechanism is introduced to automatically balance the effects of different views in model optimization. In this way, the intrinsic information of the incomplete multi-view data can be fully exploited. The experiments on several incomplete multi-view datasets show that our method outperforms the compared state-of-the-art clustering methods, which demonstrates the effectiveness of our method for IMVC.

Sequential sparse autoencoder for dynamic heading representation in ventral intraparietal area

To navigate in space, it is important to predict headings in real-time from neural responses in the brain to vestibular and visual signals, and the ventral intraparietal area (VIP) is one of the critical brain areas. However, it remains unexplored in the population level how the heading perception is represented in VIP. And there are no commonly used methods suitable for decoding the headings from the population responses in VIP, given the large spatiotemporal dynamics and heterogeneity in the neural responses. Here, responses were recorded from 210 VIP neurons in three rhesus monkeys when they were performing a heading perception task. And by specifically and separately modelling the both dynamics with sparse representation, we built a sequential sparse autoencoder (SSAE) to do the population decoding on the recorded dataset and tried to maximize the decoding performance. The SSAE relies on a three-layer sparse autoencoder to extract temporal and spatial heading features in the dataset via unsupervised learning, and a softmax classifier to decode the headings. Compared with other population decoding methods, the SSAE achieves a leading accuracy of 96.8% ± 2.1%, and shows the advantages of robustness, low storage and computing burden for real-time prediction. Therefore, our SSAE model performs well in learning neurobiologically plausible features comprising dynamic navigational information.

Facial expression recognition on partially occluded faces using component based ensemble stacked CNN

Facial Expression Recognition (FER) is the basis for many applications including human-computer interaction and surveillance. While developing such applications, it is imperative to understand human emotions for better interaction with machines. Among many FER models developed so far, Ensemble Stacked Convolution Neural Networks (ES-CNN) showed an empirical impact in improving the performance of FER on static images. However, the existing ES-CNN based FER models trained with features extracted from the entire face, are unable to address the issues of ambient parameters such as pose, illumination, occlusions. To mitigate the problem of reduced performance of ES-CNN on partially occluded faces, a Component based ES-CNN (CES-CNN) is proposed. CES-CNN applies ES-CNN on action units of individual face components such as eyes, eyebrows, nose, cheek, mouth, and glabella as one subnet of the network. Max-Voting based ensemble classifier is used to ensemble the decisions of the subnets in order to obtain the optimized recognition accuracy. The proposed CES-CNN is validated by conducting experiments on benchmark datasets and the performance is compared with the state-of-the-art models. It is observed from the experimental results that the proposed model has a significant enhancement in the recognition accuracy compared to the existing models.

Identification of Perna viridis contaminated with diarrhetic shellfish poisoning toxins in vitro using NIRS and a discriminative non-negative representation-based classifier

Diarrhetic shellfish poisoning (DSP) toxins are one of the most widespread marine biotoxins that affect aquaculture and human health, and their detection has become crucial. In this study, near-infrared reflectance spectroscopy (NIRS) with non-destructive characteristics was used to identify DSP toxins in Perna viridis. The spectral data of the DSP toxin-contaminated and non-contaminated Perna viridis samples were acquired in the 950-1700 nm range. To solve the discrimination of spectra with crossover and overlapping, a discriminative non-negative representation-based classifier (DNRC) has been proposed. Compared with collaborative and non-negative representation-based classifiers, the DNRC model exhibited better performance in detecting DSP toxins, with a classification accuracy of 99.44 %. For a relatively small-scale sample dataset in practical applications, the performance of the DNRC model was compared with those of classical models. The DNRC model achieved the best results for both identification accuracy and F-measure, and its detection performance did not significantly decrease with decreasing sample size. The experimental results validated that a combination of NIRS and the DNRC model can facilitate rapid, convenient, and non-destructive detection of DSP toxins in Perna viridis.

Lung cancer lesion detection in histopathology images using graph-based sparse PCA network

Early detection of lung cancer is critical for improvement of patient survival. To address the clinical need for efficacious treatments, genetically engineered mouse models (GEMM) have become integral in identifying and evaluating the molecular underpinnings of this complex disease that may be exploited as therapeutic targets. Assessment of GEMM tumor burden on histopathological sections performed by manual inspection is both time consuming and prone to subjective bias. Therefore, an interplay of needs and challenges exists for computer-aided diagnostic tools, for accurate and efficient analysis of these histopathology images. In this paper, we propose a simple machine learning approach called the graph-based sparse principal component analysis (GS-PCA) network, for automated detection of cancerous lesions on histological lung slides stained by hematoxylin and eosin (H&E). Our method comprises four steps: 1) cascaded graph-based sparse PCA, 2) PCA binary hashing, 3) block-wise histograms, and 4) support vector machine (SVM) classification. In our proposed architecture, graph-based sparse PCA is employed to learn the filter banks of the multiple stages of a convolutional network. This is followed by PCA hashing and block histograms for indexing and pooling. The meaningful features extracted from this GS-PCA are then fed to an SVM classifier. We evaluate the performance of the proposed algorithm on H&E slides obtained from an inducible K-ras^G12D lung cancer mouse model using precision/recall rates, F_β-score, Tanimoto coefficient, and area under the curve (AUC) of the receiver operator characteristic (ROC) and show that our algorithm is efficient and provides improved detection accuracy compared to existing algorithms.

Human Behavior Recognition via Hierarchical Patches Descriptor and Approximate Locality-Constrained Linear Coding

Human behavior recognition technology is widely adopted in intelligent surveillance, human-machine interaction, video retrieval, and ambient intelligence applications. To achieve efficient and accurate human behavior recognition, a unique approach based on the hierarchical patches descriptor (HPD) and approximate locality-constrained linear coding (ALLC) algorithm is proposed. The HPD is a detailed local feature description, and ALLC is a fast coding method, which makes it more computationally efficient than some competitive feature-coding methods. Firstly, energy image species were calculated to describe human behavior in a global manner. Secondly, an HPD was constructed to describe human behaviors in detail through the spatial pyramid matching method. Finally, ALLC was employed to encode the patches of each level, and a feature coding with good structural characteristics and local sparsity smoothness was obtained for recognition. The recognition experimental results on both Weizmann and DHA datasets demonstrated that the accuracy of five energy image species combined with HPD and ALLC was relatively high, scoring 100% in motion history image (MHI), 98.77% in motion energy image (MEI), 93.28% in average motion energy image (AMEI), 94.68% in enhanced motion energy image (EMEI), and 95.62% in motion entropy image (MEnI).

Joint feature selection and optimal bipartite graph learning for subspace clustering

Recently, there has been tremendous interest in developing graph-based subspace clustering in high-dimensional data, which does not require a priori knowledge of the number of dimensions and subspaces. The general steps of such algorithms are dictionary representation and spectral clustering. Traditional methods use the dataset itself as a dictionary when performing dictionary representation. There are some limitations that the redundant information present in the dictionary and features may make the constructed graph structure unclear and require post-processing to obtain labels. To address these problems, we propose a novel subspace clustering model that first introduces feature selection to process the input data, randomly selects some samples to construct a dictionary to remove redundant information and learns the optimal bipartite graph with K-connected components under the constraint of the (normalized) Laplacian rank. Finally, the labels are obtained directly from the graphs. The experimental results on motion segmentation and face recognition datasets demonstrate the superior effectiveness and stability of our algorithm.

Multitask Feature Learning as Multiobjective Optimization: A New Genetic Programming Approach to Image Classification

Feature learning is a promising approach to image classification. However, it is difficult due to high image variations. When the training data are small, it becomes even more challenging, due to the risk of overfitting. Multitask feature learning has shown the potential for improving generalization. However, existing methods are not effective for handling the case that multiple tasks are partially conflicting. Therefore, for the first time, this article proposes to solve a multitask feature learning problem as a multiobjective optimization problem by developing a genetic programming approach with a new representation to image classification. In the new approach, all the tasks share the same solution space and each solution is evaluated on multiple tasks so that the objectives of all the tasks can be optimized simultaneously using a single population. To learn effective features, a new and compact program representation is developed to allow the new approach to evolving solutions shared across tasks. The new approach can automatically find a diverse set of nondominated solutions that achieve good tradeoffs between different tasks. To further reduce the risk of overfitting, an ensemble is created by selecting nondominated solutions to solve each image classification task. The results show that the new approach significantly outperforms a large number of benchmark methods on six problems consisting of 15 image classification datasets of varying difficulty. Further analysis shows that these new designs are effective for improving the performance. The detailed analysis clearly reveals the benefits of solving multitask feature learning as multiobjective optimization in improving the generalization.

Structured Sparsity Optimization With Non-Convex Surrogates of l2,0-Norm: A Unified Algorithmic Framework

In this article, we present a general optimization framework that leverages structured sparsity to achieve superior recovery results. The traditional method for solving the structured sparse objectives based on l_2,0-norm is to use the l_2,1-norm as a convex surrogate. However, such an approximation often yields a large performance gap. To tackle this issue, we first provide a framework that allows for a wide range of surrogate functions (including non-convex surrogates), which exhibits better performance in harnessing structured sparsity. Moreover, we develop a fixed point algorithm that solves a key underlying non-convex structured sparse recovery optimization problem to global optimality with a guaranteed super-linear convergence rate. Building on this, we consider three specific applications, i.e., outlier pursuit, supervised feature selection, and structured dictionary learning, which can benefit from the proposed structured sparsity optimization framework. In each application, how the optimization problem can be formulated and thus be relaxed under a generic surrogate function is explained in detail. We conduct extensive experiments on both synthetic and real-world data and demonstrate the effectiveness and efficiency of the proposed framework.

Identification of DNA-binding proteins by Kernel Sparse Representation via L2,1-matrix norm

An understanding of DNA-binding proteins is helpful in exploring the role that proteins play in cell biology. Furthermore, the prediction of DNA-binding proteins is essential for the chemical modification and structural composition of DNA, and is of great importance in protein functional analysis and drug design. In recent years, DNA-binding protein prediction has typically used machine learning-based methods. The prediction accuracy of various classifiers has improved considerably, but researchers continue to spend time and effort on improving prediction performance. In this paper, we combine protein sequence evolutionary information with a classification method based on kernel sparse representation for the prediction of DNA-binding proteins, and based on the field of machine learning, a model for the identification of DNA-binding proteins by sequence information was finally proposed. Based on the confirmation of the final experimental results, we achieved good prediction accuracy on both the PDB1075 and PDB186 datasets. Our training result for cross-validation on PDB1075 was 81.37%, and our independent test result on PDB186 was 83.9%, both of which outperformed the other methods to some extent. Therefore, the proposed method in this paper is proven to be effective and feasible for predicting DNA-binding proteins.

Energy-Aware IoT-Based Method for a Hybrid On-Wrist Fall Detection System Using a Supervised Dictionary Learning Technique

In recent decades, falls have posed multiple critical health issues, especially for the older population, with their emerging growth. Recent research has shown that a wrist-based fall detection system offers an accessory-like comfortable solution for Internet of Things (IoT)-based monitoring. Nevertheless, an autonomous device for anywhere-anytime may present an energy consumption concern. Hence, this paper proposes a novel energy-aware IoT-based architecture for Message Queuing Telemetry Transport (MQTT)-based gateway-less monitoring for wearable fall detection. Accordingly, a hybrid double prediction technique based on Supervised Dictionary Learning was implemented to reinforce the detection efficiency of our previous works. A controlled dataset was collected for training (offline), while a real set of measurements of the proposed system was used for validation (online). It achieved a noteworthy offline and online detection performance of 99.8% and 91%, respectively, overpassing most of the related works using only an accelerometer. In the worst case, the system showed a battery consumption optimization by a minimum of 27.32 working hours, significantly higher than other research prototypes. The approach presented here proves to be promising for real applications, which require a reliable and long-term anywhere-anytime solution.

Robust Semisupervised Deep Generative Model Under Compound Noise

Semisupervised learning has been widely applied to deep generative model such as variational autoencoder. However, there are still limited work in noise-robust semisupervised deep generative model where the noise exists in both of the data and the labels simultaneously, which are referred to as outliers and noisy labels or compound noise. In this article, we propose a novel noise-robust semisupervised deep generative model by jointly tackling the noisy labels and outliers in a unified robust semisupervised variational autoencoder randomized generative adversarial network (URSVAE-GAN). Typically, we consider the uncertainty of the information of the input data in order to enhance the robustness of the variational encoder toward the noisy data in our unified robust semisupervised variational autoencoder (URSVAE). Subsequently, in order to alleviate the detrimental effects of noisy labels, a denoising layer is integrated naturally into the semisupervised variational autoencoder so that the variational inference is conditioned on the corrected labels. Moreover, to enhance the robustness of the variational inference in the presence of outliers, the robust β -divergence measure is employed to derive the novel variational lower bound, which already achieves competitive performance. This further motivates the development of URSVAE-GAN that collapses the decoder of URSVAE and the generator of a robust semisupervised generative adversarial network into one unit. By applying the end-to-end denoising scheme in the joint optimization, the experimental results demonstrate the superiority of the proposed framework by the evaluating on image classification and face recognition tasks and comparing with the state-of-the-art approaches.

A Sparse Representation Classification Scheme for the Recognition of Affective and Cognitive Brain Processes in Neuromarketing

In this work, we propose a novel framework to recognize the cognitive and affective processes of the brain during neuromarketing-based stimuli using EEG signals. The most crucial component of our approach is the proposed classification algorithm that is based on a sparse representation classification scheme. The basic assumption of our approach is that EEG features from a cognitive or affective process lie on a linear subspace. Hence, a test brain signal can be represented as a linear (or weighted) combination of brain signals from all classes in the training set. The class membership of the brain signals is determined by adopting the Sparse Bayesian Framework with graph-based priors over the weights of linear combination. Furthermore, the classification rule is constructed by using the residuals of linear combination. The experiments on a publicly available neuromarketing EEG dataset demonstrate the usefulness of our approach. For the two classification tasks offered by the employed dataset, namely affective state recognition and cognitive state recognition, the proposed classification scheme manages to achieve a higher classification accuracy compared to the baseline and state-of-the art methods (more than 8% improvement in classification accuracy).

Quantum face recognition protocol with ghost imaging

Face recognition is one of the most ubiquitous examples of pattern recognition in machine learning, with numerous applications in security, access control, and law enforcement, among many others. Pattern recognition with classical algorithms requires significant computational resources, especially when dealing with high-resolution images in an extensive database. Quantum algorithms have been shown to improve the efficiency and speed of many computational tasks, and as such, they could also potentially improve the complexity of the face recognition process. Here, we propose a quantum machine learning algorithm for pattern recognition based on quantum principal component analysis, and quantum independent component analysis. A novel quantum algorithm for finding dissimilarity in the faces based on the computation of trace and determinant of a matrix (image) is also proposed. The overall complexity of our pattern recognition algorithm is [Formula: see text]-N is the image dimension. As an input to these pattern recognition algorithms, we consider experimental images obtained from quantum imaging techniques with correlated photons, e.g. "interaction-free" imaging or "ghost" imaging. Interfacing these imaging techniques with our quantum pattern recognition processor provides input images that possess a better signal-to-noise ratio, lower exposures, and higher resolution, thus speeding up the machine learning process further. Our fully quantum pattern recognition system with quantum algorithm and quantum inputs promises a much-improved image acquisition and identification system with potential applications extending beyond face recognition, e.g., in medical imaging for diagnosing sensitive tissues or biology for protein identification.

DisP+V: A Unified Framework for Disentangling Prototype and Variation From Single Sample per Person

Single sample per person face recognition (SSPP FR) is one of the most challenging problems in FR due to the extreme lack of enrolment data. To date, the most popular SSPP FR methods are the generic learning methods, which recognize query face images based on the so-called prototype plus variation (i.e., P+V) model. However, the classic P+V model suffers from two major limitations: 1) it linearly combines the prototype and variation images in the observational pixel-spatial space and cannot generalize to multiple nonlinear variations, e.g., poses, which are common in face images and 2) it would be severely impaired once the enrolment face images are contaminated by nuisance variations. To address the two limitations, it is desirable to disentangle the prototype and variation in a latent feature space and to manipulate the images in a semantic manner. To this end, we propose a novel disentangled prototype plus variation model, dubbed DisP+V, which consists of an encoder-decoder generator and two discriminators. The generator and discriminators play two adversarial games such that the generator nonlinearly encodes the images into a latent semantic space, where the more discriminative prototype feature and the less discriminative variation feature are disentangled. Meanwhile, the prototype and variation features can guide the generator to generate an identity-preserved prototype and the corresponding variation, respectively. Experiments on various real-world face datasets demonstrate the superiority of our DisP+V model over the classic P+V model for SSPP FR. Furthermore, DisP+V demonstrates its unique characteristics in both prototype recovery and face editing/interpolation.

An Unsupervised Feature Extraction Approach Based on Self-Expression

Feature extraction algorithms lack good interpretability during the projection learning. To solve this problem, an unsupervised feature extraction algorithm, that is, block diagonal projection (BDP), based on self-expression is proposed. Specifically, if the original data are projected into a low-dimensional subspace by a feature extraction algorithm, although the data may be more compact, the new features obtained may not be as explanatory as the original sample features. Therefore, by imposing L2,1 norm constraint on the projection matrix, the projection matrix can be of row sparsity. On one hand, discriminative features can be selected to make the projection matrix to be more interpretable. On the other hand, irrelevant or redundant features can be suppressed. The proposed model integrates feature extraction and selection into one framework. In addition, since self-expression can well excavate the correlation between samples or sample features, the unsupervised feature extraction task can be better guided using this property between them. At the same time, the block diagonal representation regular term is introduced to directly pursue the block diagonal representation. Thus, the accuracy of pattern recognition tasks such as clustering and classification can be improved. Finally, the effectiveness of BDP in linear dimensionality reduction and classification is proved on various reference datasets. The experimental results show that this algorithm is superior to previous feature extraction counterparts.

Learning an Optimal Bipartite Graph for Subspace Clustering via Constrained Laplacian Rank

In this article, we focus on utilizing the idea of co-clustering algorithms to address the subspace clustering problem. In recent years, co-clustering methods have been developed greatly with many important applications, such as document clustering and gene expression analysis. Different from the traditional graph-based methods, co-clustering can utilize the bipartite graph to extract the duality relationship between samples and features. It means that the bipartite graph can obtain more information than other traditional graph methods. Therefore, we proposed a novel method to handle the subspace clustering problem by combining dictionary learning with a bipartite graph under the constraint of the (normalized) Laplacian rank. Besides, to avoid the effect of redundant information hiding in the data, the original data matrix is not used as the static dictionary in our model. By updating the dictionary matrix under the sparse constraint, we can obtain a better coefficient matrix to construct the bipartite graph. Based on Theorem 2 and Lemma 1, we further speed up our algorithm. Experimental results on both synthetic and benchmark datasets demonstrate the superior effectiveness and stability of our model.

Understanding metabolic alterations after SARS-CoV-2 infection: insights from the patients' oral microenvironmental metabolites

BACKGROUND

Coronavirus disease 2019 is a type of acute infectious pneumonia and frequently confused with influenza since the initial symptoms. When the virus colonized the patient's mouth, it will cause changes of the oral microenvironment. However, few studies on the alterations of metabolism of the oral microenvironment affected by SARS-CoV-2 infection have been reported. In this study, we explored metabolic alterations of oral microenvironment after SARS-CoV-2 infection.

METHODS

Untargeted metabolomics (UPLC-MS) was used to investigate the metabolic changes between oral secretion samples of 25 COVID-19 and 30 control participants. To obtain the specific metabolic changes of COVID-19, we selected 25 influenza patients to exclude the metabolic changes caused by the stress response of the immune system to the virus. Multivariate analysis (PCA and PLS-DA plots) and univariate analysis (students' t-test) were used to compare the differences between COVID-19 patients and the controls. Online hiplot tool was used to perform heatmap analysis. Metabolic pathway analysis was conducted by using the MetaboAnalyst 5.0 web application.

RESULTS

PLS-DA plots showed significant separation of COVID-19 patients and the controls. A total of 45 differential metabolites between COVID-19 and control group were identified. Among them, 35 metabolites were defined as SARS-CoV-2 specific differential metabolites. Especially, the levels of cis-5,8,11,14,17-eicosapentaenoic acid and hexanoic acid changed dramatically based on the FC values. Pathway enrichment found the most significant pathways were tyrosine-related metabolism. Further, we found 10 differential metabolites caused by the virus indicating the body's metabolism changes after viral stimulation. Moreover, adenine and adenosine were defined as influenza virus-specific differential metabolites.

CONCLUSIONS

This study revealed that 35 metabolites and tyrosine-related metabolism pathways were significantly changed after SARS-CoV-2 infection. The metabolic alterations of oral microenvironment in COVID-19 provided new insights into its molecular mechanisms for research and prognostic treatment.