Abstract Motivation Deep neural network (DNN) algorithms were utilized in predicting various biomedical phenotypes recently, and demonstrated very good prediction performances without selecting features. This study proposed a hypothesis that the DNN models may be further improved by feature selection algorithms. Results A comprehensive comparative was carried out evaluating 11 on three conventional algorithms, i.e. convolution (CNN), deep belief (DBN) recurrent (RNN), recent DNNs,...

Graph-based semi-supervised node classification (GraphSSC) has wide applications, ranging from networking and security to data mining machine learning, etc. However, existing centralized GraphSSC methods are impractical solve many real-world graph-based problems, as collecting the entire graph labeling a reasonable number of labels is time-consuming costly, privacy may be also violated. Federated learning (FL) an emerging paradigm that enables collaborative among multiple clients, which can...

This article studies an emerging practical problem called heterogeneous prototype learning (HPL). Unlike the conventional face synthesis (HFS) that focuses on precisely translating a image from source domain to another target one without removing facial variations, HPL aims at variation-free of in while preserving identity characteristics. is compounded involving two cross-coupled subproblems, is, transfer and (PL), thus making most existing HFS methods simply style images unsuitable for...

The early diagnosis of Parkinson’s disease (PD) is crucial for potential patients to receive timely treatment and prevent progression. Recent studies have shown that PD closely linked impairments in facial muscle control, resulting characteristic “masked face” symptoms. This discovery offers a novel perspective by leveraging expression recognition analysis techniques capture quantify these features, thereby distinguishing between non-PD individuals based on their expressions. However,...

Parkinson's disease (PD) is a common degenerative of the nervous system in elderly. The early diagnosis PD very important for potential patients to receive prompt treatment and avoid aggravation disease. Recent studies have found that always suffer from emotional expression disorder, thus forming characteristics "masked faces". Based on this, we propose an auto method based mixed facial expressions paper. Specifically, proposed cast into four steps: Firstly, synthesize virtual face images...

Single sample per person face recognition (SSPP FR), i.e., identifying a (i.e., data subject) with single image only for training, has several attractive potential applications, but it is still challenging problem. Existing generic learning methods usually leverage prototype plus variation (P+V) model SSPP FR provided that samples in the biometric enrolment database are variation-free and thus can be treated as prototypes of subjects. However, this condition not satisfied when these...

This article focuses on a new and practical problem in single-sample per person face recognition (SSPP FR), i.e., SSPP FR with contaminated biometric enrolment database (SSPP-ce where the SSPP-based is by nuisance facial variations wild, such as poor lightings, expression change, disguises (e.g., wearing sunglasses, hat, scarf). In SSPP-ce FR, most popular generic learning methods will suffer serious performance degradation because prototype plus variation (P+V) model used these no longer...

Single sample per person (SSPP) face recognition with a contaminated biometric enrolment database (SSPP-ce FR) is an emerging practical FR problem, where the SSPP in no longer standard but by nuisance facial variations such as expression, lighting, pose, and disguise. In this case, conventional methods, including patch-based generic learning will suffer from serious performance degradation. Few recent methods were proposed to tackle SSPP-ce either performing prototype on or discriminative...

Graph neural networks (GNNs) have achieved state-of-the-art performance in many graph learning tasks. However, recent studies show that GNNs are vulnerable to both test-time evasion and training-time poisoning attacks perturb the structure. While existing attack methods shown promising performance, we would like design an framework further enhance performance. In particular, our is inspired by certified robustness, which was originally used defenders defend against adversarial attacks. We...

Existing heterogeneous face synthesis (HFS) methods focus on performing accurate image-to-image translation across domains, while they cannot effectively remove the nuisance facial variations such as poses, expressions or occlusions. To address challenges, this paper studies a new practical prototype learning (HPL) problem. be specific, given image contaminated by from source domain, HPL aims to reconstruct variation-free in specified target domain. tackle HPL, we propose unified and...

The key ingredients of matrix factorization lie in basic learning and coefficient representation. To enhance the discriminant ability learned basis, graph embedding is usually introduced model. However, existing methods based on generally conduct analysis via a single type adjacency graph, either similarity-based graphs (e.g., Laplacian eigenmaps graph) or reconstruction-based L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -graph),...

Model quantization has drawn much attention for federated learning (FL) over the Internet of Things (IoT) since it is an effective way to address critical bottleneck communication efficiency. State-of-the-art studies have generally assumed homogeneous model quantization, where all clients' updates are quantized using same number bits and aggregated with weight at server. However, in practical IoT scenarios, various devices may apply heterogeneous due their different hardware capabilities,...

Point cloud registration is a fundamental problem in many applications. The point based on local shape descriptor has been widely researched. In order to further improve the performance of registration, novel method proposed this paper. First, binary designed establish correspondences between two clouds. high descriptiveness. Thus, more correct are established. Then, 3D transformation estimation technique developed, which multiple constraints used accelerate computation. When randomly...

For generative learning tasks, there are three crucial criteria for generating samples from the models: quality, coverage/diversity, and sampling speed. Among existing models, Generative adversarial networks (GANs) diffusion models demonstrate outstanding quality performance while suffering notable limitations. GANs can generate high-quality results enable fast sampling, their drawbacks, however, lie in limited diversity of generated samples. On other hand, excel at with a commendable...

Different kernels cause various class discriminations owing to their different geometrical structures of the data in feature space. In this paper, a method kernel optimization by maximizing measure separability empirical space with sparse representation-based classifier (SRC) is proposed solve problem automatically choosing functions and parameters learning. The first adopts so-called data-dependent generate an efficient algorithm. Then, constrained function using general gradient descent...

Single sample per person face recognition (SSPP FR) is one of the most challenging problems in FR due to extreme lack enrolment data. To date, popular SSPP methods are generic learning methods, which recognize query images based on so-called prototype plus variation (i.e., P+V) model. However, classic P+V model suffers from two major limitations: 1) it linearly combines and observational pixel-spatial space cannot generalize multiple nonlinear variations, e.g., poses, common 2) would be...

Most off-the-shelf subspace learning methods directly calculate the statistical characteristics of original input images, while ignoring different contributions image components. In fact, to extract efficient features for analysis, noise or trivial structure in images should have little contribution and intrinsic be uncovered. Motivated by this observation, we propose a new method, namely, discriminant manifold via sparse coding (DML_SC) robust feature extraction. Specifically, first...

In this paper, we propose a sparseness constraint NMF method, named graph regularized matrix factorization with sparse coding (GRNMF_SC). By combining manifold learning and techniques together, GRNMF_SC can efficiently extract the basic vectors from data space, which preserves intrinsic structure also local features of original data. The target function our method is easy to propose, while solving procedures are really nontrivial; in paper gave detailed derivation strict proof its...