Search

Federated Learning (FL) enables the multiple participating devices to collaboratively contribute to a global neural network model while keeping the training data locally. Unlike the centralized training setting, the non-IID, imbalanced (statistical heterogeneity) and distribution shifted training data of FL is distributed in the federated network, which will increase the divergences between the local models and the global model, further degrading performance. In this paper, we propose a flexible clustered federated learning (CFL) framework named FlexCFL, in which we 1) group the training of clients based on the similarities between the clients’ optimization directions for lower training divergence; 2) implement an efficient newcomer device cold start mechanism for framework scalability and practicality; 3) flexibly migrate clients to meet the challenge of client-level data distribution shift. FlexCFL can achieve improvements by dividing joint optimization into groups of sub-optimization and can strike a balance between accuracy and communication efficiency in the distribution shift environment. The convergence and complexity are analyzed to demonstrate the efficiency of FlexCFL. We also evaluate FlexCFL on several open datasets and made comparisons with related CFL frameworks. The results show that FlexCFL can significantly improve absolute test accuracy by $+10.6\%$ + 10. 6 % on FEMNIST compared with FedAvg, $+3.5\%$ + 3. 5 % on FashionMNIST compared with FedProx, $+8.4\%$ + 8. 4 % on MNIST compared with FeSEM, $+4.7\%$ + 4. 7 % on Sentiment140 compare with IFCA. The experiment results show that FlexCFL is also communication efficient in the distribution shift environment. [ABSTRACT FROM AUTHOR]

In recent years, blockchain technology has been applied in the educational domain because of its salient advantages, i.e., transparency, decentralization, and immutability. Available systems typically use public blockchain networks such as Ethereum and Bitcoin to store learning results. However, the cost of writing data on these networks is significant, making educational institutions limit data sent to the target network, typically containing only hash codes of the issued certificates. In this paper, we present a system based on a private blockchain network for lifelong learning data authentication and management named B4E (Blockchain For Education). B4E stores not only certificates but also learners' training data such as transcripts and educational programs in order to create a complete record of the lifelong education of each user and verify certificates that they have obtained. As a result, B4E can address two types of fake certificates, i.e., certificates printed by unlawful organizations and certificates issued by educational institutions for learners who have not met the training requirements. In addition, B4E is designed to allow all participants to easily deploy software packages to manage, share, and check stored information without depending on a single point of access. As such, the system enhances the transparency and reliability of the stored data. Our experiments show that B4E meets expectations for deployment in reality. [ABSTRACT FROM AUTHOR]

Radar-based gait recognition has attracted growing interest due to radar’s benefit of robust to harsh environments, and active researchers have recently studied the feasibility of semisupervised radar-based tasks to reduce the cost of labeling. However, in order to obtain sufficient radar data to meet the requirements of semisupervised learning and to obtain good generalized gait recognition models, radar data need to be collected for various cases, resulting in the “independently and identically distributed (i.i.d.) assumption” in semisupervised learning that cannot be satisfied due to the presence of “covariant perturbations.” In this article, we propose a semisupervised model for radar-based gait recognition in the wild, and this model mitigates the performance degradation observed in the non-i.i.d. semisupervised condition. The model assesses the extent to which unlabeled samples are influenced by “covariant perturbations” and then differentiates how they are incorporated in semisupervised training, which not only prevents the samples whose semantics are destroyed from training but also enables the model to place more emphasis on the “hard samples.” We create a radar measurement dataset for the performance evaluation, and our model surpasses other seven semisupervised learning algorithms, demonstrating the effectiveness of our model in radar-based gait recognition problems. [ABSTRACT FROM AUTHOR]

Sparse modeling for signal processing and machine learning, in general, has been at the focus of scientific research for over two decades. Among others, supervised sparsity-aware learning (SAL) consists of two major paths paved by 1) discriminative methods that establish direct input–output mapping based on a regularized cost function optimization and 2) generative methods that learn the underlying distributions. The latter, more widely known as Bayesian methods, enable uncertainty evaluation with respect to the performed predictions. Furthermore, they can better exploit related prior information and also, in principle, can naturally introduce robustness into the model, due to their unique capacity to marginalize out uncertainties related to the parameter estimates. Moreover, hyperparameters (tuning parameters) associated with the adopted priors, which correspond to cost function regularizers, can be learned via the training data and not via costly cross-validation techniques, which is, in general, the case with the discriminative methods. [ABSTRACT FROM AUTHOR]

There has been a persistent challenge in acquiring sufficient training image data for deep neural networks (DNNs) to enhance safety monitoring on construction sites. Given the prevalence of textual data in the construction sector and the capabilities of multi-modal AI systems, this paper presents the use of text-to-image models to generate training images that capture the relationships between objects involved in construction accidents. Through a systematic prompt design process, a synthetic dataset of 3585 images across 27 hazardous scenarios was generated. The efficacy of this method is demonstrated by the performance of DNN models trained on these virtual images, which achieved a mean Average Precision (mAP) of approximately 64% in object detection and 60% in segmentation tasks. This paper demonstrates the potential of text-to-image models in mitigating the scarcity of training data and enhancing the capability of DNNs to identify potential hazards. • Mitigating the image data scarcity in a construction safety field using synthetic data. • Training DNN models with images synthesized by text-to-image generative models. • DNN models trained with only synthetic data achieved high mAP(%) when tested with real-world images. [ABSTRACT FROM AUTHOR]

This paper presents a comparison of three different machine learning frameworks applied in the wave buoy analogy used for estimating the sea state from measured ship responses. The three frameworks output and characterise the sea state in different ways: Model 1 outputs integral parameters, Model 2 outputs a point wave spectrum and the wave direction, Model 3 outputs the full directional wave spectrum. The assessment of the models is based on simulated motion measurements, i.e. synthetic data. In the particular investigations made, the performance of Model 2, relying on a novel framework, is generally superior. However, the central take-away from the study, is the importance of considering thorough and well-prepared training data encompassing many, not to say all, possible parameter combinations and shapes in the studied wave spectra forming the training data; any machine learning model is no better than the data upon which it is trained. • Machine learning applied in the wave buoy analogy. • Comparison of three fundamentally different ML models. • Presentation of a novel framework outputting a point wave spectrum and wave direction. • Thorough investigations of the models' generalisation capability. • Well-prepared and comprehensive training data is crucial. [ABSTRACT FROM AUTHOR]

The role of artificial intelligence in the Turing test is to imitate human beings to such an extent that people will not realize it is a machine. With the rise of deep learning (a subcategory of AI), the situation is changing rapidly as the new systems do not focus on imitating human intelligence but emphasize thorough solutions to specific issues. The main difference between predefined AI and deep learning (DL) is that these systems are self-learning and have verifiable results. Firstly, we need to analyse the application of the Turing test in the Loebner Prize because, there, the primary emphasis is on aspects of human intelligence – learning, reasoning and understanding. Secondly, in the Turing test, only general intelligence is considered, and this can be questionable. If DL does not possess this form of intelligence, by this reasoning, we should consider it unintelligent. However, is such understanding correct? The third and last aspect questions whether the Turing test is beneficial for an AI designed for specific tasks because the results do not bring any new data and conclusions.

In this paper, the effect of parameters on the IPR (Inflow Performance Relationship) curves of a well is analyzed using a multilayer perceptron artificial neural network. Genetic algorithms were used to train the network to adjust the weights and bias of the network so that the RMSE (Root Mean Squared Error) of the data assigned to the network test was minimized. The algorithm also optimized the network structure (including the number of hidden layers and the neurons of each layer). Based on Eclipse software, one of the oil reservoirs was first analyzed and then by changing the average reservoir pressures (assuming the oil well is predicted to reach such pressures above the average reservoir) and pressures. For this purpose, 48 experiments were conducted concerning the research variables implemented by each of them, providing inputs and outputs for network training and testing. Finally, 85% of the data were allocated to the training section and 15% to the software-generated software using the artificial neural network test section. The basis for deciding the error rate of the test data was considered. About 100% accuracy in the regression indicates that the neural network is reliable in other performances. The defined error is less than 2%, which indicated the efficiency of the genetic algorithm to predict IPR curves. This is rooted in applying genetic algorithms as a training function because it has increased the accuracy of data training. The magnitude of ANN (Artificial Neural Network) error for each scenario is also shown in the following curve. For example, the error of the sixth experiment is reported to be around 8%.

In this article, we propose a novel method to simultaneously solve the data problem of dirty quality and poor quantity for person reidentification (ReID). Dirty quality refers to the wrong labels in image annotations. Poor quantity means that some identities have very few images (FewIDs). Training with these mislabeled data or FewIDs with triplet loss will lead to low generalization performance. To solve the label error problem, we propose a weighted label correction based on cross-entropy (wLCCE) strategy. Specifically, according to the influence range of the wrong labels, we first classify the mislabeled images into point label error and set label error. Then, we propose a weighted triplet loss (WTL) to correct the two label errors, respectively. To alleviate the poor quantity issue, we propose a feature simulation based on autoencoder (FSAE) method to generate some virtual samples for FewID. For the authenticity of the simulated features, we transfer the difference pattern of identities with multiple images (MultIDs) to FewIDs by training an autoencoder (AE)-based simulator. In this way, the FewIDs obtain richer expressions to distinguish from other identities. By dealing with a dirty and poor data problem, we can learn more robust ReID models using the triplet loss. We conduct extensive experiments on two public person ReID datasets: 1) Market-1501 and 2) DukeMTMC-reID, to verify the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

Multiview fuzzy systems aim to deal with fuzzy modeling in multiview scenarios effectively and to obtain the interpretable model through multiview learning. However, current studies of multiview fuzzy systems still face several challenges, one of which is how to achieve efficient collaboration between multiple views when there are few labeled data. To address this challenge, this article explores a novel transductive multiview fuzzy modeling method. The dependency on labeled data is reduced by integrating transductive learning into the fuzzy model to simultaneously learn both the model and the labels using a novel learning criterion. Matrix factorization is incorporated to further improve the performance of the fuzzy model. In addition, collaborative learning between multiple views is used to enhance the robustness of the model. The experimental results indicate that the proposed method is highly competitive with other multiview learning methods. [ABSTRACT FROM AUTHOR]

Thanks to the powerful learning capability, deep neural networks (DNNs) have acquired broad applications in single image reflection removal. The DNN-based algorithms relax the constraints of specific priors and learn to generate visually pleasant background layers from massive training data. However, most of them employ a single network structure to recover both the semantic information and local details of the background, which may lead to obvious reflection residue or even failure. To mitigate this deficiency, in this work, we propose a Multi-stage Curvature-guided De-Reflection Network (MCDRNet), which combines multiple network architectures in a unified framework to progressively reconstruct the background layer and refine the fine-grained details. Our framework consists of three stages, where the encoder-decoders are exploited in the first two stages to recover the semantic components of background layers with lower scales and a variant ResNet is applied in the last stage to refine the background details with the original input resolution. In the first two stages, to introduce the structural guidance for the reflection removal, we cascade another decoder branch to restore the curvature map of the background. In addition, at the end of the first two stages, instead of directly passing the intermediate estimates to the next stage, we propose a Non-local Attention Module (NAM) to augment and transmit the features from decoders. Extensive experimental results on several public datasets demonstrate that the proposed MCDRNet outperforms the state-of-the-art methods quantitatively and generates visually better reflection removal results. The source code and pre-trained models are available at https://github.com/NamecantbeNULL/MCDRNet. [ABSTRACT FROM AUTHOR]

Support vector machines (SVM) have drawn wide attention for the last two decades due to its extensive applications, so a vast body of work has developed optimization algorithms to solve SVM with various soft-margin losses. To distinguish all, in this paper, we aim at solving an ideal soft-margin loss SVM: $L_{0/1}$ L 0 / 1 soft-margin loss SVM (dubbed as $L_{0/1}$ L 0 / 1 -SVM). Many of the existing (non)convex soft-margin losses can be viewed as one of the surrogates of the $L_{0/1}$ L 0 / 1 soft-margin loss. Despite its discrete nature, we manage to establish the optimality theory for the $L_{0/1}$ L 0 / 1 -SVM including the existence of the optimal solutions, the relationship between them and P-stationary points. These not only enable us to deliver a rigorous definition of $L_{0/1}$ L 0 / 1 support vectors but also allow us to define a working set. Integrating such a working set, a fast alternating direction method of multipliers is then proposed with its limit point being a locally optimal solution to the $L_{0/1}$ L 0 / 1 -SVM. Finally, numerical experiments demonstrate that our proposed method outperforms some leading classification solvers from SVM communities, in terms of faster computational speed and a fewer number of support vectors. The bigger the data size is, the more evident its advantage appears. [ABSTRACT FROM AUTHOR]

We consider the problem of predicting a response $Y$ Y from a set of covariates $X$ X when test- and training distributions differ. Since such differences may have causal explanations, we consider test distributions that emerge from interventions in a structural causal model, and focus on minimizing the worst-case risk. Causal regression models, which regress the response on its direct causes, remain unchanged under arbitrary interventions on the covariates, but they are not always optimal in the above sense. For example, for linear models and bounded interventions, alternative solutions have been shown to be minimax prediction optimal. We introduce the formal framework of distribution generalization that allows us to analyze the above problem in partially observed nonlinear models for both direct interventions on $X$ X and interventions that occur indirectly via exogenous variables $A$ A . It takes into account that, in practice, minimax solutions need to be identified from data. Our framework allows us to characterize under which class of interventions the causal function is minimax optimal. We prove sufficient conditions for distribution generalization and present corresponding impossibility results. We propose a practical method, NILE, that achieves distribution generalization in a nonlinear IV setting with linear extrapolation. We prove consistency and present empirical results. [ABSTRACT FROM AUTHOR]

Face anti-spoofing (FAS) secures face recognition from presentation attacks (PAs). Existing FAS methods usually supervise PA detectors with handcrafted binary or pixel-wise labels. However, handcrafted labels may are not the most adequate way to supervise PA detectors learning sufficient and intrinsic spoofing cues. Instead of using the handcrafted labels, we propose a novel Meta-Teacher FAS (MT-FAS) method to train a meta-teacher for supervising PA detectors more effectively. The meta-teacher is trained in a bi-level optimization manner to learn the ability to supervise the PA detectors learning rich spoofing cues. The bi-level optimization contains two key components: 1) a lower-level training in which the meta-teacher supervises the detector’s learning process on the training set; and 2) a higher-level training in which the meta-teacher’s teaching performance is optimized by minimizing the detector’s validation loss. Our meta-teacher differs significantly from existing teacher-student models because the meta-teacher is explicitly trained for better teaching the detector (student), whereas existing teachers are trained for outstanding accuracy neglecting teaching ability. Extensive experiments on five FAS benchmarks show that with the proposed MT-FAS, the trained meta-teacher 1) provides better-suited supervision than both handcrafted labels and existing teacher-student models; and 2) significantly improves the performances of PA detectors. [ABSTRACT FROM AUTHOR]

The 3D Morphable Model (3DMM) is a powerful statistical tool for representing 3D face shapes. To build a 3DMM, a training set of face scans in full point-to-point correspondence is required, and its modeling capabilities directly depend on the variability contained in the training data. Thus, to increase the descriptive power of the 3DMM, establishing a dense correspondence across heterogeneous scans with sufficient diversity in terms of identities, ethnicities, or expressions becomes essential. In this manuscript, we present a fully automatic approach that leverages a 3DMM to transfer its dense semantic annotation across raw 3D faces, establishing a dense correspondence between them. We propose a novel formulation to learn a set of sparse deformation components with local support on the face that, together with an original non-rigid deformation algorithm, allow the 3DMM to precisely fit unseen faces and transfer its semantic annotation. We extensively experimented our approach, showing it can effectively generalize to highly diverse samples and accurately establish a dense correspondence even in presence of complex facial expressions. The accuracy of the dense registration is demonstrated by building a heterogeneous, large-scale 3DMM from more than 9,000 fully registered scans obtained by joining three large datasets together. [ABSTRACT FROM AUTHOR]

Recently, a popular line of research in face recognition is adopting margins in the well-established softmax loss function to maximize class separability. In this paper, we first introduce an Additive Angular Margin Loss (ArcFace), which not only has a clear geometric interpretation but also significantly enhances the discriminative power. Since ArcFace is susceptible to the massive label noise, we further propose sub-center ArcFace, in which each class contains $K$ K sub-centers and training samples only need to be close to any of the $K$ K positive sub-centers. Sub-center ArcFace encourages one dominant sub-class that contains the majority of clean faces and non-dominant sub-classes that include hard or noisy faces. Based on this self-propelled isolation, we boost the performance through automatically purifying raw web faces under massive real-world noise. Besides discriminative feature embedding, we also explore the inverse problem, mapping feature vectors to face images. Without training any additional generator or discriminator, the pre-trained ArcFace model can generate identity-preserved face images for both subjects inside and outside the training data only by using the network gradient and Batch Normalization (BN) priors. Extensive experiments demonstrate that ArcFace can enhance the discriminative feature embedding as well as strengthen the generative face synthesis. [ABSTRACT FROM AUTHOR]

The Tsetlin Machine (TM) is a recent machine learning algorithm with several distinct properties, such as interpretability, simplicity, and hardware-friendliness. Although numerous empirical evaluations report on its performance, the mathematical analysis of its convergence is still open. In this article, we analyze the convergence of the TM with only one clause involved for classification. More specifically, we examine two basic logical operators, namely, the “IDENTITY”- and “NOT” operators. Our analysis reveals that the TM, with just one clause, can converge correctly to the intended logical operator, learning from training data over an infinite time horizon. Besides, it can capture arbitrarily rare patterns and select the most accurate one when two candidate patterns are incompatible, by configuring a granularity parameter. The analysis of the convergence of the two basic operators lays the foundation for analyzing other logical operators. These analyses altogether, from a mathematical perspective, provide new insights on why TMs have obtained state-of-the-art performance on several pattern recognition problems. [ABSTRACT FROM AUTHOR]

The development of deep neural network (DNN) has provided transformative impacts on many fields, including computer vision and video recognition. However, the impact is limited by the need for large, labeled datasets to enable effective training. To address this fundamental problem, we propose a novel inter-network system (Dandelion), providing architecture support (Dandelion-architecture) for data augmentation that trains DNNs with rare images generated by the generative adversarial network (GAN) with orthogonal attributes modified (Dandelion-function. The approach can account for the latency requirement and resource limitation of target applications by exploiting data and computation reuses between the two networks; this amortizes the impact of bottleneck brought by GAN and facilitates design of inter-network accelerator. Moreover, we show how to implement two-network design on 3D architecture to further enhance the accelerator. Our results show that with the generated images, DNN yields 13.6% - 37.5% improvement on accuracy, depending on the data scarcity level. Our architecture achieves at least 30% speedup compared with the baseline while 40% of the overhead brought by the incorporation of GAN is reduced in our design compared with ScaleDeep, and 26.3% of performance improvement over TETRIS. [ABSTRACT FROM AUTHOR]

Magnetoencephalography (MEG) is a useful tool for clinically evaluating the localization of interictal spikes. Neurophysiologists visually identify spikes from the MEG waveforms and estimate the equivalent current dipoles (ECD). However, presently, these analyses are manually performed by neurophysiologists and are time-consuming. Another problem is that spike identification from MEG waveforms largely depends on neurophysiologists’ skills and experiences. These problems cause poor cost-effectiveness in clinical MEG examination. To overcome these problems, we fully automated spike identification and ECD estimation using a deep learning approach fully automated AI-based MEG interictal epileptiform discharge identification and ECD estimation (FAMED). We applied a semantic segmentation method, which is an image processing technique, to identify the appropriate times between spike onset and peak and to select appropriate sensors for ECD estimation. FAMED was trained and evaluated using clinical MEG data acquired from 375 patients. FAMED training was performed in two stages: in the first stage, a classification network was learned, and in the second stage, a segmentation network that extended the classification network was learned. The classification network had a mean AUC of 0.9868 (10-fold patient-wise cross-validation); the sensitivity and specificity were 0.7952 and 0.9971, respectively. The median distance between the ECDs estimated by the neurophysiologists and those using FAMED was 0.63 cm. Thus, the performance of FAMED is comparable to that of neurophysiologists, and it can contribute to the efficiency and consistency of MEG ECD analysis. [ABSTRACT FROM AUTHOR]

Owing to the considerations of cell density and low static power consumption, nonvolatile random-access memory (NVRAM) has been a promising candidate for collaborating with a dynamic random-access memory (DRAM) as the main memory in modern computer systems. As NVRAM also brings technical challenges (e.g., limited endurance and high writing cost) to computer system developers, the concept of write reduction becomes the famous doctrine in NVRAM-based system design. Unfortunately, a well-known machine learning algorithm, random forest, will generate a massive amount of write traffic to the main memory space during its construction phase. In other words, a random forest hits the Achilles’ heel of NVRAM-based systems. For remedying this pain, our work proposes an NVRAM-friendly random forest algorithm, namely, Amine, for an NVRAM-based system. The design principle of Amine is to replace write operations with read accesses without raising the read complexity of the random forest algorithm. According to experimental results, Amine can effectively decrease the latency of random forest construction by 64%, compared with the original random forest algorithm. [ABSTRACT FROM AUTHOR]

Quantum entanglement brings high-speed and inherently privacy-preserving transmission for information communication in quantum networks. The qubit scarcity is an important issue that cannot be ignored in quantum networks due to the limited storage capacity of quantum devices, the short lifespan of qubits, and so on. In this article, we first formulate the qubit competition problem as the Cooperative-Qubit-Allocation-Problem (CQAP) by taking into account both the waiting time and the fidelity of end-to-end entanglement with the given transmission link set. We then model the CQAP as a Markov Decision Process (MDP) and adopt a Reinforcement Learning (RL) algorithm to self-adaptively and cooperatively allocate qubits among quantum repeaters. Further, we introduce an Active Learning (AL) algorithm to improve the efficiency of the RL algorithm by reducing its trial-error times. Simulation results demonstrate that our proposed algorithm outperforms the benchmark algorithms, with 23.5 ms reduction on the average waiting time and 19.2 improvement on the average path maturity degree, respectively. [ABSTRACT FROM AUTHOR]

The Right to be Forgotten gives a data owner the right to revoke their data from an entity storing it. In the context of federated learning, the Right to be Forgotten requires that, in addition to the data itself, any influence of the data on the FL model must disappear, a process we call “federated unlearning.” The most straightforward and legitimate way to implement federated unlearning is to remove the revoked data and retrain the FL model from scratch. Yet the computational and time overhead associated with fully retraining FL models can be prohibitively expensive. In this article, we take the first step to comprehensively investigate the way to settle the unlearning paradigm in the context of federated learning. First, we define the problem of efficient federated unlearning, including its challenges and goals, and we identify three common types of federated unlearning requests: class unlearning, client unlearning, and sample unlearning. Based on those challenges and goals, a general pipeline is proposed for federated unlearning for the above three types of requests. We revisit how the training data affects the final FL model performance and thereby empowers the proposed framework with the reverse stochastic gradient ascent (SGA) and elastic weight consolidation (EWC). Various experiments are conducted to verify effectiveness of the proposed method in both aspects of unlearning efficacy and efficiency. We believe the proposed method will perform as an essential component for future machine unlearning systems. [ABSTRACT FROM AUTHOR]

In this article, we study the problem of wireless human sensing, which refers to human activity recognition (HAR). HAR based on wireless signals plays an important role in security, human-computer interaction, and healthcare in the 5G era. Most state-of-the-art human activity recognition applications rely on deep learning approaches, which require a large amount of training data to achieve good performance. However, wireless signal data is difficult to collect and label, and it also carries private information, making it challenging to construct large-scale datasets.The recent advances in federated learning provide a chance to aggregate a wide range of users to collaboratively train a HAR model using decentralized datasets under data-preserving constraints. However, since a wireless signal is easily interrupted by the environment, the data across all participants is non-IID, thus decreasing the performance of an aggregated model. Additionally, due to the resource-constrained nature of edge devices, training the HAR model on an end user usually takes too long, resulting in straggler problems in federated learning training. In this article, we proposed a cross-domain federated learning framework (CDFL) to address the lack of labeled wireless data. A transfer learning approach was proposed to simulate wireless data by converting from widely available image datasets, and solving the distribution mismatch problem by domain adaption. Additionally, a customized federated learning approach was proposed to reduce the computational overhead of local model training. Using a case study of ultrasonic signal-based gesture recognition, we demonstrate the effectiveness of the proposed framework. Our method achieves over 90 percent accuracy on a 5-category task without real data, and 88 percent accuracy on a 10-category task when the user collects only one piece of data. [ABSTRACT FROM AUTHOR]

The KNN algorithm is one of the most popular data mining algorithms. It has been widely and successfully applied to data analysis applications across a variety of research topics in computer science. This paper illustrates that, despite its success, there remain many challenges in KNN classification, including K computation, nearest neighbor selection, nearest neighbor search and classification rules. Having established these issues, recent approaches to their resolution are examined in more detail, thereby providing a potential roadmap for ongoing KNN-related research, as well as some new classification rules regarding how to tackle the issue of training sample imbalance. To evaluate the proposed approaches, some experiments were conducted with 15 UCI benchmark datasets. [ABSTRACT FROM AUTHOR]

We study the problem of mismatched binary hypothesis testing between i.i.d. distributions. We analyze the tradeoff between the pairwise error probability exponents when the actual distributions generating the observation are different from the distributions used in the likelihood ratio test, sequential probability ratio test, and Hoeffding’s generalized likelihood ratio test in the composite setting. When the real distributions are within a small divergence ball of the test distributions, we find the deviation of the worst-case error exponent of each test with respect to the matched error exponent. In addition, we consider the case where an adversary tampers with the observation, again within a divergence ball of the observation type. We show that the tests are more sensitive to distribution mismatch than to adversarial observation tampering. [ABSTRACT FROM AUTHOR]

With the advent of deep learning, the research on facial expression recognition (FER) has received a lot of interest. Different deep convolutional neural network (DCNN) architectures have been developed for real-time and efficient FER. One of the challenges in FER is obtaining trustworthy features that are strongly associated with facial expression changes. Furthermore, traditional DCNNs for FER problems have two significant issues: insufficient training data, which leads to overfitting, and intra-class facial appearance variations. FLEPNet, a texture-based feature-level ensemble parallel network for FER, is proposed in this study and proved to solve the aforementioned problems. Our parallel network FLEPNet uses multi-scale convolutional and multi-scale residual block-based DCNN as building blocks. First, we consider modified homomorphic filtering to normalize the illumination effectively, which minimizes the intra-class difference. The deep networks are then protected against having insufficient training data by using texture analysis on face expression images to identify multiple attributes. Four texture features are extracted and combined with the image's original characteristics. Finally, the integrated features retrieved by two networks are used to classify seven facial expressions. Experimental results reveal that the proposed technique achieves an average accuracy of 0.9914, 0.9894, 0.9796, 0.8756, and 0.8072 on Japanese Female Facial Expressions, Extended CohnKanade, Karolinska Directed Emotional Faces, Real-world Affective Face Database, and Facial Expression Recognition 2013 databases, respectively. Moreover, experimental outcomes depict significant reliability when compared to competing approaches. [ABSTRACT FROM AUTHOR]

In this article, a reinforced iterative fuzzy radial basis function neural networks (IFRBFNN) is introduced as an augmented FRBFNN architecture generated through the iterative refinement process of weighted fuzzy C-means (WFCM) clustering and weighted least square error estimation. It is well known that the location of the clusters have an effect on the classification performance of the FRBFNN. The underlying idea behind this article is how to define the center points of clusters based on the data distribution analysis as well as the improvement of classification performance. In a nutshell, while the fuzzy C-means clustering to define the positions of radial basis functions (RBFs) based in the unsupervised learning manner is usually used in a FRBFNN, the parameters (cluster centers and their ensuing polynomial coefficients) refinement of FRBFNN through the proposed iterative method lead to superb classification performance by relocating the positions of RBFs over the input space implied by supervised learning. The idea of the proposed approach is to relocate the centers (prototypes) of the fuzzy clusters by using WFCM clustering algorithm and to re-estimate the coefficients by weighted least square estimation with the aid of the cross-entropy loss values of data so that the classification performance is improved. The weight related to each data is determined by some auxiliary information (i.e., the modified version of the cross-entropy loss value of each data). When it comes to the estimation of the coefficients of the consequent polynomial in IFRBFNN, they are estimated by the weighted least square error estimation technique, where the weights of the coefficients are defined based on the cross-entropy loss values. The auxiliary information such as the weights for relocation of the centers of RBFs and the weights for the re-estimation of the polynomial coefficients should be defined from the viewpoint of enhancement of the classification performance. The cross-entropy loss value of each data is able to meet the necessity of the required supervision signal. Several numerical experiments are provided to demonstrate the usefulness of the proposed design method for the FRBFNN for classification problems. [ABSTRACT FROM AUTHOR]

Cartoon is a common form of art in our daily life and automatic generation of cartoon images from photos is highly desirable. However, state-of-the-art single-style methods can only generate one style of cartoon images from photos and existing multi-style image style transfer methods still struggle to produce high-quality cartoon images due to their highly simplified and abstract nature. In this article, we propose a novel multi-style generative adversarial network (GAN) architecture, called MS-CartoonGAN, which can transform photos into multiple cartoon styles. MS-CartoonGAN uses only unpaired photos and cartoon images of multiple styles for training. To achieve this, we propose to use (1) a hierarchical semantic loss with sparse regularization to retain semantic content and recover flat shading in different abstract levels, (2) a new edge-promoting adversarial loss for producing fine edges, and (3) a style loss to enhance the difference between output cartoon styles and make training process more stable. We also develop a multi-domain architecture, where the generator consists of a shared encoder and multiple decoders for different cartoon styles, along with multiple discriminators for individual styles. By observing that cartoon images drawn by different artists have their unique styles while sharing some common characteristics, our shared network architecture exploits the common characteristics of cartoon styles, achieving better cartoonization and being more efficient than single-style cartoonization. We show that our multi-domain architecture can theoretically guarantee to output desired multiple cartoon styles. Through extensive experiments including a user study, we demonstrate the superiority of the proposed method, outperforming state-of-the-art single-style and multi-style image style transfer methods. [ABSTRACT FROM AUTHOR]

Expensive multiobjective optimization problems pose great challenges to evolutionary algorithms due to their costly evaluation. Building cheap surrogate models to replace the expensive real models has been proved to be a practical way to reduce the number of costly evaluations. Supervised learning techniques from the community of machine learning have been widely applied to build either regressors, which approximate the fitness values of candidate solutions, or classifiers, which estimate the categories of candidate solutions. Considering the characteristics of the data produced in optimization, this article proposes a new surrogate model, called a relation model, for evolutionary multiobjective optimization. Instead of estimating the qualities of candidate solutions directly, the relation model tries to estimate the relationship between a pair of solutions $\langle \mathbf {x}, \mathbf {y}\rangle $ , i.e., $\mathbf {x}$ dominates $\mathbf {y}$ , $\mathbf {x}$ is dominated by $\mathbf {y}$ , or $\mathbf {x}$ is nondominated with $\mathbf {y}$ in the case of multiobjective optimization. To implement this idea, first a balanced training set is prepared, then a classifier is built based on the training data set to learn the relationship, and finally, the classifier with a voting-scoring strategy is applied to estimate the relationship between the candidate solutions and parent solutions. By this way, the promising candidate solutions are recognized and evaluated by the real models. The new approach is applied to three well-known benchmark suites and two real-world applications, and the experimental results suggest that the proposed method outperforms some state-of-the-art methods based on regression and classification models on the given instances. [ABSTRACT FROM AUTHOR]

Synthetic aperture sonar (SAS) allows for high-resolution imagery of the seafloor with a quality that approaches that of an optical camera. Identifying seabed types within SAS images can prove very useful in a setting where autonomous underwater vehicles (AUVs) are used, since information about the bottom type can be used to determine the behavior of the AUV. In this article, we present a new algorithm for the automatic detection of different seabed types based on SAS image texture. Using a Gaussian process classification (GPC) model, we were able to predict the seabed type, as well as an uncertainty estimate on this prediction. The GPC-based model is able to learn the most informative textures from a training data set, and by construction is robust against outliers or data points that are very different from those in the training set. An additional benefit of the model is that it can be easily configured to hit the desired tradeoff between computational complexity and accuracy. [ABSTRACT FROM AUTHOR]

Vision-based intelligent driving assistance systems and transportation systems can be improved by enhancing the visibility of the scenes captured in extremely challenging conditions. In particular, many low-image image enhancement (LIE) algorithms have been proposed to facilitate such applications in low-light conditions. While deep learning-based methods have achieved substantial success in this field, most of them require paired training data, which is difficult to be collected. This paper advocates a novel Unsupervised Decomposition and Correction Network (UDCN) for LIE without depending on paired data for training. Inspired by the Retinex model, our method first decomposes images into illumination and reflectance components with an image decomposition network (IDN). Then, the decomposed illumination is processed by an illumination correction network (ICN) and fused with the reflectance to generate a primary enhanced result. In contrast with fully supervised learning approaches, UDCN is an unsupervised one which is trained only with low-light images and corresponding histogram equalized (HE) counterparts (can be derived from the low-light image itself) as input. Both the decomposition and correction networks are optimized under the guidance of hybrid no-reference quality-aware losses and inter-consistency constraints between the low-light image and its HE counterpart. In addition, we also utilize an unsupervised noise removal network (NRN) to remove the noise previously hidden in the darkness for further improving the primary result. Qualitative and quantitative comparison results are reported to demonstrate the efficacy of UDCN and its superiority over several representative alternatives in the literature. The results and code will be made public available at https://github.com/myd945/UDCN. [ABSTRACT FROM AUTHOR]

For safety planning in crowd evacuation, it is important to predict the evacuation decisions made by different individuals and understand the reasons behind these decisions. To this end, this paper proposes an automated approach that can learn prioritized fuzzy decision rules from crowd data to predict and understand the evacuation decisions of a real human. A coevolutionary fuzzy rule miner based on genetic fuzzy-system is designed to select necessary decision features from available ones and learn both rule structure and associated rule parameters from training data. The learned fuzzy rule contains multiple sub-rules, each of which can represent evacuation strategies of different individuals in a given scenario and the features in the fuzzy condition of the sub-rule are organized and evaluated in a sequential order to reflect the priorities of different features. Based on training and testing on four evacuations scenarios of two real-world datasets, it is shown that our proposed approach can learn decision rules that are competitive to the existing evacuation decision models in terms of prediction accuracy. More importantly, it is also demonstrated that our learned rules complying with the proposed prioritized fuzzy rule representation can facilitate the interpretation of evacuation behaviors, such as “herding under zero visibility of exit” and “diminished importance on the distance to exit”, which are aligned to the field observations from real crowd evacuation. [ABSTRACT FROM AUTHOR]

Practical industrial processes usually have nonstationary properties, which make the monitoring more challenging because the fault information may be buried by nonstationary trends. For nonstationary processes, many methods have been proposed for fault detection based on continuous variables. However, binary variables may appear together with continuous variables in modern industrial processes. To address the issue of process monitoring with hybrid variables and nonstationarity, a model named recursive hybrid variable monitoring (RHVM) is proposed in this paper. For RHVM, recursive strategy is utilized to suppress nonstationary trend and to reveal fault information. In addition, RHVM has the ability of model self-updating with arriving samples. The closed-form updates of required parameters are derived in detail and the improvement of performance is analyzed. At last, the superiority of the proposed model is demonstrated by a simulation example and a practical nonstationary process of a power plant. [ABSTRACT FROM AUTHOR]

In order to improve mobile data transparency, various approaches have been proposed to inspect network traffic generated by mobile devices and detect exposure of personally identifiable information (PII), ad requests, etc. State-of-the-art approaches use features extracted from HTTP packets and train classifiers in a centralized way: users collect and label network packets on their mobile devices, then upload data to a central server; the server uses the data contributed by all users to train a packet classifier. However, training datasets from network traffic collected on user devices may contain sensitive information that users may not want to upload. In this article, we propose a federated learning approach to mobile packet classification, which enables devices to collaboratively train a global model, without uploading the training data collected on devices. We apply our framework to two packet classification tasks (i.e., to predict PII exposure or ad requests in individual packets) and we demonstrate its effectiveness in terms of classification performance, communication and computation cost, using three real-world datasets. Methodological challenges we address in the process include model and feature selection, as well as tuning the federated learning parameters specifically for our packet classification tasks. We also discuss privacy limitations and mitigation approaches. [ABSTRACT FROM AUTHOR]

In order to solve insufficient training data problem in remote sensing, a web platform was created so that registered users can generate labeled data for various classes in a dynamic structure. Users were asked to select representative pixel groups for the forest, hazelnut, shadow, soil, tea, and building classes with the polygon tool, and then assign a class label corresponding to each created polygon thanks to the help document displaying descriptive information regarding the locations, colors, textures and distributions of the classes in the image. Crowdsourcing was again used to test the accuracy of the tagged data produced by crowdsourcing. The created data set was overlaid with the original WV-2 image, and the correctness of the labels ​​of the polygons was once visually verified. Finally, the WV-2 image, consisting of 40 patches, was classified with CNN and an average of over 95% accuracy was achieved. [ABSTRACT FROM AUTHOR]

Although hand pose estimation has achieved a great success in recent years, there are still challenges with RGB-based estimation tasks, the most significant of which is the absence of labeled training data. At present, the synthetic dataset has plenty of images with accurate annotation, but the difference from real-world datasets affects generalization. Therefore, a transfer learning strategy, which tries to transfer knowledge from a labeled source domain to an unlabeled target domain, is a frequent solution. Existing methods such as mean-teacher, Cyclegan, and MCD will train models with the help of some easily accessible domains such as synthetic data. However, these methods are not guaranteed to operate well in real-world settings due to the domain shift. In this paper, we design a new unsupervised domain adaptation method named Multi-branch Adversarial Regressors (MarsDA) in hand pose estimation, where it could be better for feature migration. Specifically, we first generate pseudo-labels for unlabeled target domain data. Then, the new adversarial training loss between multiple regression branches we designed for hand pose estimation is introduced to narrow the domain gap. In this way, our model can reduce the noise of pseudo labels caused by the domain gap and improve the accuracy of pseudo labels. We evaluate our method on two publicly available real-world datasets, H3D and STB. Experimental results show that our method outperforms existing methods by a large margin. [ABSTRACT FROM AUTHOR]

Deep neural networks are trained so as to achieve a kind of the maximum overall accuracy through a learning process using given training data. Therefore, it is difficult to fix them to improve the accuracies of specific problematic classes or classes of interest that may be valuable to some users or applications. To address this issue, we propose the synaptic join method to tweak neural networks by adding certain additional synapses from the intermediate hidden layers to the output layer across layers and additionally training only these synapses, if necessary. To select the most effective synapses, the synaptic join method evaluates the performance of all the possible candidate synapses between the hidden neurons and output neurons based on the distribution of all the possible proper weights. The experimental results show that the proposed method can effectively improve the accuracies of specific classes in a controllable way. [ABSTRACT FROM AUTHOR]

Automating sleep staging is vital to scale up sleep assessment and diagnosis to serve millions experiencing sleep deprivation and disorders and enable longitudinal sleep monitoring in home environments. Learning from raw polysomnography signals and their derived time-frequency image representations has been prevalent. However, learning from multi-view inputs (e.g., both the raw signals and the time-frequency images) for sleep staging is difficult and not well understood. This work proposes a sequence-to-sequence sleep staging model, XSleepNet,1 that is capable of learning a joint representation from both raw signals and time-frequency images. Since different views may generalize or overfit at different rates, the proposed network is trained such that the learning pace on each view is adapted based on their generalization/overfitting behavior. In simple terms, the learning on a particular view is speeded up when it is generalizing well and slowed down when it is overfitting. View-specific generalization/overfitting measures are computed on-the-fly during the training course and used to derive weights to blend the gradients from different views. As a result, the network is able to retain the representation power of different views in the joint features which represent the underlying distribution better than those learned by each individual view alone. Furthermore, the XSleepNet architecture is principally designed to gain robustness to the amount of training data and to increase the complementarity between the input views. Experimental results on five databases of different sizes show that XSleepNet consistently outperforms the single-view baselines and the multi-view baseline with a simple fusion strategy. Finally, XSleepNet also outperforms prior sleep staging methods and improves previous state-of-the-art results on the experimental databases. [ABSTRACT FROM AUTHOR]

Category-level 6-D object pose tracking is very challenging in the field of 3-D computer vision. Keypoint-based object pose estimation has demonstrated its effectiveness in dealing with it. However, current approaches first estimate the keypoints through a neural network and further compute the interframe pose change via least-squares optimization. They estimate rotation and translation in the same way, ignoring the differences between them. In this work, we propose a keypoint-based disentangled pose network, which disentangles the 6-D object pose change to 3-D rotation and 3-D translation. Specifically, the translation is directly estimated by the network and the rotation is indirectly calculated by singular value decomposition according to the keypoints. Extensive experiments on the NOCS-REAL275 dataset demonstrate the superiority of our method. [ABSTRACT FROM AUTHOR]

The performance of a deep-learning-based model primarily relies on the diversity and size of the training dataset. However, obtaining such a large amount of labeled data for practical remote sensing (RS) applications is expensive and labor-intensive. Training protocols have been previously proposed for few-shot learning (FSL) and zero-shot learning (ZSL). However, FSL is not compatible with handling unobserved class data at the inference phase, while ZSL requires many training samples of the seen classes. In this work, we propose a novel training protocol for image retrieval and name it as label-deficit zero-shot learning (LDZSL). We use this novel LDZSL training protocol for the challenging task of cross-sensor data retrieval in RS. This protocol uses very few labeled data samples of the seen classes during training and interprets unobserved class data samples at the inference phase. This strategy is critical as some data modalities are hard to annotate without domain experts. This work proposes a novel bilevel Siamese network to perform the LDZSL cross-sensor retrieval of multispectral and synthetic aperture radar (SAR) images. We use the available georeferenced SAR and multispectral data to domain align the embedding features of the two modalities. We experimentally demonstrate the proposed model’s efficacy using the So2Sat dataset compared with the existing state-of-the-art models of the ZSL framework trained under a reduced training set. We also show the generalizability of the proposed model using a sketch-based image retrieval task. Experimental results on the Earth on the Canvas dataset exhibit comparative performance over the literature. [ABSTRACT FROM AUTHOR]

Objective: Statistical shape models have been successfully used in numerous biomedical image analysis applications where prior shape information is helpful such as organ segmentation or data augmentation when training deep learning models. However, training such models requires large data sets, which are often not available and, hence, shape models frequently fail to represent local details of unseen shapes. This work introduces a kernel-based method to alleviate this problem via so-called model localization. It is specifically designed to be used in large-scale shape modeling scenarios like deep learning data augmentation and fits seamlessly into the classical shape modeling framework. Method: Relying on recent advances in multi-level shape model localization via distance-based covariance matrix manipulations and Grassmannian-based level fusion, this work proposes a novel and computationally efficient kernel-based localization technique. Moreover, a novel way to improve the specificity of such models via normalizing flow-based density estimation is presented. Results: The method is evaluated on the publicly available JSRT/SCR chest X-ray and IXI brain data sets. The results confirm the effectiveness of the kernelized formulation and also highlight the models’ improved specificity when utilizing the proposed density estimation method. Conclusion: This work shows that flexible and specific shape models from few training samples can be generated in a computationally efficient way by combining ideas from kernel theory and normalizing flows. Significance: The proposed method together with its publicly available implementation allows to build shape models from few training samples directly usable for applications like data augmentation. [ABSTRACT FROM AUTHOR]

Machine learning models are established with a variety of data collected from individual users who are concerned about their privacy. Various cloud service providers (e.g., Amazon, Google, Alibaba) commercialize their models by selling them or providing limited access via APIs. However, existing studies have shown that such cloud-based models are susceptible to training data inference attacks (i.e., membership inference attacks, attribute inference attacks, and model inversion attacks) since the trained models remember information about the training data. In this article, we perform an exhaustive investigation on the current training data inference attacks against cloud-based models. According to the attack scenario, we divide the existing inference attacks into two categories: centralized and collaborative learning scenarios. We first give a comprehensive review of attacks in each category and then give a qualitative comparison of the existing attacks. Further, we quantitatively compare the performance of different attack strategies by experiments. Last but not least, we pinpoint some open issues and potential future directions that need to be investigated in this area. [ABSTRACT FROM AUTHOR]

We propose a machine-learning-based workflow to solve the spectral estimation problem of time signals from electromagnetic simulations of highly resonant devices. An activation function for a three-layer radial basis function network is defined and used to fit the signal originated from a grid-based computation. Due to the proper choice of the activation function and a careful minimization of the loss expression, we are able to train this network using only a small part of the original time signal and finally extrapolate the signal up to a point it has sufficiently decayed. In the end, we obtain an estimation of the signal’s spectral content which is considerably more accurate than a Fourier transform of the training data alone. Therefore, using our approach one may abort a time domain simulation earlier, and still calculate important key quantities such as resonance frequencies and $Q$ -values with high precision. In order to test our method, we perform the spectral estimation of a highly resonant filter’s response. [ABSTRACT FROM AUTHOR]

Class imbalance problems have been reported as a major issue in various applications. Classification becomes further complicated when an imbalance occurs in time series data sets. To address time series data, it is necessary to consider their characteristics (i.e., high dimensionality, high correlations, and multimodality). Oversampling is a well-known approach for addressing this problem; however, such an approach does not appropriately consider the characteristics of time series data. This paper addresses these limitations by presenting a model-based oversampling approach, a boundary-focused generative adversarial network (BFGAN). The proposed BFGAN employs a specifically designed additional label for reflecting the importance of a sample's position in data space. Furthermore, the BFGAN generates artificial samples after taking into consideration a sample's multimodality and importance by using a suitable modified GAN structure. We present empirical results that reveal a significant improvement in the quality of the generated data when the proposed BFGAN is used as an oversampling algorithm for an imbalanced multimodal time series data set. [ABSTRACT FROM AUTHOR]

Heterogeneous Information Network (HIN) collecitve classification studies the problem of predicting labels for one type of nodes in a HIN which contains multiple types of nodes multiple types of links among them. Previous studies have revealed that exploiting relative importance of links is quite useful to improve node classification performance as connected nodes tend to have similar labels. Most existing approaches exploit the relative importance of links either by directly counting the number of connections among nodes or by learning the weight of each type of link from labeled data only. However, these approaches either neglect the importance of types of links to the class labels or may lead to overfitting problem. We propose a Tensor-based Markov chain (T-Mark) approach, which is able to automatically and simultaneously predict the labels for unlabeled nodes and give the relative importance of types of links that actually improve the classification accuracy. Specifically, we build two tensor equations by using the HIN and features of nodes from both labeled and unlabeled data. A Markov chain-based model is proposed and it is solved by an iterative process to obtain the stationary distributions. Theoretical analyses of the existence and uniqueness of such probability distributions are given. Extensive experimental results demonstrate that T-Mark is able to achieve superior performance in the comparison and obtain reasonable relative importance of links. [ABSTRACT FROM AUTHOR]

On-line detection of anomalies in time series is a key technique used in various event-sensitive scenarios such as robotic system monitoring, smart sensor networks and data center security. However, the increasing diversity of data sources and the variety of demands make this task more challenging than ever. First, the rapid increase in unlabeled data means supervised learning is becoming less suitable in many cases. Second, a large portion of time series data have complex seasonality features. Third, on-line anomaly detection needs to be fast and reliable. In light of this, we have developed a prediction-driven, unsupervised anomaly detection scheme, which adopts a backbone model combining the decomposition and the inference of time series data. Further, we propose a novel metric, Local Trend Inconsistency (LTI), and an efficient detection algorithm that computes LTI in a real-time manner and scores each data point robustly in terms of its probability of being anomalous. We have conducted extensive experimentation to evaluate our algorithm with several datasets from both public repositories and production environments. The experimental results show that our scheme outperforms existing representative anomaly detection algorithms in terms of the commonly used metric, Area Under Curve (AUC), while achieving the desired efficiency. [ABSTRACT FROM AUTHOR]