Multi-source subdomain negative transfer suppression and multiple pseudo-labels guidance alignment: A method for fault diagnosis under cross-working conditions.

Extensive researches have been conducted on transfer learning based fault diagnosis. However, negative information transfer may arise due to significant differences in the subdomain distribution across multiple source domains (MSDs). Most existing methods focus solely on the impact of subdomains from a single source domain (SSD) on the target domain (TD). Therefore, this paper proposed a novel multi-stage alignment multi-source subdomain adaptation (MAMSA) method. The global feature extractor is designed to extract domain-invariant features. Three domain-specific feature extractors capture high-level fault features from different domains with a customized adaptation strategy, which combines adversarial learning and distribution alignment based on multiple pseudo-label-guided local maximum mean discrepancy (MP-LMMD) to learn subdomain-invariant features. MP-LMMD utilizes pseudo-labels generated from all classifiers in the TD to guide the alignment of subdomains, suppressing negative transfer from the source domains (SDs). The experimental results indicate that the MAMSA method has excellent capabilities to suppress negative transfer, and the diagnostic performance can be greatly promoted with MAMSA under cross-working conditions.

EEG-based emotion recognition using a temporal-difference minimizing neural network.

Electroencephalogram (EEG) emotion recognition plays an important role in human-computer interaction. An increasing number of algorithms for emotion recognition have been proposed recently. However, it is still challenging to make efficient use of emotional activity knowledge. In this paper, based on prior knowledge that emotion varies slowly across time, we propose a temporal-difference minimizing neural network (TDMNN) for EEG emotion recognition. We use maximum mean discrepancy (MMD) technology to evaluate the difference in EEG features across time and minimize the difference by a multibranch convolutional recurrent network. State-of-the-art performances are achieved using the proposed method on the SEED, SEED-IV, DEAP and DREAMER datasets, demonstrating the effectiveness of including prior knowledge in EEG emotion recognition.

Fast Model Selection and Hyperparameter Tuning for Generative Models.

Generative models have gained significant attention in recent years. They are increasingly used to estimate the underlying structure of high-dimensional data and artificially generate various kinds of data similar to those from the real world. The performance of generative models depends critically on a good set of hyperparameters. Yet, finding the right hyperparameter configuration can be an extremely time-consuming task. In this paper, we focus on speeding up the hyperparameter search through adaptive resource allocation, early stopping underperforming candidates quickly and allocating more computational resources to promising ones by comparing their intermediate performance. The hyperparameter search is formulated as a non-stochastic best-arm identification problem where resources like iterations or training time constrained by some predetermined budget are allocated to different hyperparameter configurations. A procedure which uses hypothesis testing coupled with Successive Halving is proposed to make the resource allocation and early stopping decisions and compares the intermediate performance of generative models by their exponentially weighted Maximum Means Discrepancy (MMD). The experimental results show that the proposed method selects hyperparameter configurations that lead to a significant improvement in the model performance compared to Successive Halving for a wide range of budgets across several real-world applications.

A new domain adaption residual separable convolutional neural network model for cross-domain remaining useful life prediction.

In order to realize the remaining useful life (RUL) prediction of mechanical equipment under different operating conditions, a domain adaption residual separable convolutional neural network (DRSCN) model is proposed in this paper. In the DRSCN model, instead of the traditional convolutional layer, a residual separable convolutional module is developed to improve the feature extraction ability of the model. Moreover, a multi-kernel maximum mean discrepancy metric function and an adversarial learning mechanism are embedded in the DRSCN model to enhance its ability to resist domain shifts, thus improving the cross-domain RUL prediction accuracy of the model. The effectiveness of the DRSCN model is verified on an aircraft engine dataset. The experimental results show that the proposed model can realize high-accuracy RUL prediction.

Aero-Engine Remaining Useful Life Prediction Based on Bi-Discrepancy Network.

Most unsupervised domain adaptation (UDA) methods align feature distributions across different domains through adversarial learning. However, many of them require introducing an auxiliary domain alignment model, which incurs additional computational costs. In addition, they generally focus on the global distribution alignment and ignore the fine-grained domain discrepancy, so target samples with significant domain shifts cannot be detected or processed for specific tasks. To solve these problems, a bi-discrepancy network is proposed for the cross-domain prediction task. Firstly, target samples with significant domain shifts are detected by maximizing the discrepancy between the outputs of the dual regressor. Secondly, the adversarial training mechanism is adopted between the feature generator and the dual regressor for global domain adaptation. Finally, the local maximum mean discrepancy is used to locally align the fine-grained features of different degradation stages. In 12 cross@-domain prediction tasks generated on the C-MAPSS dataset, the root-mean-square error (RMSE) was reduced by 77.24%, 61.72%, 38.97%, and 3.35% on average, compared with the four mainstream UDA methods, which proved the effectiveness of the proposed method.

Possibilistic distribution distance metric: a robust domain adaptation learning method.

The affective Brain-Computer Interface (aBCI) systems, which achieve predictions for individual subjects through training on multiple subjects, often cannot achieve satisfactory results due to the differences in Electroencephalogram (EEG) patterns between subjects. One tried to use Subject-specific classifiers, but there was a lack of sufficient labeled data. To solve this problem, Domain Adaptation (DA) has recently received widespread attention in the field of EEG-based emotion recognition. Domain adaptation (DA) learning aims to solve the problem of inconsistent distributions between training and test datasets and has received extensive attention. Most existing methods use Maximum Mean Discrepancy (MMD) or its variants to minimize the problem of domain distribution inconsistency. However, noisy data in the domain can lead to significant drift in domain means, which can affect the adaptability performance of learning methods based on MMD and its variants to some extent. Therefore, we propose a robust domain adaptation learning method with possibilistic distribution distance measure. Firstly, the traditional MMD criterion is transformed into a novel possibilistic clustering model to weaken the influence of noisy data, thereby constructing a robust possibilistic distribution distance metric (P-DDM) criterion. Then the robust effectiveness of domain distribution alignment is further improved by a fuzzy entropy regularization term. The proposed P-DDM is in theory proved which be an upper bound of the traditional distribution distance measure method MMD criterion under certain conditions. Therefore, minimizing P-DDM can effectively optimize the MMD objective. Secondly, based on the P-DDM criterion, a robust domain adaptation classifier based on P-DDM (C-PDDM) is proposed, which adopts the Laplacian matrix to preserve the geometric consistency of instances in the source domain and target domain for improving the label propagation performance. At the same time, by maximizing the use of source domain discriminative information to minimize domain discrimination error, the generalization performance of the learning model is further improved. Finally, a large number of experiments and analyses on multiple EEG datasets (i.e., SEED and SEED-IV) show that the proposed method has superior or comparable robustness performance (i.e., has increased by around 10%) in most cases.

Harmonizing Healthy Cohorts to Support Multicenter Studies on Migraine Classification using Brain MRI Data.

Multicenter and multi-scanner imaging studies might be needed to provide sample sizes large enough for developing accurate predictive models. However, multicenter studies, which likely include confounding factors due to subtle differences in research participant characteristics, MRI scanners, and imaging acquisition protocols, might not yield generalizable machine learning models, that is, models developed using one dataset may not be applicable to a different dataset. The generalizability of classification models is key for multi-scanner and multicenter studies, and for providing reproducible results. This study developed a data harmonization strategy to identify healthy controls with similar (homogenous) characteristics from multicenter studies to validate the generalization of machine-learning techniques for classifying individual migraine patients and healthy controls using brain MRI data. The Maximum Mean Discrepancy (MMD) was used to compare the two datasets represented in Geodesic Flow Kernel (GFK) space, capturing the data variabilities for identifying a "healthy core". A set of homogeneous healthy controls can assist in overcoming some of the unwanted heterogeneity and allow for the development of classification models that have high accuracy when applied to new datasets. Extensive experimental results show the utilization of a healthy core. One dataset consists of 120 individuals (66 with migraine and 54 healthy controls) and another dataset consists of 76 (34 with migraine and 42 healthy controls) individuals. A homogeneous dataset derived from a cohort of healthy controls improves the performance of classification models by about 25% accuracy improvements for both episodic and chronic migraineurs.

FEM Simulation-Based Adversarial Domain Adaptation for Fatigue Crack Detection Using Lamb Wave.

Lamb wave-based damage detection technology shows great potential for structural integrity assessment. However, conventional damage features based damage detection methods and data-driven intelligent damage detection methods highly rely on expert knowledge and sufficient labeled data for training, for which collecting is usually expensive and time-consuming. Therefore, this paper proposes an automated fatigue crack detection method using Lamb wave based on finite element method (FEM) and adversarial domain adaptation. FEM-simulation was used to obtain simulated response signals under various conditions to solve the problem of the insufficient labeled data in practice. Due to the distribution discrepancy between simulated signals and experimental signals, the detection performance of classifier just trained with simulated signals will drop sharply on the experimental signals. Then, Domain-adversarial neural network (DANN) with maximum mean discrepancy (MMD) was used to achieve discriminative and domain-invariant feature extraction between simulation source domain and experiment target domain, and the unlabeled experimental signals samples will be accurately classified. The proposed method is validated by fatigue tests on center-hole metal specimens. The results show that the proposed method presents superior detection ability compared to other methods and can be used as an effective tool for cross-domain damage detection.

Unsupervised model adaptation for multivariate calibration by domain adaptation-regularization based kernel partial least square.

In chemometrics, calibration model adaptation is desired when training- and test-samples come from different distributions. Domain-invariant feature representation is currently a successful strategy to realize model adaptation and has received wide attention. The paper presents a nonlinear unsupervised model adaptation method termed as domain adaption regularization-based kernel partial least squares regression (DarKPLS). DarKPLS aims to minimize the source and target distributions in a low-dimensional latent space projected from the reproducing kernel Hilbert space (RKHS) generated with the labeled source data and unlabeled target data. Specially, the distributional means and variances between source and target latent variables are aligned in the RKHS. By extending existing domain invariant partial least square regression (di-PLS) with the projected maximum mean discrepancy (PMMD) to reduce the mean discrepancy in the RKHS further, DarKPLS could realize fine-grained domain alignment that further improves the adaptation performance. DarKPLS is applied to the γ-polyglutamic acid fermentation dataset, tobacco dataset and corn dataset, and it demonstrates improved prediction results in comparison with No adaptation partial least squares (PLS), null augmented regression (NAR), extended linear joint trained framework (ExtJT), scatter component analysis (SCA) and domain-invariant iterative partial least squares (DIPALS).

Margin-Based Modal Adaptive Learning for Visible-Infrared Person Re-Identification.

Visible-infrared person re-identification (VIPR) has great potential for intelligent transportation systems for constructing smart cities, but it is challenging to utilize due to the huge modal discrepancy between visible and infrared images. Although visible and infrared data can appear to be two domains, VIPR is not identical to domain adaptation as it can massively eliminate modal discrepancies. Because VIPR has complete identity information on both visible and infrared modalities, once the domain adaption is overemphasized, the discriminative appearance information on the visible and infrared domains would drain. For that, we propose a novel margin-based modal adaptive learning (MMAL) method for VIPR in this paper. On each domain, we apply triplet and label smoothing cross-entropy functions to learn appearance-discriminative features. Between the two domains, we design a simple yet effective marginal maximum mean discrepancy (M3D) loss function to avoid an excessive suppression of modal discrepancies to protect the features' discriminative ability on each domain. As a result, our MMAL method could learn modal-invariant yet appearance-discriminative features for improving VIPR. The experimental results show that our MMAL method acquires state-of-the-art VIPR performance, e.g., on the RegDB dataset in the visible-to-infrared retrieval mode, the rank-1 accuracy is 93.24% and the mean average precision is 83.77%.

Reinforcement learning-based integrated active fault diagnosis and tracking control.

Reliable and real-time active diagnosis of system faults with uncertainties is strongly dependent on the input design. This paper establishes a data-driven framework for integrated design of active fault diagnosis and control while ensuring the tracking performance. To be specific, the input design is formulated as a constrained optimization problem that can be solved with the aid of constrained reinforcement learning algorithms. Moreover, based on the maximum mean discrepancy metric, a novel active fault isolation scheme is proposed to implement model discrimination using system outputs. At the end, the effectiveness of the proposed approach is evaluated in two case studies in the presence of probabilistic disturbances and uncertainties.

A dual alignment-based multi-source domain adaptation framework for motor imagery EEG classification.

Domain adaptation, as an important branch of transfer learning, can be applied to cope with data insufficiency and high subject variabilities in motor imagery electroencephalogram (MI-EEG) based brain-computer interfaces. The existing methods generally focus on aligning data and feature distribution; however, aligning each source domain with the informative samples of the target domain and seeking the most appropriate source domains to enhance the classification effect has not been considered. In this paper, we propose a dual alignment-based multi-source domain adaptation framework, denoted DAMSDAF. Based on continuous wavelet transform, all channels of MI-EEG signals are converted respectively and the generated time-frequency spectrum images are stitched to construct multi-source domains and target domain. Then, the informative samples close to the decision boundary are found in the target domain by using entropy, and they are employed to align and reassign each source domain with normalized mutual information. Furthermore, a multi-branch deep network (MBDN) is designed, and the maximum mean discrepancy is embedded in each branch to realign the specific feature distribution. Each branch is separately trained by an aligned source domain, and all the single branch transfer accuracies are arranged in descending order and utilized for weighted prediction of MBDN. Therefore, the most suitable number of source domains with top weights can be automatically determined. Extensive experiments are conducted based on 3 public MI-EEG datasets. DAMSDAF achieves the classification accuracies of 92.56%, 69.45% and 89.57%, and the statistical analysis is performed by the kappa value and t-test. Experimental results show that DAMSDAF significantly improves the transfer effects compared to the present methods, indicating that dual alignment can sufficiently use the different weighted samples and even source domains at different levels as well as realizing optimal selection of multi-source domains.

Distribution Adaptation and Classification Framework Based on Multiple Kernel Learning for Motor Imagery BCI Illiteracy.

A brain-computer interface (BCI) translates a user's thoughts such as motor imagery (MI) into the control of external devices. However, some people, who are defined as BCI illiteracy, cannot control BCI effectively. The main characteristics of BCI illiterate subjects are low classification rates and poor repeatability. To address the problem of MI-BCI illiteracy, we propose a distribution adaptation method based on multi-kernel learning to make the distribution of features between the source domain and target domain become even closer to each other, while the divisibility of categories is maximized. Inspired by the kernel trick, we adopted a multiple-kernel-based extreme learning machine to train the labeled source-domain data to find a new high-dimensional subspace that maximizes data divisibility, and then use multiple-kernel-based maximum mean discrepancy to conduct distribution adaptation to eliminate the difference in feature distribution between domains in the new subspace. In light of the high dimension of features of MI-BCI illiteracy, random forest, which can effectively handle high-dimensional features without additional cross-validation, was employed as a classifier. The proposed method was validated on an open dataset. The experimental results show that that the method we proposed suits MI-BCI illiteracy and can reduce the inter-domain differences, resulting in a reduction in the performance degradation of both cross-subjects and cross-sessions.

Fine-Grained Ship Recognition from the Horizontal View Based on Domain Adaptation.

Ship recognition is a fundamental and essential step in maritime activities, and it can be widely used in maritime rescue, vessel management, and other applications. However, most studies conducted in this area use synthetic aperture radar (SAR) images and space-borne optical images, and those studies utilizing visible images are limited to the coarse-grained level. In this study, we constructed a fine-grained ship dataset with real images and simulation images that consisted of five categories of ships. To solve the problem of low accuracy in fine-grained ship classification with different angles in visible images, a network based on domain adaptation and a transformer was proposed. Concretely, style transfer was first used to reduce the gap between the simulation images and real images. Then, with the goal of utilizing the simulation images to execute classification tasks on the real images, a domain adaptation network based on local maximum mean discrepancy (LMMD) was used to align the different domain distributions. Furthermore, considering the innate attention mechanism of the transformer, a vision transformer (ViT) was chosen as the feature extraction module to extract the fine-grained features, and a fully connected layer was used as the classifier. Finally, the experimental results showed that our network had good performance on the fine-grained ship dataset with an overall accuracy rate of 96.0%, and the mean average precision (mAP) of detecting first and then classifying with our network was 87.5%, which also verified the feasibility of using images generated by computer simulation technology for auxiliary training.

Robust Latent Multi-Source Adaptation for Encephalogram-Based Emotion Recognition.

In practical encephalogram (EEG)-based machine learning, different subjects can be represented by many different EEG patterns, which would, in some extent, degrade the performance of extant subject-independent classifiers obtained from cross-subjects datasets. To this end, in this paper, we present a robust Latent Multi-source Adaptation (LMA) framework for cross-subject/dataset emotion recognition with EEG signals by uncovering multiple domain-invariant latent subspaces. Specifically, by jointly aligning the statistical and semantic distribution discrepancies between each source and target pair, multiple domain-invariant classifiers can be trained collaboratively in a unified framework. This framework can fully utilize the correlated knowledge among multiple sources with a novel low-rank regularization term. Comprehensive experiments on DEAP and SEED datasets demonstrate the superior or comparable performance of LMA with the state of the art in the EEG-based emotion recognition.

A New Method of Image Classification Based on Domain Adaptation.

Deep neural networks can learn powerful representations from massive amounts of labeled data; however, their performance is unsatisfactory in the case of large samples and small labels. Transfer learning can bridge between a source domain with rich sample data and a target domain with only a few or zero labeled samples and, thus, complete the transfer of knowledge by aligning the distribution between domains through methods, such as domain adaptation. Previous domain adaptation methods mostly align the features in the feature space of all categories on a global scale. Recently, the method of locally aligning the sub-categories by introducing label information achieved better results. Based on this, we present a deep fuzzy domain adaptation (DFDA) that assigns different weights to samples of the same category in the source and target domains, which enhances the domain adaptive capabilities. Our experiments demonstrate that DFDA can achieve remarkable results on standard domain adaptation datasets.

Classification logit two-sample testing by neural networks for differentiating near manifold densities.

The recent success of generative adversarial networks and variational learning suggests that training a classification network may work well in addressing the classical two-sample problem, which asks to differentiate two densities given finite samples from each one. Network-based methods have the computational advantage that the algorithm scales to large datasets. This paper considers using the classification logit function, which is provided by a trained classification neural network and evaluated on the testing set split of the two datasets, to compute a two-sample statistic. To analyze the approximation and estimation error of the logit function to differentiate near-manifold densities, we introduce a new result of near-manifold integral approximation by neural networks. We then show that the logit function provably differentiates two sub-exponential densities given that the network is sufficiently parametrized, and for on or near manifold densities, the needed network complexity is reduced to only scale with the intrinsic dimensionality. In experiments, the network logit test demonstrates better performance than previous network-based tests using classification accuracy, and also compares favorably to certain kernel maximum mean discrepancy tests on synthetic datasets and hand-written digit datasets.

On the capacity of deep generative networks for approximating distributions.

We study the efficacy and efficiency of deep generative networks for approximating probability distributions. We prove that neural networks can transform a low-dimensional source distribution to a distribution that is arbitrarily close to a high-dimensional target distribution, when the closeness is measured by Wasserstein distances and maximum mean discrepancy. Upper bounds of the approximation error are obtained in terms of the width and depth of neural network. Furthermore, it is shown that the approximation error in Wasserstein distance grows at most linearly on the ambient dimension and that the approximation order only depends on the intrinsic dimension of the target distribution. On the contrary, when f-divergences are used as metrics of distributions, the approximation property is different. We show that in order to approximate the target distribution in f-divergences, the dimension of the source distribution cannot be smaller than the intrinsic dimension of the target distribution.

A deep neural network with subdomain adaptation for motor imagery brain-computer interface.

BACKGROUND: The nonstationarity problem of EEG is very serious, especially for spontaneous signals, which leads to the poor effect of machine learning related to spontaneous signals, especially in related tasks across time, which correspondingly limits the practical use of brain-computer interface (BCI). OBJECTIVE: In this paper, we proposed a new transfer learning algorithm, which can utilize the labeled motor imagery (MI) EEG data at the previous time to achieve better classification accuracies for a small number of labeled EEG signals at the current time. METHODS: We introduced an adaptive layer into the full connection layer of a deep convolution neural network. The objective function of the adaptive layer was designed to minimize the Local Maximum Mean Discrepancy (LMMD) and the prediction error while minimizing the distance within each class (DWC) and maximizing the distance between classes within each domain (DBCWD). We verified the effectiveness of the proposed algorithm on two public datasets. RESULTS: The classification accuracy of the proposed algorithm was higher than other comparison algorithms, and the paired t-test results also showed that the performance of the proposed algorithm was significantly different from that of other algorithms. The results of the confusion matrix and feature visualization showed the effectiveness of the proposed algorithm. CONCLUSION: Experimental results showed that the proposed algorithm can achieve higher classification accuracy than other algorithms when there was only a small amount of labeled MI EEG data at the current time. It can be promising to be applied to the field of BCI.

The energy distance for ensemble and scenario reduction.

Scenario reduction techniques are widely applied for solving sophisticated dynamic and stochastic programs, especially in energy and power systems, but are also used in probabilistic forecasting, clustering and estimating generative adversarial networks. We propose a new method for ensemble and scenario reduction based on the energy distance which is a special case of the maximum mean discrepancy. We discuss the choice of energy distance in detail, especially in comparison to the popular Wasserstein distance which is dominating the scenario reduction literature. The energy distance is a metric between probability measures that allows for powerful tests for equality of arbitrary multivariate distributions or independence. Thanks to the latter, it is a suitable candidate for ensemble and scenario reduction problems. The theoretical properties and considered examples indicate clearly that the reduced scenario sets tend to exhibit better statistical properties for the energy distance than a corresponding reduction with respect to the Wasserstein distance. We show applications to a Bernoulli random walk and two real data-based examples for electricity demand profiles and day-ahead electricity prices. This article is part of the theme issue 'The mathematics of energy systems'.