Developing a reliable EEG-based Brain Computer Interface (BCI) system typically requires large and diverse training datasets, but collecting sufficient data remains challenging due to subject fatigue and inter-individual variability. To address these limitations, this study proposes a reinforcement learning-enhanced EEG diffusion (RLED) framework for adaptive data augmentation in endogenous EEG tasks, with a focus on motor imagery and emotion recognition. The framework integrates a reinforcement learning mechanism to dynamically regulate the diffusion training process and achieve a flexible balance among temporal, spectral, and class-related features. Experiments on four datasets demonstrate that the proposed method generates high-quality synthetic EEG signals and consistently improves classification performance. These findings show that the proposed RLED framework may serve as a promising tool for EEG data augmentation and generalization in practical BCI applications.

Brain-computer interfaces (BCIs) rely on accurately decoding electroencephalography (EEG) motor imagery (MI) signals for effective device control. Graph neural networks (GNNs) outperform convolutional neural networks (CNNs) in this regard, by leveraging the spatial relationships between EEG electrodes through adjacency matrices. The EEG graph lottery ticket framework, EEG_GLT-Net, featuring the state-of-the-art (SOTA) EEG_GLT adjacency matrix method, has notably enhanced EEG MI signal classification, evidenced by an average accuracy of 83.95% across 20 subjects on the PhysioNet dataset. This significantly exceeds the 76.10% accuracy rate achieved using the Pearson correlation coefficient (PCC) method in the same framework. In this research, we advance the field by applying a reinforcement learning (RL) approach to the classification of EEG MI signals. Our innovative method empowers the RL agent, enabling not only the classification of EEG MI data points with higher accuracy but effective identification of EEG MI data points that are less distinct. We present the EEG_RL-Net, an enhancement of the EEG_GLT-Net framework, which incorporates the trained EEG_GCN Block from EEG_GLT-Net at an adjacency matrix density of 13.39% alongside the RL-centric dueling deep Q network (Dueling DQN) block. The EEG_RL-Net model showcases exceptional classification performance, achieving an unprecedented average accuracy of 96.40% across 20 subjects within 25 ms. This model illustrates the transformative effect of the RL in EEG MI time point classification.

Brain-Computer Interfaces (BCIs) rely on accurately decoding electroencephalography (EEG) motor imagery (MI) signals for effective device control. Graph Neural Networks (GNNs) outperform Convolutional Neural Networks (CNNs) in this regard, by leveraging the spatial relationships between EEG electrodes through adjacency matrices. The EEG_GLT-Net framework, featuring the state-of-the-art EEG_GLT adjacency matrix method, has notably enhanced EEG MI signal classification, evidenced by an average accuracy of 83.95% across 20 subjects on the PhysioNet dataset. This significantly exceeds the 76.10% accuracy rate achieved using the Pearson Correlation Coefficient (PCC) method within the same framework. In this research, we advance the field by applying a Reinforcement Learning (RL) approach to the classification of EEG MI signals. Our innovative method empowers the RL agent, enabling not only the classification of EEG MI data points with higher accuracy, but effective identification of EEG MI data points that are less distinct. We present the EEG_RL-Net, an enhancement of the EEG_GLT-Net framework. The EEG_RL-Net model showcases exceptional classification performance, achieving an unprecedented average accuracy of 95.36% across 20 subjects within 18.20 milliseconds. This model illustrates the transformative effect of the RL in EEG MI real time point classification.

As the simplest feature extraction, traditional hand-crafted transient-state features have been widely used in the area of electronic noses (e-noses). However, the influence of noise in the calculation of numerical differentiation leads to inaccuracy and instability in extracting these features. To tackle this issue, a novel numerical differentiation algorithm is proposed, which uses kernel-based regularization. The proposed method can provide accurate and stable transient-state features by directly estimating high-order derivatives from the noise-contaminated sensor’s reading. The feature representation is a prerequisite for the good performance of e-noses. Nevertheless, it should be noted that this performance in real applications can still be affected by other factors, such as sensor drift and the disturbance of nontarget odors. These issues can be addressed by applying a framework of domain adaptation and one-class classification. The proposed method and the adopted framework are verified in a field experiment, which identifies the odor of four targets and two disturbance whiskies measured by a self-designed e-nose system. The classification accuracy with traditional features is improved from$\mathbf{71.90\%}$to$\mathbf{86.36\%}$, showing the good potential of the proposed method for application in the area of e-noses.

Accurate assessment of fish quality is difficult in practice due to the lack of trusted fish provenance and quality tracking information. Working with Sydney Fish Market (SFM), we develop a Blockchain-enabled fish provenance and quality tracking (BeFAQT) system. A multilayer Blockchain architecture based on attribute-based encryption (ABE) is proposed to tackle the privacy issue caused by applying Blockchain to secure supply chain data and achieve trusted and confidential data sharing among parties in fish supply chains. An Internet-of-Things (IoT) chain saves encrypted fish provenance and quality tracking data, and an ABE chain is specifically designed for the access control to the data in the IoT chain. Latest IoT and artificial intelligence (AI) technologies, including NarrowBand-IoT, image processing, and biosensing, are developed for fish origin proof, supply chain tracking, and objective fish quality assessment. As proven by field trials with SFM and a local fish supply chain, the BeFAQT is able to provide trusted and comprehensive fish provenance and quality tracking information in real time.

In the electronic nose (e-nose), a stable feature representation of the gas sensor’s response is a key step to realize subsequent odor identification algorithms. However, the noises in gas sensors hinder the acquisition of such features. In order to solve this problem, this article proposes a stable feature extraction algorithm which takes the impulse response of the e-nose system as the feature. The impulse response is estimated from a nonparametric model constrained by a multiscale wavelet kernel regularization matrix. The kernel regularization matrix equips the proposed feature extraction method with an ability in resistance to random noise. A numerical experiment proves that compared with single-scale kernel regularization, the use of multiscale wavelet kernel helps to achieve more stable and accurate impulse response estimation. Then, a field experiment is conducted to demonstrate the performance of the proposed features. This experiment aims to identify four different whiskies measured by a self-designed e-nose with four commercial gas sensors. Under the framework of transfer learning, the classification result based on the proposed features outperforms those using other considered features. The accuracy of whisky identification reaches 92.00%, showing a good potential of applying the proposed feature representations in the area of e-noses.

Differentiable ARchiTecture Search(DARTS) has recently become the mainstream in the neural architecture search (NAS) due to its efficiency and simplicity. With a gradient-based bi-level optimization, DARTS alternately optimizes the inner model weights and the outer architecture parameter in a weight-sharing supernet. A key challenge to the scalability and quality of the learned architectures is the need for differentiating through the inner-loop optimisation. While much has been discussed about several potentially fatal factors in DARTS, the architecture gradient, a.k.a. hypergradient, has received less attention. In this paper, we tackle the hypergradient computation in DARTS based on the implicit function theorem, making it only depends on the obtained solution to the inner-loop optimization and agnostic to the optimization path. To further reduce the computational requirements, we formulate a stochastic hypergradient approximation for differentiable NAS, and theoretically show that the architecture optimization with the proposed method is expected to converge to a stationary point. Comprehensive experiments on two NAS benchmark search spaces and the common NAS search space verify the effectiveness of our proposed method. It leads to architectures outperforming, with large margins, those learned by the baseline methods.

Lung cancer is the leading cause of cancer death worldwide. The critical reason for the deaths is delayed diagnosis and poor prognosis. With the accelerated development of deep learning techniques, it has been successfully applied extensively in many real-world applications, including health sectors such as medical image interpretation and disease diagnosis. By combining more modalities that being engaged in the processing of information, multimodal learning can extract better features and improve the predictive ability. The conventional methods for lung cancer survival analysis normally utilize clinical data and only provide a statistical probability. To improve the survival prediction accuracy and help prognostic decision-making in clinical practice for medical experts, we for the first time propose a multimodal deep learning framework for non-small cell lung cancer (NSCLC) survival analysis, named DeepMMSA. This framework leverages CT images in combination with clinical data, enabling the abundant information held within medical images to be associate with lung cancer survival information. We validate our model on the data of 422 NSCLC patients from The Cancer Imaging Archive (TCIA). Experimental results support our hypothesis that there is an underlying relationship between prognostic information and radiomic images. Besides, quantitative results show that our method could surpass the state-of-the-art methods by 4% on concordance.

One-shot neural architecture search (NAS) has recently become mainstream in the NAS community because it significantly improves computational efficiency through weight sharing. However, the supernet training paradigm in one-shot NAS introduces catastrophic forgetting, where each step of the training can deteriorate the performance of other architectures that contain partially-shared weights with current architecture. To overcome this problem of catastrophic forgetting, we formulate supernet training for one-shot NAS as a constrained continual learning optimization problem such that learning the current architecture does not degrade the validation accuracy of previous architectures. The key to solving this constrained optimization problem is a novelty search based architecture selection (NSAS) loss function that regularizes the supernet training by using a greedy novelty search method to find the most representative subset. We applied the NSAS loss function to two one-shot NAS baselines and extensively tested them on both a common search space and a NAS benchmark dataset. We further derive three variants based on the NSAS loss function, the NSAS with depth constrain (NSAS-C) to improve the transferability, and NSAS-G and NSAS-LG to handle the situation with a limited number of constraints. The experiments on the common NAS search space demonstrate that NSAS and it variants improve the predictive ability of supernet training in one-shot NAS with remarkable and efficient performance on the CIFAR-10, CIFAR-100, and ImageNet datasets. The results with the NAS benchmark dataset also confirm the significant improvements these one-shot NAS baselines can make.

This article investigates deep neural networks (DNNs)-based lung nodule classification with hyperparameter optimization. Hyperparameter optimization in DNNs is a computationally expensive problem, and a surrogate-assisted evolutionary algorithm has been recently introduced to automatically search for optimal hyperparameter configurations of DNNs, by applying computationally efficient surrogate models to approximate the validation error function of hyperparameter configurations. Different from existing surrogate models adopting stationary covariance functions (kernels) to measure the difference between hyperparameter points, this article proposes a nonstationary kernel that allows the surrogate model to adapt to functions whose smoothness varies with the spatial location of inputs. A multilevel convolutional neural network (ML-CNN) is built for lung nodule classification, and the hyperparameter configuration is optimized by the proposed nonstationary kernel-based Gaussian surrogate model. Our algorithm searches with a surrogate for optimal setting via a hyperparameter importance-based evolutionary strategy, and the experiments demonstrate our algorithm outperforms manual tuning and several well-established hyperparameter optimization methods, including random search, grid search, the tree-structured parzen estimator (TPE) approach, Gaussian processes (GP) with stationary kernels, and the recently proposed hyperparameter optimization via RBF and dynamic (HORD) coordinate search.

One-Shot Neural Architecture Search (NAS) significantly improves the computational efficiency through weight sharing. However, this approach also introduces multi-model forgetting during the supernet training (architecture search phase), where the performance of previous architectures degrade when sequentially training new architectures with partially-shared weights. To overcome such catastrophic forgetting, the state-of-the-art method assumes that the shared weights are optimal when jointly optimizing a posterior probability. However, this strict assumption is not necessarily held for One-Shot NAS in practice. In this paper, we formulate the supernet training in the One-Shot NAS as a constrained optimization problem of continual learning that the learning of current architecture should not degrade the performance of previous architectures during the supernet training. We propose a Novelty Search based Architecture Selection (\textbf{NSAS}) loss function and demonstrate that the posterior probability could be calculated without the strict assumption when maximizing the diversity of the selected constraints. A greedy novelty search method is devised to find the most representative subset to regularize the supernet training. We apply our proposed approach to two One-Shot NAS baselines, random sampling NAS (RandomNAS) and gradient-based sampling NAS (GDAS). Extensive experiments demonstrate that our method enhances the predictive ability of the supernet in One-Shot NAS and achieves remarkable performance on CIFAR-10, CIFAR-100, and PTB with efficiency.

One-Shot Neural architecture search (NAS) has received wide attentions due to its computational efficiency. Most state-of-the-art One-Shot NAS methods use the validation accuracy based on inheriting weights from the supernet as the stepping stone to search for the best performing architecture, adopting a bilevel optimization pattern with assuming this validation accuracy approximates to the test accuracy after re-training. However, recent works have found that there is no positive correlation between the above validation accuracy and test accuracy for these One-Shot NAS methods, and this reward based sampling for supernet training also entails the rich-get-richer problem. To handle this deceptive problem, this paper presents a new approach, Efficient Novelty-driven Neural Architecture Search, to sample the most abnormal architecture to train the supernet. Specifically, a single-path supernet is adopted, and only the weights of a single architecture sampled by our novelty search are optimized in each step to reduce the memory demand greatly. Experiments demonstrate the effectiveness and efficiency of our novelty search based architecture sampling method.

Recent works on One-Shot Neural Architecture Search (NAS) mostly adopt a bilevel optimization scheme to alternatively optimize the supernet weights and architecture parameters after relaxing the discrete search space into a differentiable space. However, the non-negligible incongruence in their relaxation methods is hard to guarantee the differentiable optimization in the continuous space is equivalent to the optimization in the discrete space. Differently, this paper utilizes a variational graph autoencoder to injectively transform the discrete architecture space into an equivalently continuous latent space, to resolve the incongruence. A probabilistic exploration enhancement method is accordingly devised to encourage intelligent exploration during the architecture search in the latent space, to avoid local optimal in architecture search. As the catastrophic forgetting in differentiable One-Shot NAS deteriorates supernet predictive ability and makes the bilevel optimization inefficient, this paper further proposes an architecture complementation method to relieve this deficiency. We analyze the effectiveness of the proposed method, and a series of experiments have been conducted to compare the proposed method with state-of-the-art One-Shot NAS methods.

Bayesian optimization (BO) has been broadly applied to computational expensive problems, but it is still challenging to extend BO to high dimensions. Existing works are usually under strict assumption of an additive or a linear embedding structure for objective functions. This paper directly introduces a supervised dimension reduction method, Sliced Inverse Regression (SIR), to high dimensional Bayesian optimization, which could effectively learn the intrinsic sub-structure of objective function during the optimization. Furthermore, a kernel trick is developed to reduce computational complexity and learn nonlinear subset of the unknowing function when applying SIR to extremely high dimensional BO. We present several computational benefits and derive theoretical regret bounds of our algorithm. Extensive experiments on synthetic examples and two real applications demonstrate the superiority of our algorithms for high dimensional Bayesian optimization.

This paper presents an adaptive neural network (NN) control of a two-degree-of-freedom manipulator driven by an electrohydraulic actuator. To restrict the system output in a prescribed performance constraint, a weighted performance function is designed to guarantee the dynamic and steady tracking errors of joint angle in a required accuracy. Then, a radial-basis-function NN is constructed to train the unknown model dynamics of a manipulator by traditional backstepping control (TBC) and obtain the preliminary estimated model, which can replace the preknown dynamics in the backstepping iteration. Furthermore, an adaptive estimation law is adopted to self-tune every trained-node weight, and the estimated model is online optimized to enhance the robustness of the NN controller. The effectiveness of the proposed control is verified by comparative simulation and experimental results with Proportional-integral-derivative and TBC methods.

The distribution power subsystems are usually interconnected to each other, so the design of the interconnected optimal filtering algorithm for distributed state estimation is a challenging task. Driven by this motivation, this paper proposes a novel consensus filter based dynamic state estimation algorithm with its convergence analysis for modern power systems. The novelty of the scheme is that the algorithm is designed based on the mean squared error and semidefinite programming approaches. Specifically, the optimal local gain is computed after minimizing the mean squared error between the true and estimated states. The consensus gain is determined by a convex optimization process with a given suboptimal local gain. Furthermore, the convergence of the proposed scheme is analyzed after stacking all the estimation error dynamics. The Laplacian operator is used to represent the interconnected filter structure as a compact error dynamic for deriving the convergence condition of the algorithm. The developed approach is verified by using the renewable microgrid. It shows that the distributed scheme being explored is effective as it takes only 0.00004 seconds to properly estimate the system states and does not need to transmit the remote sensing signals to the central estimator.

Due to dramatically rising energy demand worldwide power system is often run near the operational and technical limits, where unexpected trivial disturbances can cause possibly massive blackouts. Cyber attacks on smart grid communication networks are one of the impending threats to cause large-scale cascading outage. In contrast to the traditional cyber attack protection techniques, this paper presents a recursive systematic convolutional code based defending technique from the signal processing perspective. This code introduces redundancy in the system for protecting the grid information. Furthermore, an optimal control law is designed to stabilize the power network. Specifically, the performance index for control is converted to a convex semidefinite programming problem. The proposed controller can work well for any initial values. The efficacy of the developed approach is verified through numerical simulations. Results show that the proposed strategy has stronger attack protection performance and the controller can stabilize the grid in a fairly short time. This approach provides a fundamental framework for the design of the smart grid energy management system and reliable communication infrastructure scheme with renewable integration applications.

The penetration of renewable distributed energy resources such as wind turbine has been dramatically integrated in distribution networks due to clean, sustainable, and economical green energy. Due to intermittent wind speed, the power generation patterns vary which can risk the distribution network operation. So, it is intrinsically required to monitor the wind turbines in a distributed way. This paper presents an adaptive-then-combine distributed dynamic approach for monitoring the grid under lossy communication link between the wind turbines and energy management system. Firstly, the wind turbine is represented by a state-space linear equation, with sensors deployed to obtain system state information. Based on the mean squared error principle, an adaptive approach is proposed to estimate the local state information. The global estimation is designed by combining estimation results with weighting factors which are calculated by minimizing estimation error covariance based on semi-definite programming. Finally, the convergence analysis indicates that estimation error is gradually decreased, so the estimated state converges to the actual state. The efficacy of the developed approach is verified using the turbine model. The research is valuable for green energy technologies, households and interconnected industrial information societies, and also provides the knowledge towards a smart and secure green energy future.

The penetration of renewable distributed energy resources, such as wind turbine, has been dramatically increased in distribution networks. Due to the intermittent property, the wind power generation patterns vary, which may risk distribution network operations. So, it is intrinsically necessary to monitor wind turbines in a distributed way. This paper presents an adaptive-then-combine distributed dynamic approach for monitoring the grid under lossy communication links between the wind turbines and energy management system. First, the wind turbine is represented by a state-space linear model, with sensors deployed to obtain the system state information. Based on the mean squared error principle, an adaptive approach is proposed to estimate the local state information. The global estimation is designed by combining estimation results with weighting factors which are calculated by minimizing the estimation error covariance based on semidefinite programming. Finally, the convergence analysis indicates that the estimation error is gradually decreased, so the estimated state converges to the actual state. The efficacy of the developed approach is verified using the wind turbine and the IEEE 6-bus distribution system.

Ultracapacitors (UCs) are an important energy storage technology in automotive and grid applications. They have several advantages, including high power density and extraordinarily long lifespan. Accurate State-of-Charge (SOC) tracking of UCs is critical for the reliability, resilience, and safety in system operation. This paper presents a novel robust H infinity observer in order to realize the SOC estimation of a UC in real time. It is computationally efficient because the observer gain involved in the real-time computation can be readily synthesized offline. In comparison to state-of-the-art Kalman filtering (KF), the developed robust scheme can ensure high estimation accuracy even without prior knowledge of the process and noise measurement statistical properties. More significantly, the H infinity observer proves to be more robust and tolerant to modeling uncertainties arising from the change of operating conditions and/or cell health status. These benefits are experimentally verified.

This paper describes a new algorithm, called Ensemble Rapid Centroid Estimation (ERCE), designed to handle large-scale non-convex cluster optimization tasks, and estimate the number of clusters with quasi-linear complexity. ERCE stems from a recently developed Rapid Centroid Estimation (RCE) algorithm. RCE was originally developed as a lightweight simplification of the Particle Swarm Clustering (PSC) algorithm. RCE retained the quality of PSC, greatly reduced the computational complexity, and increased the stability. However, RCE has certain limitations with respect to complexity, and is unsuitable for non-convex clusters. The new ERCE algorithm presented here addresses these limitations.

Prior work suggests that particle swarm clustering (PSC) can be a powerful tool for solving clustering problems. This paper reviews parts of the PSC algorithm, and shows how and why a new class of algorithms is proposed in an attempt to improve the efficiency and repeatability of PSC. This new implementation is referred to as rapid centroid estimation (RCE). RCE simplifies the update rules of PSC, and greatly reduces computational complexity by enhancing the efficiency of the particle trajectories. On benchmark evaluations with an artificial dataset that has 80 dimensions and a volume of 5000, the RCE variants have iteration times of less than 0.1 s, which compares to iteration times of 2 s for PSC and modified PSC (mPSC). On UC Irvine (UCI) machine learning benchmark datasets, the RCE variants are much faster than PSC and mPSC, and produce clusters with higher purity and greatly improved optimization speeds. For example, the RCE variants are more than 100 times faster than PSC and mPSC on the UCI breast cancer dataset. It can be concluded that the RCE variants are leaner and faster than PSC and mPSC, and that the new optimization strategies also improve clustering quality and repeatability.

Clustering can be especially effective where the data is irregular, noisy and/or not differentiable. A major obstacle for many clustering techniques is that they are computationally expensive, hence limited to smaller data volume and dimension. We propose a lightweight swarm clustering solution called Rapid Centroid Estimation (RCE). Based on our experiments, RCE has significantly quickened optimization time of its predecessors, Particle Swarm Clustering (PSC) and Modified Particle Swarm Clustering (mPSC). Our experimental results show that on benchmark datasets, RCE produces generally better clusters compared to PSC, mPSC, K-means and Fuzzy C-means. Compared with K-means and Fuzzy C-means which produces clusters with 62% and 55% purities on average respectively, thyroid dataset has successfully clustered on average 71% purity in 14.3 seconds.

Data clustering is a process where a set of data points is divided into groups of similar points. Recent approaches for data clustering have seen the development of unsupervised learning algorithms based on Particle Swarm Optimization (PSO) techniques. These include Particle Swarm Clustering (PSC) and Modified PSC (mPSC) algorithms for solving clustering problems. However, the PSC and mPSC algorithms tend to be computationally expensive when applied to datasets that have higher levels of dimensionality and large volumes. This paper presents a novel and more efficient swarm clustering strategy we call Rapid Centroid Estimation (RCE). We compare the performance of RCE with the performance of PSC and mPSC in several ways including complexity analyses and particle behavior analyses. Our benchmark testing suggests that RCE can reach a solution 274 times quicker than PSC and 270 times quicker than mPSC for a clustering task where the dataset has a dimension of 80 and a volume of 500. We also investigated particle behaviors on two-class two-dimensional datasets with volume of 500, presenting 250 data for each well-separated class with known Gaussian centers. We found that RCE converged to the appropriate centers at 70 updates on average, compared to 19802 updates for PSC and 23006 updates for mPSC. An ANOVA indicates RCE is significantly faster than both PSC and mPSC.

Teaching multivariable control usually involves a certain level of mathematical sophistication and hence requires some labaratorial exemplification of the material given in formal lectures. This paper reports on a hands-on approach to multi-variable control education via the implementation of a model predictive controller on a two-input, two output coupled drive apparatus. This scaled-down system represents many industrial processes while provides an excellent set-up for demonstrating the cross-coupled effects in multi-input multi-output systems. Here, a model predictive controller (MPC) is developed and implemented on the basis of a constrained optimization problem to show control performance via the belt tension and velocity outputs, demonstrate the decoupling capability, and also illustrate such issues as control input saturation, the selection of operating point, reference inputs, and system robustness to external disturbance and varying parameters. The implementation is based on Labview and MATLAB Model Predictive Control Toolbox.

This paper proposes an integral controller design scheme for nonlinear systems based on optimal control and the passivity theorem in order to suppress the effect of external disturbances. The main strategy is to augment an optimal controller with a PI type controller. To guarantee the proposed controller has a desired stability margin, the passivity-based design method is introduced. Here, the inverse optimal control technique is employed to avoid the need of solving a Hamilton-Jacobi equation. An illustrative example is given to show the design procedure and the controller effectiveness.