Research on continual learning (CL) mainly focuses on data represented in the Euclidean space, while research on graph-structured data is scarce. Furthermore, most graph learning models are tailored for static graphs. However, graphs usually evolve continually in the real world. Catastrophic forgetting also emerges in graph learning models when being trained incrementally. This leads to the need to develop robust, effective and efficient continual graph learning approaches. Continual graph learning (CGL) is an emerging area aiming to realize continual learning on graph-structured data. This survey is written to shed light on this emerging area. It introduces the basic concepts of CGL and highlights two unique challenges brought by graphs. Then it reviews and categorizes recent state-of-the-art approaches, analyzing their strategies to tackle the unique challenges in CGL. Besides, it discusses the main concerns in each family of CGL methods, offering potential solutions. Finally, it explores the open issues and potential applications of CGL.

While deep learning-based robotic grasping technology has demonstrated strong adaptability, its computational complexity has also significantly increased, making it unsuitable for scenarios with high real-time requirements. Therefore, we propose a low computational complexity and high accuracy model named vision mamba grasping network (VMGNet) for 2-D planar robotic grasping. For the first time, we introduce the visual state space (VSS) into the robotic grasping field to achieve linear computational complexity, thereby greatly reducing the model's computational cost. Meanwhile, to improve the accuracy of the model, we propose an efficient and lightweight multiscale feature fusion module, named fusion bridge module (FBM), to extract and fuse information at different scales. We also present a new loss function calculation method to enhance the importance differences between subtasks, improving the model's fitting ability. Experiments show that VMGNet has only 8.7 gigafloating point operations and an inference time (IT) of 8.1 ms on our devices. VMGNet also achieved state-of-the-art (SOTA) performance on the Cornell and Jacquard public datasets. To validate VMGNet's effectiveness in practical applications, we conducted real grasping experiments in multiobject scenarios, and VMGNet achieved an excellent performance with a 94.4% success rate in real-world grasping tasks. The code and video for the real-world robotic grasping experiments are available at https://github.com/momo7941/VMGNet

Predicting the performance of photovoltaic (PV) systems is crucial for optimizing renewable energy utilization. However, traditional time-series methods focus only on temporal patterns, overlooking environmental variations, while dynamic conditions such as partial shading further complicate power prediction. To address this shading-induced variability, we propose a Temporal and Environment-Informed Prediction (TEIP) framework, which enhances PV power prediction by dynamically structuring temporal and environmental data through a novel multi-spatial attention LSTM (MSAL) network. This framework utilizes the TE matrix to capture structured environmental conditions over time, including the variability caused by partial shading. A dual-branch MSAL model uniquely processes environmental data through spatial feature extraction, which is then sequentially processed by LSTM to capture temporal dependencies. This hierarchical spatial-temporal processing enables dynamic adaptation to changing environmental conditions. Experimental results show the framework achieves superior prediction accuracy with R2 of 0.952 under sunny conditions, significantly outperforming traditional approaches. The framework demonstrates exceptional robustness by maintaining consistent performance (R2 of 0.948) even under challenging cloudy conditions, validating its effectiveness for real-world applications.

The prediction of bifacial photovoltaic (bPV) system performance under variable conditions has persistently challenged researchers and practitioners alike, largely due to the unstable and imprecise irradiance measurements and the extensive training processes required for machine learning-based methods. Addressing these issues, this study introduces an innovative digital twin system that integrates a novel circuit-long short-term memory (LSTM) model with the newly proposed triangle-shading pattern estimation method, eliminating dependencies on direct irradiance measurements and historical data. Our approach uniquely combines the adaptability of LSTM networks with circuit models, facilitating real-time power prediction with unprecedented accuracy and efficiency. Comprehensive evaluations across various shading scenarios demonstrate the proposed model's superior performance, consistently reducing mean absolute error, mean squared error, and root mean squared error by over 50% compared with existing methods. This breakthrough offers a scalable, cost-effective solution for optimizing the deployment and management of bPV systems, marking a significant advancement in the field of photovoltaic research.

In past work (Onokpasa, Wild, Wong, DCC 2023), we showed that (a) for joint compression of RNA sequence and structure, stochastic context-free grammars are the best known compressors and (b) that grammars which have better compression ability also show better performance in ab initio structure prediction. Previous grammars were manually curated by human experts. In this work, we develop a framework for automatic and systematic search algorithms for stochastic grammars with better compression (and prediction) ability for RNA. We perform an exhaustive search of small grammars and identify grammars that surpass the performance of human-expert grammars.

In this work, we study a scheduling problem with explorable uncertainty. Each job comes with an upper limit of its processing time, which could be potentially reduced by testing the job, which also takes time. The objective is to schedule all jobs on a single machine with a minimum total completion time. The challenge lies in deciding which jobs to test and the order of testing/processing jobs. The online problem was first introduced with unit testing time [5, 6] and later generalized to variable testing times [1]. For this general setting, the upper bounds of the competitive ratio are shown to be 4 and 3.3794 for deterministic and randomized online algorithms [1]; while the lower bounds for unit testing time stands [5, 6], which are 1.8546 (deterministic) and 1.6257 (randomized). We continue the study on variable testing times setting. We first enhance the analysis framework in [1] and improve the competitive ratio of the deterministic algorithm in [1] from 4 to 1+√(2)≈ 2.4143 . Using the new analysis framework, we propose a new deterministic algorithm that further improves the competitive ratio to 2.316513. The new framework also enables us to develop a randomized algorithm improving the expected competitive ratio from 3.3794 to 2.152271.

Water resources have spurred the need for advanced maritime technologies, particularly for effective exploration and harnessing. Traditional maritime resource investigation methodologies are hampered by challenges such as elevated costs and risks. Unmanned Surface Vehicles (USVs) have emerged as a transformative solution, offering applications previously unattainable or challenging with conventional ships. An essential element in enhancing the capabilities of USVs is the incorporation of “Digital Twin” (DT) technology, a digital representation mirroring a system's real-world attributes and behaviours. While DT's application has gained traction in other sectors, notably in autonomous vehicles and unmanned aerial vehicles, a focused exploration of its integration in USVs remains sparse. This paper bridges this gap by offering a comprehensive survey of DT within the context of USVs. We review prevailing DT methodologies, delve into multifaceted perspectives surrounding DT-including definitions, classifications and key features, and highlight the prospective directions and challenges for its future application in USVs. Through this study, we underscore the potential of DT in enhancing the capabilities, productivity, and sustainability of USVs in complex maritime environments.

In this paper, we use the biological domain knowledge incorporated into stochastic models for ab initio RNA secondary-structure prediction to improve the state of the art in joint compression of RNA sequence and structure data (Liu et al., BMC Bioinformatics, 2008). Moreover, we show that, conversely, compression ratio can serve as a cheap and robust proxy for comparing the prediction quality of different stochastic models, which may help guide the search for better RNA structure prediction models. Our results build on expert stochastic context-free grammar models of RNA secondary structures (Dowell & Eddy, BMC Bioinformatics, 2004; Nebel & Scheid, Theory in Bio-sciences, 2011) combined with different (static and adaptive) models for rule probabilities and arithmetic coding. We provide a prototype implementation and an extensive empirical evaluation, where we illustrate how grammar features and probability models affect compression ratios.

Artificial intelligence (AI) systems are becoming wiser, even surpassing human performances in some fields, such as image classification, chess, and Go. However, most high-performance AI systems, such as deep learning models, are black boxes (i.e., only system inputs and outputs are visible, but the internal mechanisms are unknown) and, thus, are notably challenging to understand. Thereby a system with better explainability is needed to help humans understand AI. This paper proposes a dual-track AI approach that uses reinforcement learning to supplement fine-grained deep learning-based sentiment classification. Through lifelong machine learning, the dual-track approach can gradually become wiser and realize high performance (while keeping outstanding explainability). The extensive experimental results show that the proposed dual-track approach can provide reasonable fine-grained sentiment analyses to product reviews and remarkably achieve a 133% promotion of the Macro-F1 score on the Twitter sentiment classification task and a 27.12% promotion of the Macro-F1 score on an Amazon iPhone 11 sentiment classification task, respectively.

Accurate estimated model parameters are of great significance to the modeling, characterization, and optimization of photovoltaic (PV) systems. The existing parameter estimation algorithms cannot predict parameters of partially shaded PV systems. This paper proposes a low-cost shading pattern identification algorithm to estimate essential environmental information for partial shading conditions (PSCs). With the help of a comprehensive PV model, the optimal environmental factors and model parameters can be obtained via a proposed exploration-exploitation-Jaya (EE-Jaya) algorithm, which has the capacity to balance the exploration and exploitation of candidate solutions. The proposed method is tested experimentally to evaluate its performance under a variety of test conditions. Results show that it can balance the local and global search and enable the developed model to fit the measured data well, outperforming other state-of-the-art global optimization algorithms in terms of accuracy and reliability.

Bifacial photovoltaic (PV) modules show higher output than monofacial PV modules, and therefore, an accurate model is of significance to bifacial PV system management. Traditional bifacial PV models treat the bifacial PV cell as two monofacial PV cells in parallel, and most of these studies do not address the effects of different irradiance and materials at both sides on total output in simulation. This paper presents a simplified bifacial alpha-beta single double diode model. The measurement error rate alpha is proposed to demonstrate deviation between measured data and real data which happens during the I-V or irradiance measurement. A bifacial contribution rate beta is introduced to present the ratio of output from two sides in the total generated current resulting from the different irradiance and materials. The experiment results show that the proposed model obtained the lowest errors compared with existing bifacial PV models.

Reading product reviews is the best way to know the product quality in online shopping. Due to the huge review number, customers and merchants need product analysis algorithms to help with quality analysis. Current researches use sentiment analysis to replace quality analysis. However, it has a significant drawback. This paper proves that the sentiment-based analysis algorithms are insufficient for online product quality analysis. They ignore the relationship between aspect and its description and cannot detect noise (unrelated description). So this paper raises a Lifelong Product Quality Analysis algorithm LPQA to learn the relationship between aspects. It can detect the noise and improve the opinion classification performance. It improves the classification F1 score to 77.3% on the Amazon iPhone dataset and 69.99% on Semeval Laptop dataset.

When Photovoltaic (PV) strings are under partial shading conditions, the current-voltage (I-V) characteristic curve shows multiple stairs, which brings challenges to power management, especially in real-time. In this paper, a characteristic matrix is used to characterize the multi-peak current-voltage characteristics curve. In addition, a light characteristics identification algorithm is proposed to extract the characteristic matrix from the measured I-V data. The proposed algorithm only requires limited points in the I-V curve, and the information of turning points on the I-V curve can be estimated. The technique has been evaluated under four environmental conditions in experiments, and the results verify the accuracy of the proposed method.

Humans always play the role of a teacher to AI in the last decades. One day, AI will become wiser than humans in various fields. At that time, humans need to learn from AI to improve themselves. However, most of the current high-performance machine learning models are black boxes and challenging to understand. So a system with better explainability is needed to help humans to understand AI. Therefore, the authors proposed a double-track approach to use expert systems to supplement the current machine learning paradigm to solve this problem. Under lifelong machine learning, the double-track approach can be wiser and wiser and achieve high performance but keeps outstanding explainability.

Artificial intelligence (AI) has entered tourism and become a new service in a tour guide. AI technology can help tourism by providing customized services and attracting visitors to fight with the crisis of the COVID-19 epidemic. This paper introduces how AI tour guide services contribute to tourism and its main issues. The future development of AI tour guides also was discussed at the end and the authors believe lifelong machine learning is the key to developing AI tour guides.

This paper presents a rapid response system architecture for the distributed management of warehouses in logistics by applying the concept of tiered edge computing. A tiered edge node architecture is proposed for the system to process computing tasks of different complexity, and a corresponding rapid response algorithm is introduced. The paper emphasizes the classification of abstracted outlier sensing data which could better match different sensing types and transplant to various application fields. A software-defined simulation is used to evaluate the system performance on response time and response accuracy, from which it can be concluded that common predefined emergency cases can be detected and responded to, rapidly.

We study a scheduling problem arising in demand response management in smart grid. Consumers send in power requests with a flexible feasible time interval during which their requests can be served. The grid controller, upon receiving power requests, schedules each request within the specified interval. The electricity cost is measured by a convex function of the load in each timeslot. The objective is to schedule all requests with the minimum total electricity cost. Previous work has studied cases where jobs have unit power requirement and unit duration. We extend the study to arbitrary power requirement and duration, which has been shown to be NP-hard. We give the first online algorithm for the general problem and prove that the problem is fixed parameter tractable. We also show that the online algorithm is the best-possible in an asymptotically sense when the objective is to minimize the peak load. In addition, we observe that the classical non-preemptive machine minimization problem is a special case of the smart grid problem with min-peak objective and show that we can achieve the best-possible competitive ratio in an asymptotically sense when solving the non-preemptive machine minimization problem.

Millimeter wave (mmWave) communication not only provides ultra-high speed radio access but is also ideally suited for efficient and flexible wireless backhauling. Specifically for dense deployments, a mmWave macro base station (MBS) that serves a large number of mmWave micro base stations (μBSs) is much more cost effective than legacy cellular architectures which connect μBSs to the core network through fibers. In addition, μBSs can cooperate with each other by acting as relay nodes. The directional nature of mmWave communication allows for spatial reuse, even in the presence of interference, which can be exploited to optimize mmWave wireless backhaul performance. The optimization opportunistically prioritizes the use of good connections at the MBS and further leverages compact and concurrent transmissions between μBS. Relays and directional antennas speed up communication, but increase the complexity of the scheduling problem. In this work, we study the mmWave backhaul scheduling problem and derive an MILP formulation for it as well as upper and lower bounds. We prove that the problem is NP-hard and can be approximated, but only if interference is negligible. By means of numerical simulations, we compare theoretical results with heuristics in small system sizes. Results validate the analysis and demonstrate the high performance of our heuristics in realistic cellular settings.

Benefiting from the rapid development of big data and high-performance computing, more data is available and more tasks could be solved by machine learning now. Even so, it is still difficult to maximum the power of big data due to each dataset is isolated with others. Although open source datasets are available, algorithms' performance is asymmetric with the data volume. Hence, the AI community wishes to raise a symmetric continuous learning architecture which can automatically learn and adapt to different tasks. Such a learning architecture also is commonly called as lifelong machine learning (LML). This learning paradigm could manage the learning process and accumulate meta-knowledge by itself during learning different tasks. The meta-knowledge is shared among all tasks symmetrically to help them to improve performance. With the growth of meta-knowledge, the performance of each task is expected to be better and better. In order to demonstrate the application of lifelong machine learning, this paper proposed a novel and symmetric lifelong learning approach for sentiment classification as an example to show how it adapts different domains and keeps efficiency meanwhile.

In this paper, we consider the pair-wise semiglobal sequence alignment problem with gaps, which is motivated by the re-sequencing problem that requires to assemble short reads sequences into a genome sequence by referring to a reference sequence. The problem has been studied before for single gap and bounded number of gaps. For single gap, there is a GPU-based algorithm proposed (Barton et al., 2015). In our work, we propose a GPU-based algorithm for the bounded number of gaps case, called GPUGapsMis. We implement the algorithm and compare the performance with the CPU-based algorithm, called CPUGapsMis. The algorithm has two distinct stages: the alignment phase, and the backtrack phase. We investigate several different approaches, in order to determine the most favorable for this problem, by means of a Hybrid model or a wholly-GPU based model, as well as the alignment of single text sequences or multiple text sequences on the GPU at a time. We show that the alignment phase of the algorithm is a good candidate for parallelization, with peak speedup of 11 times. We show that although the backtracking phase is sequential, it is more beneficial to perform it on the GPU, as opposed to returning to the CPU and performing there. When performing both phases on the GPU, GPUGapsMis achieves a peak speedup of 10.4 times against CPUGapsMis. Our data parallel GPU algorithm achieves results which are an improvement on those of an existing GPU data parallel implementation (Ojiaku, 2014).