Search

expert finding in community question answering sites aims to match target questions with experts who are most likely to provide satisfactory answers. Network embedding techniques have been highly successful in expert finding. Nevertheless, most network embedding techniques generate only a single feature vector for experts based on historically answered question to match the target question, often ignoring multi-granularity linguistic matching information. As a result, these methods cannot fully capture the similarity between questions and experts. To tackle these challenges, this study proposes a multi-granularity encoded community question answering expert finding model incorporating graph neural networks, i.e., LG-ERMG: 1) LG-ERMG constructs a relationship graph based on the experts and their history of answered questions and utilizes a lightweight graph convolutional network to capture the potential connections among the experts, which can help to enhance the representation of expert expertise; and 2) multi-granularity coding technique is used to learn different granularity of semantic matching information between target questions and experts’ historical answer questions. In this study, experiments on two real community question answering datasets are carried out to demonstrate the effectiveness of this approach.

The deep integration of Artificial Intelligence (AI) is gradually becoming a key force in innovating the teaching of English as a Foreign Language (EFL). This study aims to assess the practical effects of AI technology in providing customized instructional support and learning pathways in EFL instruction. The study reveals the benefits of AI in the instruction of English vocabulary, utilizing the Apriori algorithm from association rule mining and empirical analysis from survey data of 110 second-year university students across four different majors using AI-powered language learning platforms and AI-powered mobile language learning applications (such as UNIPUS AIGC platform and iTEST, intelligent assessment mobile application). It also deduces related teaching strategies and learning models. The results indicate that the use of AI-powered language learning platforms positively impacts English vocabulary learning outcomes in EFL instruction, and the combined use of AI-powered mobile language learning applications for self-testing and in-class tests effectively enhances vocabulary learning efficiency. The findings and conclusions of this study provide valuable insights for EFL educational practice and demonstrate the potential of AI in boosting the effectiveness of language learning, offering empirical support and guidance for future educational decision-making.

With the continuous development of the communication industry, there is a shift in real-time services from 4G networks to Delay Tolerable (DT) services in the context of 5G/B5G networks. Additionally, energy consumption control poses significant challenges in the current communication industry. Therefore, we study algorithms and schemes to improve the Energy Efficiency (EE) of DT services in the context of Non-Orthogonal Multiple Access (NOMA) downlink two-user communication system.First, we transformed the EE enhancement problem into a convex optimization problem based on transmission power by derivation. Secondly, we propose to use Approximate Statistical Dynamic Programming (ASDP) algorithm, Deep Deterministic Policy Gradient (DDPG), and Proximal Policy Optimization (PPO) to solve the problem that convex optimization cannot be decided in real time. Finally, we perform an interpretability analysis on whether the decision schemes of the agents trained by the DDPG algorithm and the PPO algorithm are reasonable. The simulation results show that the decisions made by the agent trained by the DDPG algorithm perform better compared to the ASDP algorithm and the PPO algorithm.

Classical two-dimensional chaotic systems are not safe enough due to few control parameters and limited chaotic range. To cope with this problem, a new wide-range 2D-Logistic-Sine hyperchaotic map (2D-LSHM) is proposed in this paper. By analyzing the bifurcation map and Lyapunov exponent of 2D-LSHM, the results prove that the map has good ergodicity and unpredictability. In addition, this paper proposes a 2D-LSHM-based image encryption scheme, LSHM-IES, to solve the problem that the dislocation diffusion algorithm based on specific rules is vulnerable to attacks. The scheme uses a $3\times 3$ convolutional kernel to replace the pixel values in the dislocation process. An improved Zigzag transform is developed in the diffusion phase to make the algorithm more secure and the key space larger. Experimental results from a variety of performance tests on different images indicate that the LSHM-IES encryption scheme possesses favorable encryption performance, low time cost, and high robustness against data missing attacks and noise effects.

Recommender systems are an effective way to address the information overload problem. Due to the rapid development of deep learning, recommender systems based on deep learning can efficiently process massive training samples and integrate a variety of additional information, which can alleviate the inherent data sparsity and cold start problems of recommender systems. However, deep learning recommender systems need to collect massive amounts of user data for training models, which poses challenges for data security and user privacy protection. By using users’ local, not centrally collected, data, federated learning can collaboratively train deep learning recommendation models and improve the accuracy of recommendations without compromising user privacy. In this paper, we propose FedDeepFM, a deep recommender system model based on federated learning. First, FedDeepFM follows the standard paradigm of federated learning by uploading each client’s model parameters instead of the original data for model updates, according to the data processing inequality, which ensures that the server cannot collect users’ private information directly but at the same time is able to use the private data held by each client to jointly train a global model with generalization capabilities. Second, to prevent the server from learning users’ private information through indirect inference methods, a pseudo-interaction filling method is used to expand the client data to mask the real user data. The experimental results show that FedDeepFM, the deep recommender system model based on federated learning proposed in this paper, can provide high-quality recommendations while enhancing users’ privacy.

Given a set of facilities and a set of clients, a reverse $k$ nearest neighbors ( $\text{R}k$ NN) query returns every client for which the query facility is one of the $k$ -closest facilities. $\text{R}k$ NN query has been studied thoroughly for its importance in various fields such as facility location. In this paper, we propose a brand new variant of $\text{R}k$ NN query, namely, maximal group reverse $k$ nearest neighbors (MaxGroupR $k$ NN) query. Given a set of clients, a set of candidate facilities and parameters $k$ and $m$ , MaxGroupR $k$ NN query returns a set of $m$ facilities out of the candidates, such that the total number of $\text{R}k$ NNs of the result set is maximized. The MaxGroupR $k$ NN query is important for multi-facility location problem, which aims to maximize the total potential clients of a group of facilities providing the same service such as chain stores, charging stations and logistic centers. A straightforward solution is to enumerate all possible combinations which is obviously time consuming. In order to address this problem, we present an efficient solution namely MGR, which is based on a well-designed pruning technique. The proposed pruning technique is able to filter out the candidates that cannot contribute to the final result and reduce the computation cost dramatically. Moreover, we propose a well-designed optimization technique that can further reduce the computation cost. A detailed theoretical analysis of our methods is provided and the experimental results also confirm that our proposed methods have high efficiency and scalability.

Ridesharing services aim at reducing the users’ travel cost and optimizing the drivers’ routes to satisfy passengers’ expected maximum matching times in practice request dispatching. Existing works can be roughly classified into two types, i.e., online-based and batch-based methods, in which the former mainly focus on responding quickly to the requests and the latter focuses on enumerating request combinations meticulously to improve the service quality. However, online-based methods perform poorly in terms of service quality due to the neglect of the sharing relationship between requests, while batch-based methods fail on efficiency. None of these works can smoothly balance the service quality and matching time cost since the matching window is not sufficiently explored or even neglected. To cope with this problem, we propose a novel framework $\mathcal {E}$ -Ride, which comprehensively leverages the matching time window based on the event model. Specifically, an adaptive windowed matching algorithm is proposed to adaptively consider personalized matching time and provide a matching solution with higher service rates at lower latencies. Besides, we maintain the request groups through a mixed graph and further integrate the subsequent arrival requests to optimize the matching results, which can scale to or satisfy online use demands. The extensive experimental results demonstrate the efficiency and effectiveness of our proposed method.

Rapid advances of sensing and cloud technologies transform the manufacturing system into a data-rich environment and make production scheduling increasingly complex. Traditional offline scheduling methods are limited in the ability to handle low-volume-high-mix workorders with diverse design specifications. Simulation-based methods show the promise for distributed scheduling of manufacturing jobs but are mostly implemented with historical data and empirical rules in a static manner. Recently, artificial intelligence (AI) algorithms fuel increasing interests to solve dynamic scheduling problems in the manufacturing setting. However, it's difficult to utilize high-dimensional data for production scheduling while considering multiple practical objectives for smart manufacturing (e.g., minimize the makespan, reduce production costs, balance workloads). Therefore, this paper presents a new AI scheduler with composite reward functions for data-driven dynamic scheduling of manufacturing jobs under uncertainty in a smart factory. Internet-enabled sensor networks are deployed in the smart factory to track real-time statuses of workorders, machines, and material handling systems. A novel manufacturing value network is developed to take high-dimensional data as the input and then learn the state-action values for real-time decision making. Based on reinforcement learning (RL), composite rewards help the AI scheduler learn efficiently to achieve multiple objectives for production scheduling in real time. The proposed methodology is evaluated and validated with experimental studies in a smart manufacturing setting. Experimental results show that the new AI scheduler not only improves the multi-objective performance metrics in the production scheduling problem but also effectively copes with unexpected events (e.g., urgent workorders, machine failures) in manufacturing systems.

It is well known that the stochastic optimization problem can be regarded as one of the most hard problems since, in most of the cases, the values of $f$ and its gradient are often not easily to be solved, or the $F(\cdot, \xi)$ is normally not given clearly and (or) the distribution function $P$ is equivocal. Then an effective optimization algorithm is successfully designed and used to solve this problem that is an interesting work. This paper designs stochastic bigger subspace algorithms for solving nonconvex stochastic optimization problems. A general framework for such algorithm is presented for convergence analysis, where the so-called the sufficient descent property, the trust region feature, and the global convergence of the stationary points are proved under the suitable conditions. In the worst-case, we will turn out that the complexity is competitive under a given accuracy parameter. We will proved that the $SFO$ -calls complexity of the presented algorithm with diminishing steplength is $O\left({\epsilon ^{-{\frac {1}{1-\beta }}}}\right)$ and the $SFO$ -calls complexity of the given algorithm with random constant steplength is $O(\epsilon ^{-2})$ respectively, where $\beta \in (0.5,1)$ and $\epsilon $ is accuracy and the needed conditions are weaker than the quasi-Newton methods and the normal conjugate gradient algorithms. The detail algorithm framework with variance reduction is also proposed for experiments and the nonconvex binary classification problem is done to demonstrate the performance of the given algorithm.

The multi-objective evolutionary algorithm based on decomposition (MOEA/D) uses a fixed neighborhood size and allocates the same algorithm resources for all sub-problems. This approach makes it harder to effectively optimize the sub-problems in different periods of time, slows the convergence of the algorithm and reduces the quality of the decomposition. This paper proposes an adaptive neighborhood adjustment strategy designed to solve this problem. The neighborhood size of each generation of different subproblems can be adjusted adaptively, and limited algorithm resources can be allocated more efficiently to balance the convergence and diversity of the algorithm. In the algorithm performance comparison experiment, this paper compares the proposed algorithm with the MOEA/D, MOEA/D-GR, MOEA/D-DU and MOEA/D-DN in ZDT and DTLZ series test problems. The experimental results show that the proposed algorithm can efficiently allocate limited algorithm resources, improve algorithm convergence, and achieve better overall performance of the decomposition set.

Preference-inspired co-evolutionary algorithms (PICEAs) consider the target vectors as the preferences, and then use the domination relationship between the candidate solutions and target vectors to increase their selection pressure. However, the size of dominating objective space varies with the different positions of candidate solutions and it leads to the imbalance of the evolutionary ability of whole population. To solve this problem, this paper proposes a preference-inspired coevolutionary algorithm based on a differentiated allocation strategy (PICEAg-DS). First, it sets up an external archive to save the nondominated solutions and then extracts the convergence and diversity information from it. Second, it divides the objective space into several subspaces and designs a space distance operator to evaluate their optimization difficulty. Finally, it dynamically assigns the target vectors and guides more computational resource to the difficult to optimize subspaces, and thus drives the whole population evolution. To prove the advantages of differentiated resource allocation strategy, the PICEAg-DS is compared with two classic coevolutionary algorithms (PICEAg, CMOPSO) and two classic MOEAs based on resource allocation strategy (EAG-MOEAD, MOEAD-DRA). The experimental results show that PICEAg-DS performs better than the other algorithms on many WFG test problems. To further analysis the effectiveness of PICEAg-DS, compare it with two MOEAs based on domination relationship (NSGAII, SPEA2) and two MOEAs based on decomposition (RVEA, MOEA/D-M2M) on MOP and UF test suite. The experimental results show the PICEAg-DS has a better convergence than the other comparison algorithms, especially on 3-objective MOP6-7 and UF8-9.

Photon-counting analysis plays a key role in many areas, such as biology, chemistry, and medicine. In this paper, we present an integrated time-correlated single photon counting (TCSPC) lifetime analytical system with a complete signal path-from fluorophore excitation, emission detection, to lifetime extraction. The time-to-digital module of the proposed TCSPC system achieves a root-mean-square differential non-linearity of 4% of the least significant bit and a full width at half maximum temporal resolution from 121 to 145 ps within the 500-ns full-scale range. To evaluate the lifetime extraction and detection limit of the proposed TCSPC system, a wide variety of samples, such as fluorescein in water, coumarin 6 in dimethyl sulfoxide, and rhodamine 6G in water, each prepared in 14 concentrations from 0.5 nM (nanomolar, 10-9 mol/L) to 25 μM (micromolar, 10-6 mol/L), are tested. With the optimized hardware and firmware design, the proposed TCSPC system can accurately extract the fluorescence lifetime of fluorescein, coumarin 6, and rhodamine 6G down to the concentration of 1, 1, and 2.5 nM, respectively, significantly outperforming similar fluorescence lifetime analysis systems.

Oscillating grinding is a high-efficiency machining method for crankshaft pin journal, and motion model is one of the important foundations of oscillating grinding, which heavily affects the grinding quality. This paper proposes an analysis method of oscillating grinding motion model for crankshaft pin journal, which can provide an effective tool for engineers and technicians to quickly evaluate different motion models in practice. First, the method is described by performing characteristics analysis of oscillating grinding, including the velocity characteristics at grinding point and the performance of driving axis. A complete analysis flow is also given. Then, a numerical comparison example is given between two main oscillating grinding motion models, i.e. the constant linear velocity motion model (CLVMM) and the constant angular velocity motion model (CAVMM), and the results show that the CAVMM is more practical than the CLVMM by using the proposed method. Finally, a crankshaft pin journal grinding experiment is carried out on a prototype, and the result also indicates that the CAVMM is much better in practice, which validates the effectiveness of the proposed analysis method of oscillating grinding motion model.

In surveillance applications, the extent states and measurements of extended targets received by sensors are time-varying. In this paper, we propose a joint tracking and classification (JTC) method for single extended target under the presence of clutter and detection uncertainty. The extent state is modeled as elliptic shape via random matrix model (RMM), and is used as the feature for target classification. To adapt to the time-varying conditions of an extended target, the RMM proposed by Lan et al. is used. Besides, the RMM is integrated into Bernoulli filter to detect an extended target with clutter and detection uncertainty. The resulting method is called joint tracking and classification Gaussian inverse Wishart Bernoulli (JTC-GIW-Ber) filter, and the closed expressions for JTC-GIW-Ber filter recursions are derived under the necessary assumptions and approximations. Comprehensive simulations are carried out to test the performance, and the results demonstrate that the proposed JTC-GIW-Ber filter not only outperforms the JTC-GIW probability hypothesis density (JTC-GIW-PHD) filter and the GIW-Ber filter in extent state estimation, but also outperforms the JTC-GIW-PHD filter in target classification.

In marine engineering design, the threshold selection is a basic and very important part for the analysis of measured data and subsequent acquisition of sample data for probability analysis. In this paper, threshold selection methods are deeply studied from the two perspectives of time domain and frequency domain, and the obtained calculation results with different thresholds are provided. First, the theory of mean residual life plot is used to quantify the threshold value of the measured tide levels of Hangzhou Bay. For the 100-year return periods design values calculated in the Gumbel distribution, the results under the threshold u = 3.9 m shows an increasing of 1.70% over those under the threshold u = 3.75 m; and the results under the threshold u = 3.75 m shows an increase of 0.97% over those under the threshold u = 3.5 m. The quantitative selection method of a threshold is simple and practical, and eliminates the tedious process of multiple trials based on empirical threshold determination. Besides, for the first time, the coherence spectrum threshold is used to explore the significant correlations of different tide level sequences in the frequency domain. Based on the threshold selection method used in the WOSA method-based coherence spectrum significance test, the coherence spectrum thresholds of different tide level in Hangzhou Bay are provided to effectively avoid the false identification of false peaks. It is suggested by the calculation results that, at the overlapping ratio of 50%, the segment number of three and the segment length of 16, and the spectral analysis shows that the extreme tide level of Hangzhou Bay has a vibration period of 21.3 years.

Time-resolved fluorescence is a widely adopted technique for DNA detection due to its high sensitivity and selectivity. However, due to stringent requirements on optics and electronics, instrumentation with time-resolved capability is bulky and expensive, prohibiting their use in portable and point-of-care applications. In this work, a fully-integrated DNA biosensor based on liquid-core waveguide (LCW) optics for fluorescence lifetime analysis is presented. The DNA biosensor encompasses all-custom bioassay, optics and electronics into a microsystem, delivering a near sample-to-answer level of integration. Lifetime, rather than intensity, is exploited as the analytical signal from the V-carbazole probe for the first time. Excitation propagation within the LCW is investigated both analytically and in simulations to achieve high excitation rejection and low temporal dispersion, enabling the proposed LCW-based system with much smaller instrumentation size to deliver comparable lifetime measurement accuracy to traditional systems. Detection of DNA down to 15 base pairs at a low detection limit of 1.38 nM demonstrates the high applicability of the proposed biosensor for compact, application-specific, and low-cost diagnostics devices.

In this paper, a fractional calculus operator for image denoising is constructed based on the characteristic of local entropy and the gradient feature, and an adaptive fractional calculus image denoising algorithm is proposed. First, the effects on the entropy and gradient by noise are analyzed, respectively. Second, the noise points are regarded as small probability events in an image, and the noise points, edges, texture regions, and smooth regions are divided combining with the local structure. Finally, for improving the image denoising effect, we consider employing different fractional orders to deal with different pixels and a piecewise function is constructed to make the differential order to be adaptive. The function is with respect to the local entropy and gradient on the pixel. The experimental results show that the peak signal-to-noise ratio and the entropy (ENTROPY) of the proposed adaptive fractional calculus image denoising algorithm are higher than that of the other algorithms compared in this paper. The proposed algorithm can not only preserve image edges and texture information while removing the noise, but also obtain a good visual effect.