The level of service (LOS) serves as a crucial metric in assessing the efficiency of intelligent transportation systems (ITSs). Precise forecasting of LOS is vital to improving the service quality provided by ITSs. However, within a transportation network, the LOS at each node typically exhibits spatial and temporal correlations due to the complex dynamics of traffic patterns. This poses a significant difficulty in accurately predicting traffic LOS. In this article, we propose a deep learning model, namely spatiotemporal cross-attention neural network (STCANN), to predict the LOS of a transportation network. On the one hand, the spatiotemporal dependencies between network nodes are captured through a set of stacked spatial and temporal cross-attention modules. On the other hand, the nonlinear interactions between traffic flow and speed are also resolved using a deliberate cross-attention mechanism. To evaluate the performance of the proposed model, we conduct experiments on two public datasets available from California highways. The results demonstrate that the proposed STCANN model outperforms the state-of-the-art methods in terms of prediction accuracy. The findings show that the incorporation of the cross-attention mechanism into the spatiotemporal neural network architecture shows great promise in improving the accuracy and reliability of LOS prediction.

This study addresses the issue of uneven energy consumption in electric vehicle (EV) platoons, arising from the static sequencing of vehicles within the platoon. Such an imbalance can negatively impact the efficiency of individual vehicles and the driving performance of the entire platoon. Our approach proposes dynamically altering the formation of the platoon during transit to balance energy use. The core challenge is to identify the most efficient vehicle sequence at predetermined re-sequencing points during the journey. To address this, we introduce three innovative methods based on deep reinforcement learning, chosen for their ability to handle complex, dynamic optimization problems. Our experimental studies, conducted on actual transportation networks, demonstrate these methods significantly enhance energy management and distribution efficiency in EV platoons, highlighting their potential for practical applications in intelligent transportation systems.

Sea surface temperature (SST) is a fundamental parameter in the field of oceanography as it significantly influences various physical, chemical, and biological processes within the marine environment. In this study, we propose an attention-based context fusion network (ACFN) model for short-term prediction of SST based on the operational SST and sea ice analysis (OSTIA) data. The ACFN model integrates an attention-based context fusion block into the convolutional long short-term memory (ConvLSTM) model. Specifically, the attention mechanism generates attention maps sequentially across both the channel and spatial dimensions, allowing for the detailed exploration of intricate spatiotemporal correlations between the previous context state and the current input state in ConvLSTM. To assess the performance of the ACFN model, we apply it to predict SST in the Bohai Sea with lead times ranging from one to ten days. The results demonstrate that our proposed model outperforms several state-of-the-art models, i.e., ConvLSTM, predictive RNN (PredRNN), SimVP, and motion details RNN (MoDeRNN), in terms of mean absolute error (MAE) and coefficient of determination ($R^{2}$). In particular, our analysis reveals that the prediction errors are relatively higher in the coastal areas compared to those in the central Bohai Sea.

Accurate prediction of sea surface temperature (SST) is crucial in the field of oceanography, as it has a significant impact on various physical, chemical, and biological processes in the marine environment. In this study, we propose a physics-guided attention-based neural network (PANN) to address the spatiotemporal SST prediction problem. The PANN model incorporates data-driven spatiotemporal convolution operations and the underlying physical dynamics of SSTs using a cross-attention mechanism. First, we construct a spatiotemporal convolution module (SCM) using convolutional long short-term memory (ConvLSTM) to capture the spatial and temporal correlations present in the time series of the SST data. We then introduce a physical constraint module (PCM) to mimic the transport dynamics in fluids based on data assimilation techniques used to solve partial differential equations (PDEs). Consequently, we employ an attention fusion module (AFM) to effectively combine the data-driven and PDE-constrained predictions obtained from the SCM and PCM, aiming at enhancing the accuracy of the predictions. To evaluate the performance of the proposed model, we conduct short-term SST forecasts in the East China Sea (ECS) with forecast lead times ranging from one to ten days, by comparing it with several state-of-the-art models, including ConvLSTM, PredRNN, temporal convolutional transformer network (TCTN), convolutional gated recurrent unit (ConvGRU), and SwinLSTM. The experimental results demonstrate that our proposed model outperforms these models in terms of multiple evaluation metrics for short-term predictions.

Online electroencephalograph (EEG) classification is a core service of recently booming brain e-health, but its performance often becomes unstable because (1) conventional end-to-end models (e.g., deep neural network, DNN) largely remain static, while brain states of diseases are highly dynamic and exhibits significant individuality; and (2) EEG analytics are too complicated and have to be sustained by advanced computing services. This study adopts an automatic machine learning method to construct a dual-CNN (convolutional neural network) of high performance in terms of both accuracy and efficiency. The model can optimize its hyperparameters continuously on its own initiative. Experimental results in the evaluation of depression using real EEG datasets indicate that (1) the proposed method executes 3.5 times faster compared with a conventional counterpart; (2) the dual-CNN gains a significant performance improvement (versus CapsuleNet and Resnet-16) in identifying Major Depression Disorder (MDD) with accuracy, sensitivity, and specificity up to 98.81, 98.36, and 99.31 percent respectively; and those for treatment outcome are 99.52, 99.63, and 99.37 percent respectively, and (3) classification can be completed several hundred times faster than EEG being collected upon a COTS computer.

Understanding the dynamic properties of epidemic spreading on complex social networks is essential to make effective and efficient public health policies for epidemic prevention and control. In recent years, the concept of network embedding has attracted lots of attention to deal with various network analytic tasks, the purpose of which is to encode relationships or information of networked elements into a low-dimensional vector space. However, most existing embedding methods have focused mainly on preserving static network information, such as structural proximity, node/edge attributes, and labels. On the contrary, in this paper, we focus on the embedding problem of preserving dynamic characteristics of epidemic spreading on social networks. We propose a novel embedding method, namely EpiRep, to learn node representations of a network by maximizing the likelihood of preserving groups of infected nodes due to the epidemics starting from every single node on the network. Specifically, the Susceptible-Infectious model is adopted to simulate the epidemic dynamics on networks, and the Continuous Bag-of-Words model with negative sampling is used to obtain node representations. Experimental results show that the EpiRep method outperforms two benchmark random-walk based embedding methods in terms of node clustering and classification on several synthetic and real-world networks. The proposed method and findings in this paper may offer new insight for source identification and infection prevention in the face of epidemic spreading on social networks.

BACKGROUND: Novel sequence motifs detection is becoming increasingly essential in computational biology. However, the high computational cost greatly constrains the efficiency of most motif discovery algorithms. RESULTS: In this paper, we accelerate MEME algorithm targeted on Intel Many Integrated Core (MIC) Architecture and present a parallel implementation of MEME called MIC-MEME base on hybrid CPU/MIC computing framework. Our method focuses on parallelizing the starting point searching method and improving iteration updating strategy of the algorithm. MIC-MEME has achieved significant speedups of 26.6 for ZOOPS model and 30.2 for OOPS model on average for the overall runtime when benchmarked on the experimental platform with two Xeon Phi 3120 coprocessors. CONCLUSIONS: Furthermore, MIC-MEME has been compared with state-of-arts methods and it shows good scalability with respect to dataset size and the number of MICs. Source code: https://github.com/hkwkevin28/MIC-MEME .

BACKGROUND: Frequent subgraphs mining is a significant problem in many practical domains. The solution of this kind of problem can particularly used in some large-scale drug molecular or biological libraries to help us find drugs or core biological structures rapidly and predict toxicity of some unknown compounds. The main challenge is its efficiency, as (i) it is computationally intensive to test for graph isomorphisms, and (ii) the graph collection to be mined and mining results can be very large. Existing solutions often require days to derive mining results from biological networks even with relative low support threshold. Also, the whole mining results always cannot be stored in single node memory. RESULTS: In this paper, we implement a parallel acceleration tool for classical frequent subgraph mining algorithm called cmFSM. The core idea is to employ parallel techniques to parallelize extension tasks, so as to reduce computation time. On the other hand, we employ multi-node strategy to solve the problem of memory constraints. The parallel optimization of cmFSM is carried out on three different levels, including the fine-grained OpenMP parallelization on single node, multi-node multi-process parallel acceleration and CPU-MIC collaborated parallel optimization. CONCLUSIONS: Evaluation results show that cmFSM clearly outperforms the existing state-of-the-art miners even if we only hold a few parallel computing resources. It means that cmFSM provides a practical solution to frequent subgraph mining problem with huge number of mining results. Specifically, our solution is up to one order of magnitude faster than the best CPU-based approach on single node and presents a promising scalability of massive mining tasks in multi-node scenario. More source code are available at:Source Code: https://github.com/ysycloud/cmFSM .

Co-evolution exists ubiquitously in biological systems. At the molecular level, interacting proteins, such as ligands and their receptors and components in protein complexes, co-evolve to maintain their structural and functional interactions. Many proteins contain multiple functional domains interacting with different partners, making co-evolution of interacting domains occur more prominently. Multiple methods have been developed to predict interacting proteins or domains within proteins by detecting their co-variation. This strategy neglects the fact that interacting domains can be highly co-conserved due to their functional interactions. Here we report a novel algorithm to detect signals of both co-positive selection (co-variation) and co-purifying selection (co-conservation). Preliminary results show that our algorithm performs well and outperforms the popular co-variation analysis program CAPS. Our algorithm can be widely used to predict interacting domains in protein and nucleotide sequences and to analyze protein-ncRNA complexes.

Active surveillance is a desirable way to prevent the spread of infectious diseases in that it aims to timely discover individual incidences through an active searching for patients. However, in practice active surveillance is difficult to implement especially when monitoring space is large but available resources are limited. Therefore, it is extremely important for public health authorities to know how to distribute their very sparse resources to high-priority regions so as to maximize the outcomes of active surveillance. In this paper, we raise the problem of active surveillance planning and provide an effective method to address it via modeling and mining spatiotemporal patterns of infection risks from heterogeneous data sources. Taking malaria as an example, we perform an empirical study on real-world data to validate our method and provide our new findings.

Most real-world distribution systems can be modeled as distribution networks, where a commodity can flow from source nodes to sink nodes through junction nodes. One of the fundamental characteristics of distribution networks is the functional robustness, which reflects the ability of maintaining its function in the face of internal or external disruptions. In view of the fact that most distribution networks do not have any centralized control mechanisms, we consider the problem of how to improve the functional robustness in a decentralized way. To achieve this goal, we study two important problems: 1) how to formally measure the functional robustness, and 2) how to improve the functional robustness of a network based on the local interaction of its nodes. First, we derive a utility function in terms of network entropy to characterize the functional robustness of a distribution network. Second, we propose a decentralized network pricing mechanism, where each node need only communicate with its distribution neighbors by sending a "price" signal to its upstream neighbors and receiving "price" signals from its downstream neighbors. By doing so, each node can determine its outflows by maximizing its own payoff function. Our mathematical analysis shows that the decentralized pricing mechanism can produce results equivalent to those of an ideal centralized maximization with complete information. Finally, to demonstrate the properties of our mechanism, we carry out a case study on the U.S. natural gas distribution network. The results validate the convergence and effectiveness of our mechanism when comparing it with an existing algorithm.

Bolstering resource coallocation is essential for realizing the Grid vision, because computationally intensive applications often require multiple computing resources from different administrative domains. Given that resource providers and consumers may have different requirements, successfully obtaining commitments through concurrent negotiations with multiple resource providers to simultaneously access several resources is a very challenging task for consumers. The impetus of this paper is that it is one of the earliest works that consider a concurrent negotiation mechanism for Grid resource coallocation. The concurrent negotiation mechanism is designed for 1) managing (de)commitment of contracts through one-to-many negotiations and 2) coordination of multiple concurrent one-to-many negotiations between a consumer and multiple resource providers. The novel contributions of this paper are devising 1) a utility-oriented coordination (UOC) strategy, 2) three classes of commitment management strategies (CMSs) for concurrent negotiation, and 3) the negotiation protocols of consumers and providers. Implementing these ideas in a testbed, three series of experiments were carried out in a variety of settings to compare the following: 1) the CMSs in this paper with the work of others in a single one-to-many negotiation environment for one resource where decommitment is allowed for both provider and consumer agents; 2) the performance of the three classes of CMSs in different resource market types; and 3) the UOC strategy with the work of others [e.g., the patient coordination strategy (PCS )] for coordinating multiple concurrent negotiations. Empirical results show the following: 1) the UOC strategy achieved higher utility, faster negotiation speed, and higher success rates than PCS for different resource market types; and 2) the CMS in this paper achieved higher final utility than the CMS in other works. Additionally, the properties of the three classes of CMSs in different kinds of resource markets are also verified.

Automated negotiation provides a means for resolving differences among interacting agents. For negotiation with complete information, this paper provides mathematical proofs to show that an agent's optimal strategy can be computed using its opponent's reserve price (RP) and deadline. The impetus of this work is using the synergy of Bayesian learning (BL) and genetic algorithm (GA) to determine an agent's optimal strategy in negotiation (N) with incomplete information. BLGAN adopts: 1) BL and a deadline-estimation process for estimating an opponent's RP and deadline and 2) GA for generating a proposal at each negotiation round. Learning the RP and deadline of an opponent enables the GA in BLGAN to reduce the size of its search space (SP) by adaptively focusing its search on a specific region in the space of all possible proposals. SP is dynamically defined as a region around an agent's proposal P at each negotiation round. P is generated using the agent's optimal strategy determined using its estimations of its opponent's RP and deadline. Hence, the GA in BLGAN is more likely to generate proposals that are closer to the proposal generated by the optimal strategy. Using GA to search around a proposal generated by its current strategy, an agent in BLGAN compensates for possible errors in estimating its opponent's RP and deadline. Empirical results show that agents adopting BLGAN reached agreements successfully, and achieved: 1) higher utilities and better combined negotiation outcomes (CNOs) than agents that only adopt GA to generate their proposals, 2) higher utilities than agents that adopt BL to learn only RP, and 3) higher utilities and better CNOs than agents that do not learn their opponents' RPs and deadlines.

Whereas many extant works only adopt utility as the performance measure for evaluating negotiation agents, this work formulates strategies that optimize combined negotiation outcomes in terms of utilities, success rates, and negotiation speed. In some applications (e.g., Grid resource management), negotiation agents should be designed such that they are more likely to acquire resources more rapidly and with more certainty (in addition to optimizing utility0. For negotiations with complete information, mathematical proofs show that the negotiation strategy set in this work optimizes the utilities of agents while guaranteeing that agreements are reached. A novel algorithm BLGAN is devised to guide agents in negotiations with incomplete information. BLGAN adopts 1) a Bayesian learning (BL) approach for estimating the reserve price of an agent’ opponent, and 2) a multi-objective genetic algorithm (GA) for generating a proposal at each negotiation (N) round. In bilateral negotiations with incomplete information, empirical results show that when both agents adopt BLGAN to learn each other’s reserve price, they are both guaranteed to reach agreements, and complete negotiations with much fewer negotiation rounds. When only one agent adopts BLGAN, the agent was highly successful n reaching agreements, achieved average utlities that were much closer to optimal, and used fewer negotiation rounds than the agent that did not adopt BLGAN.