Long-term temporal information yields crucial cues for video action understanding. Previous researches always rely on sequential models such as recurrent networks, memory units, segmental models, ...self-attention mechanism to integrate the local temporal features for long-term temporal modeling. Recurrent or memory networks record temporal patterns (or relations) by memory units, which are proved to be difficult to capture long-term information in machine translation. Self-attention mechanisms directly aggregate all local information with attention weights which is more straightforward and efficient than the former. However, the attention weights from self-attention ignore the relations between local information and global information which may lead to unreliable attention. To this end, we propose a new attention network architecture, termed as Cascade multi-head ATtention Network (CATNet), which constructs video representations with two-level attentions, namely multi-head local self-attentions and relation based global attentions. Starting from the segment features generated by backbone networks, CATNet first learns multiple attention weights for each segment to capture the importance of local features in a self-attention manner. With the local attention weights, CATNet integrates local features into several global representations, and then learns the second level attention for the global information by a relation manner. Extensive experiments on Kinetics, HMDB51, and UCF101 show that our CATNet boosts the baseline network with a large margin. With only RGB information, we respectively achieve 75.8%, 75.2%, and 96.0% on these three datasets, which are comparable or superior to the state of the arts.
•Propose a Cascade multi-head attention Network to construct video representations.•Provide visual analysis for multi-head attention weights.•Achieve performance comparable or superior to the state of the arts.
The fate and dispersal of oil in the ocean is dependent upon ocean dynamics, as well as transformations resulting from the interaction with the microbial community and suspended particles. These ...interaction processes are parameterized in many models limiting their ability to accurately simulate the fate and dispersal of oil for subsurface oil spill events. This paper presents a coupled ocean-oil-biology-sediment modeling system developed by the Consortium for Simulation of Oil-Microbial Interactions in the Ocean (CSOMIO) project. A key objective of the CSOMIO project was to develop and evaluate a modeling framework for simulating oil in the marine environment, including its interaction with microbial food webs and sediments. The modeling system developed is based on the Coupled Ocean-Atmosphere-Wave-Sediment Transport model (COAWST). Central to CSOMIO’s coupled modeling system is an oil plume model coupled to the hydrodynamic model (Regional Ocean Modeling System, ROMS). The oil plume model is based on a Lagrangian approach that describes the oil plume dynamics including advection and diffusion of individual Lagrangian elements, each representing a cluster of oil droplets. The chemical composition of oil is described in terms of three classes of compounds: saturates, aromatics, and heavy oil (resins and asphaltenes). The oil plume model simulates the rise of oil droplets based on ambient ocean flow and density fields, as well as the density and size of the oil droplets. The oil model also includes surface evaporation and surface wind drift. A novel component of the CSOMIO model is two-way Lagrangian-Eulerian mapping of the oil characteristics. This mapping is necessary for implementing interactions between the ocean-oil module and the Eulerian sediment and biogeochemical modules. The sediment module is a modification of the Community Sediment Transport Modeling System. The module simulates formation of oil-particle aggregates in the water column. The biogeochemical module simulates microbial communities adapted to the local environment and to elevated concentrations of oil components in the water column. The sediment and biogeochemical modules both reduce water column oil components. This paper provides an overview of the CSOMIO coupled modeling system components and demonstrates the capabilities of the modeling system in the test experiments.
Phosphate ions significantly enhance the water resistance of magnesium oxychloride cement (MOC). In this study, MOC was prepared with a MgO/MgCl2 molar ratio of 7 and a H2O/MgCl2 molar ratio of 12, ...while varying the dosage of sodium dihydrogen phosphate. The aim was to observe the microstructural changes in 5Mg(OH)2·MgCl2·8H2O (Phase 5) crystals and the macroscopic mechanical properties before and after immersion. The research found that phosphate ions increase the content of air-cured Phase 5 and reduce the formation of Mg(OH)2 by lowering the required Mg2+ concentration for Phase 5 production. However, excessive phosphate ions can damage the crystal structure of Phase 5, leading to a decrease in MOC strength. An appropriate amount of phosphate ions increases the output of Phase 5, resulting in a highly dense microstructure, which significantly improves the water resistance of MOC. Additionally, this enhances the pore structure of MOC, making the overall structure more compact and resistant to water penetration. This research offers valuable theoretical insights into Phase 5 crystal variations during air curing and water-immersion damage under low Mg2+ concentrations.
•MOC with a relatively low molar ratio of MgO/MgCl2 at 7.•Additional 1–2 % Phase 5 crystal formation.•Excessive sodium dihydrogen phosphate depletes Phase 5 quantity.•Dosage of sodium dihydrogen phosphate should be kept below 4 %.
The use of recycled aggregates (RCAs) could further enhance the sustainability of reactive magnesia cement (RMC) formulations, but the performances of RMC-RCA are still unknown to researchers. ...Therefore, in this study, the strength development of CO2-cured RMC-RCA was explored for the first time, and the strength variation among groups was further interpreted by microstructural analyses. The initial results show that compared with samples with natural aggregates (NAs), a 28 d strength boost (i.e., 37 vs. 15 MPa) in RMC-RCA can be observed. The strength improved in RMC-RCA could be related to its refined microstructures, ITZs and larger carbonation depths, which resulted from the easiness of CO2 penetration within samples. Then the presence of porous ITZs in RCA contributes to better CO2 diffusion, as described in the proposed revised CO2 diffusion model. Overall, while enabling a pronounced CO2 capture potential of RMC concrete, this preliminary study could pave the path for the upcycling of RCAs in RMC concrete on a large scale.
•RMC with RCAs and NAs (various w/b ratios) in CO2 and air curing is prepared.•Use of RCAs led to better strength improvement due to improved CO2 diffusion and ITZs.•Porous ITZs in RCAs also promoted enhanced compacted structures and carbonation depths.•Pore structrues and carbonation degrees could be both critical for RMC samples' strength.•The revised CO2 diffusion model provided enables better RMC-RCA preparation.
Summary
In service‐oriented cloud computing systems (CCSs), the aim of cloud service organizers (CSOs) is to achieve maximum profit by collecting metadata with low cost from big data reporters (BDRs) ...and to provide advanced services to customers at a high price. In these systems, BDRs receive payoffs by reporting metadata to CSOs and exchanging metadata with other BDRs, and customers expect to get high‐quality services at a low price. However, because the missing of a critical parameters decision model in such service‐oriented CCSs, it is difficult to measure key parameters in CCSs such as price and quality of services in the competitive market. In this paper, we propose a multiple game (MG) model to formulate the critical parameters decision process. In the MG model, there are multiple games: games among BDRs and games among CSOs under the rule of “survival of the fittest,” games between BDRs and CSOs under the rule of “the highest payoff first,” and games between customers and CSOs under the rule of “the lowest price and the highest quality of service (QoS) first.” With the proposed multiple game (MG) model, the optimal key parameters can be obtained and the Pareto‐optimal equilibrium point can be achieved. Extensive simulation results demonstrate the effectiveness and efficiency of the proposed MG model in dynamically deciding key parameters in CCSs.
A major current focus in software quality is how to identify and interpret Self-admitted technical debt(SATD). While many methods have been proposed to identify SATD, these methods are neither ...interpretable nor generic. There remains a need for an efficient method that can interpret SATD. In this paper, we propose a two-stage approach to identify and interpret SATD using interpretable methods. In the first stage, the decision tree model is combined into an integrated model to identify SATD better. We apply SHAP, LIME, and Anchors models in the second stage to interpret the result. The experiments of 10 projects show that our method not only can effectively detect and explain SATD both in within-project and cross-project experiments, but also has a good explanation for self-generated data outside the dataset.
To mitigate land losses in the Mississippi River Delta, sediment diversions are being employed to enable the flow of river water and sediments into wetlands experiencing degradation. A ...two-dimensional coupled flow-wave Delft3D model was used in this study to explore the hydrodynamics and sediment transport in Fourleague Bay (FLB), Louisiana, USA, which has been considered an analog site for studying the efficiency of sediment diversion projects. In-situ measurements of sediment accretion and hydrodynamic characteristics from 2015 to 2016 were utilized to calibrate and validate the morphodynamic model. The validated model was then applied to quantify sediment transport in FLB and surrounding marshes between May 2015 and May 2016. The results show that more sediment could be deposited to the surrounding marshes with high river discharges and strong winds. Thus, by strategically aligning the timing of pulses of river water from the diversion with the seasonal intensification of atmospheric forcing, it is possible to sustain and promote the growth of the surrounding wetlands. Moreover, we found that multiple sediment transport processes occurred during the entire study period, including the deposition of riverine sediment into the bay floor, direct deposition of riverine sediment in the surrounding marshes, resuspension of bay floor sediment, and redistribution of resuspended sediment to adjacent marshes and the Gulf of Mexico (GoM). The results indicate that the riverine sediment tended to be directly deposited in the marshes when the river discharge was high. During calm weather conditions and normal river discharge, FLB acted as a reservoir, storing sediment from the upper river, and later acted as a sediment source to the nearby wetlands and the GoM during energetic atmospheric conditions. This suggests that using the bay floor as a reservoir can extend the distance over which wetlands can benefit from the sediment diversions, as the supply of sediment to the wetlands becomes a multi-step process. Thus, it is important to retain sediments from river diversions in shallow bays and allow storms to redistribute them to adjacent wetlands.
•A morphodynamic model was developed to understand the fine sediment dispersal process from a river to coastal wetlands.•The model was calibrated and validated using unique measurements of hydrodynamics and sediment accretion rates.•More sediment could be deposited to the surrounding marshes with high river discharges and strong winds.•Sediment deposited on the surrounding marshes was primarily from the river and bay floor.•Riverine sediment tends to be directly deposited onto the marshes during high discharge stages.
To investigate the interactions among geomorphology, hydrodynamics, and sediment dynamics on the inner shelf offshore Louisiana, multiple acoustic and optical sensors were deployed during a 58-day ...intermediate-energy period from May 23 to July 22, 2016. Time series results show that an elongated bathymetric “trough” between Ship Shoal and Isles Dernieres partially confines flow in the E-W (shore-parallel) direction. Warm water with lower salinity was observed in the mid to upper water column with cool water with higher salinity in the lower water column. High sediment concentrations of 1–10 g/L were observed in the bottom boundary layer during intermediate-energy conditions in response to sustained winds of up to 11 m/s, significant waves heights of up to 1.5 m, occasional 8 s period swells, and a spring tidal range of 0.6 m. The dominant current and sediment transport directions were westward during the study period. About 77% of the sediment flux occurred during three 2-day-long periods (only 10% of the observation period), revealing the nonlinear and episodic nature of sediment transport in this study area. Although intermediate-energy conditions are less energetic than hurricanes and storms, they occur more often and contribute greatly to the long-term net sediment transport. Based on preliminary estimates, ~51.0 million tons of sediment passes along the Louisiana inner shelf annually, comparable with the annual sediment exiting the Mississippi Delta and sourced from marsh edge erosion in coastal Louisiana combined. The inner shelf sediment flux is an integral part of the coastal sediment budget and may provide important mineral sediment for wetland accretion if transported onshore during storms.
PDF
