基于Deep Forest算法的对虾急性肝胰腺坏死病(AHPND)预警数学模型构建

为预报池塘养殖凡纳对虾(Penaeus vannamei)急性肝胰腺坏死病(AHPND)的发生,自2020年开始,笔者对凡纳对虾养殖区开展了连续监测工作,包括与疾病发生相关的环境理化因子、微生物因子、虾体自身健康状况等18个候选预警因子指标,通过数据标准化处理后分析病原、宿主与环境之间的相关性,对候选预警因子进行筛选,基于Python语言编程结合Deep Forest、Light GBM、XGBoost算法进行数据建模和预测性能评判,仿真环境为Python2.7,以预警因子指标作为输入样本(即警兆),以对虾是否发病指标作为输出结果(即警情),根据输入样本和输出结果各自建立输入数据矩阵和目标数据矩阵,利用原始数据矩阵对输入样本进行初始化,结合函数方程进行拟合,拟合的源代码能利用已知环境、病原及对虾免疫指标数据对目标警情进行预测。最终建立了基于Deep Forest算法的虾体(肝胰腺内)细菌总数、虾体弧菌(Vibrio)占比、水体细菌总数和盐度的4维向量预警预报模型,准确率达89.00%。本研究将人工智能算法应用到对虾AHPND发生的预测预报,相关研究结果为对虾AHPND疾病预警预报建立了预警数学模型,并为对虾健康养殖和疾病防控提供了技术支撑和有力保障。

基于GRU-DRSN的双通道人体活动识别

人体活动识别(human activity recognizition, HAR)在医疗、军工、智能家居等领域有很大的应用空间。传统机器学习方法特征提取难度较大且精度不高。针对上述问题并结合传感器时序特性,提出了一种融合CBAM(convolutional block attention module)注意力机制的GRU-DRSN双通道并行模型,有效避免了传统串行模型因网络深度加深引起梯度爆炸和消失问题。同时并行结构使得两条支路具有相同的优先级,使用深度残差收缩网络(deep residual shrinkage network, DRSN)提取数据的深层空间特征,同时使用门控循环结构(gated recurrent unit, GRU)学习活动样本在时间序列上的特征,同时进行提取样本不同维度的特征,并通过CBAM模块进行特征的权重分配,最后通过Softmax层进行识别,实现了端对端的人体活动识别。使用公开数据集(wireless sensor data mining, WISDM)进行验证,模型平均精度达到了97.6%,与传统机器学习模型和前人所提神经网络模型相比,有更好的识别效果。

基于视频和BCE-YOLO模型的奶牛采食行为检测

【目的】动物采食行为是一个重要的动物福利评价指标,本研究旨在解决复杂饲养环境下奶牛采食行为识别精度差、特征提取不充分的问题,实现对奶牛采食行为的自动监控。【方法】本文提出了一种基于改进BCEYOLO模型的识别方法,该方法通过添加BiFormer、CoT、EMA 3个增强模块,提高YOLOv8模型特征提取的能力,进一步与优于Staple、SiameseRPN算法的Deep SORT算法结合,实现对奶牛采食时头部轨迹的跟踪。在奶牛采食过程的俯视和正视视频中提取11288张图像,按照6∶1比例分为训练集和测试集,构建采食数据集。【结果】改进的BCE-YOLO模型在前方和上方拍摄的数据集上精确度分别为77.73%、76.32%,召回率分别为82.57%、86.33%,平均精确度均值分别为83.70%、76.81%;相较于YOLOv8模型,整体性能提升6~8个百分点。Deep SORT算法相比于Staple、SiameseRPN算法,综合性能提高1~4个百分点;并且改进的BCE-YOLO模型与Deep SORT目标跟踪算法结合良好,能对奶牛采食行为进行准确跟踪且有效地抑制了奶牛ID(Identity document)的变更。【结论】本文提出的方法能有效解决复杂饲养环境下奶牛采食行为识别精度差、特征提取不充分的问题,为智能畜牧与精确养殖提供重要参考。

Machine learning applications in stroke medicine:advancements,challenges,and future prospectives

Stroke is a leading cause of disability and mortality worldwide,necessitating the development of advanced technologies to improve its diagnosis,treatment,and patient outcomes.In recent years,machine learning techniques have emerged as promising tools in stroke medicine,enabling efficient analysis of large-scale datasets and facilitating personalized and precision medicine approaches.This abstract provides a comprehensive overview of machine learning’s applications,challenges,and future directions in stroke medicine.Recently introduced machine learning algorithms have been extensively employed in all the fields of stroke medicine.Machine learning models have demonstrated remarkable accuracy in imaging analysis,diagnosing stroke subtypes,risk stratifications,guiding medical treatment,and predicting patient prognosis.Despite the tremendous potential of machine learning in stroke medicine,several challenges must be addressed.These include the need for standardized and interoperable data collection,robust model validation and generalization,and the ethical considerations surrounding privacy and bias.In addition,integrating machine learning models into clinical workflows and establishing regulatory frameworks are critical for ensuring their widespread adoption and impact in routine stroke care.Machine learning promises to revolutionize stroke medicine by enabling precise diagnosis,tailored treatment selection,and improved prognostication.Continued research and collaboration among clinicians,researchers,and technologists are essential for overcoming challenges and realizing the full potential of machine learning in stroke care,ultimately leading to enhanced patient outcomes and quality of life.This review aims to summarize all the current implications of machine learning in stroke diagnosis,treatment,and prognostic evaluation.At the same time,another purpose of this paper is to explore all the future perspectives these techniques can provide in combating this disabling disease.

Estimating shear strength of high-level pillars supported with cemented backfilling using the HoekeBrown strength criterion

Deep metal mines are often mined using the high-level pillars with subsequent cementation backfilling(HLSCB)mining method.At the design stage,it is therefore important to have a reasonable method for determining the shear strength of the high-level pillars(i.e.cohesion and internal friction angle)when they are supported by cemented backfilling.In this study,a formula was derived for the upper limit of the confining pressure σ3max on a high-level pillar supported by cemented backfilling in a deep metal mine.A new method of estimating the shear strength of such pillars was then proposed based on the Hoek eBrown failure criterion.Our analysis indicates that the horizontal stress σhh acting on the cemented backfill pillar can be simplified by expressing it as a constant value.A reasonable and effective value for σ3max can then be determined.The value of s3max predicted using the proposed method is generally less than 3 MPa.Within this range,the shear strength of the high-level pillar is accurately calculated using the equivalent MohreCoulomb theory.The proposed method can effectively avoid the calculation of inaccurate shear strength values for the high-level pillars when the original HoekeBrown criterion is used in the presence of large confining pressures,i.e.the situation in which the cohesion value is too large and the friction angle is too small can effectively be avoided.The proposed method is applied to a deep metal mine in China that is being excavated using the HLSCB method.The shear strength parameters of the high-level pillars obtained using the proposed method were input in the numerical simulations.The numerical results show that the recommended level heights and sizes of the high-level pillars and rooms in the mine are rational.

Prediction models of burst strength degradation for casing with considerations of both wear and corrosion

Casing wear and casing corrosion are serious problems affecting casing integrity failure in deep and ultra-deep wells.This paper aims to predict the casing burst strength with considerations of both wear and corrosion.Firstly,the crescent wear shape is simplified into three categories according to common mathematical models.Then,based on the mechano-electrochemical(M-E)interaction,the prediction model of corrosion depth is built with worn depth as the initial condition,and the prediction models of burst strength of the worn casing and corroded casing are obtained.Secondly,the accuracy of different prediction models is validated by numerical simulation,and the main influence factors on casing strength are obtained.At last,the theoretical models are applied to an ultra-deep well in Northwest China,and the dangerous well sections caused by wear and corrosion are predicted,and the corrosion rate threshold to ensure the safety of casing is obtained.The results show that the existence of wear defects results in a stress concentration and enhanced M-E interaction on corrosion depth growth.The accuracy of different mathematical models is different:the slot ring model is most accurate for predicting corrosion depth,and the eccentric model is most accurate for predicting the burst strength of corroded casing.The burst strength of the casing will be overestimated by more than one-third if the M-E interaction is neglected,so the coupling effect of wear and corrosion should be sufficiently considered in casing integrity evaluation.

Superiority of a Convolutional Neural Network Model over Dynamical Models in Predicting Central Pacific ENSO

The application of deep learning is fast developing in climate prediction,in which El Ni?o–Southern Oscillation(ENSO),as the most dominant disaster-causing climate event,is a key target.Previous studies have shown that deep learning methods possess a certain level of superiority in predicting ENSO indices.The present study develops a deep learning model for predicting the spatial pattern of sea surface temperature anomalies(SSTAs)in the equatorial Pacific by training a convolutional neural network(CNN)model with historical simulations from CMIP6 models.Compared with dynamical models,the CNN model has higher skill in predicting the SSTAs in the equatorial western-central Pacific,but not in the eastern Pacific.The CNN model can successfully capture the small-scale precursors in the initial SSTAs for the development of central Pacific ENSO to distinguish the spatial mode up to a lead time of seven months.A fusion model combining the predictions of the CNN model and the dynamical models achieves higher skill than each of them for both central and eastern Pacific ENSO.

Pluggable multitask diffractive neural networks based on cascaded metasurfaces

Optical neural networks have significant advantages in terms of power consumption,parallelism,and high computing speed,which has intrigued extensive attention in both academic and engineering communities.It has been considered as one of the powerful tools in promoting the fields of imaging processing and object recognition.However,the existing optical system architecture cannot be reconstructed to the realization of multi-functional artificial intelligence systems simultaneously.To push the development of this issue,we propose the pluggable diffractive neural networks(P-DNN),a general paradigm resorting to the cascaded metasurfaces,which can be applied to recognize various tasks by switching internal plug-ins.As the proof-of-principle,the recognition functions of six types of handwritten digits and six types of fashions are numerical simulated and experimental demonstrated at near-infrared regimes.Encouragingly,the proposed paradigm not only improves the flexibility of the optical neural networks but paves the new route for achieving high-speed,low-power and versatile artificial intelligence systems.

Evolutionary Decision-Making and Planning for Autonomous Driving Based on Safe and Rational Exploration and Exploitation

Decision-making and motion planning are extremely important in autonomous driving to ensure safe driving in a real-world environment.This study proposes an online evolutionary decision-making and motion planning framework for autonomous driving based on a hybrid data-and model-driven method.First,a data-driven decision-making module based on deep reinforcement learning(DRL)is developed to pursue a rational driving performance as much as possible.Then,model predictive control(MPC)is employed to execute both longitudinal and lateral motion planning tasks.Multiple constraints are defined according to the vehicle’s physical limit to meet the driving task requirements.Finally,two principles of safety and rationality for the self-evolution of autonomous driving are proposed.A motion envelope is established and embedded into a rational exploration and exploitation scheme,which filters out unreasonable experiences by masking unsafe actions so as to collect high-quality training data for the DRL agent.Experiments with a high-fidelity vehicle model and MATLAB/Simulink co-simulation environment are conducted,and the results show that the proposed online-evolution framework is able to generate safer,more rational,and more efficient driving action in a real-world environment.

Point Cloud Classification Using Content-Based Transformer via Clustering in Feature Space

Recently, there have been some attempts of Transformer in 3D point cloud classification. In order to reduce computations, most existing methods focus on local spatial attention,but ignore their content and fail to establish relationships between distant but relevant points. To overcome the limitation of local spatial attention, we propose a point content-based Transformer architecture, called PointConT for short. It exploits the locality of points in the feature space(content-based), which clusters the sampled points with similar features into the same class and computes the self-attention within each class, thus enabling an effective trade-off between capturing long-range dependencies and computational complexity. We further introduce an inception feature aggregator for point cloud classification, which uses parallel structures to aggregate high-frequency and low-frequency information in each branch separately. Extensive experiments show that our PointConT model achieves a remarkable performance on point cloud shape classification. Especially, our method exhibits 90.3% Top-1 accuracy on the hardest setting of ScanObjectN N. Source code of this paper is available at https://github.com/yahuiliu99/PointC onT.

Deep learning-based inpainting of saturation artifacts in optical coherence tomography images

Limited by the dynamic range of the detector,saturation artifacts usually occur in optical coherence tomography(OCT)imaging for high scattering media.The available methods are difficult to remove saturation artifacts and restore texture completely in OCT images.We proposed a deep learning-based inpainting method of saturation artifacts in this paper.The generation mechanism of saturation artifacts was analyzed,and experimental and simulated datasets were built based on the mechanism.Enhanced super-resolution generative adversarial networks were trained by the clear–saturated phantom image pairs.The perfect reconstructed results of experimental zebrafish and thyroid OCT images proved its feasibility,strong generalization,and robustness.

A review of reservoir damage during hydraulic fracturing of deep and ultra-deep reservoirs

Deep and ultra-deep reservoirs have gradually become the primary focus of hydrocarbon exploration as a result of a series of significant discoveries in deep hydrocarbon exploration worldwide.These reservoirs present unique challenges due to their deep burial depth(4500-8882 m),low matrix permeability,complex crustal stress conditions,high temperature and pressure(HTHP,150-200℃,105-155 MPa),coupled with high salinity of formation water.Consequently,the costs associated with their exploitation and development are exceptionally high.In deep and ultra-deep reservoirs,hydraulic fracturing is commonly used to achieve high and stable production.During hydraulic fracturing,a substantial volume of fluid is injected into the reservoir.However,statistical analysis reveals that the flowback rate is typically less than 30%,leaving the majority of the fluid trapped within the reservoir.Therefore,hydraulic fracturing in deep reservoirs not only enhances the reservoir permeability by creating artificial fractures but also damages reservoirs due to the fracturing fluids involved.The challenging“three-high”environment of a deep reservoir,characterized by high temperature,high pressure,and high salinity,exacerbates conventional forms of damage,including water sensitivity,retention of fracturing fluids,rock creep,and proppant breakage.In addition,specific damage mechanisms come into play,such as fracturing fluid decomposition at elevated temperatures and proppant diagenetic reactions at HTHP conditions.Presently,the foremost concern in deep oil and gas development lies in effectively assessing the damage inflicted on these reservoirs by hydraulic fracturing,comprehending the underlying mechanisms,and selecting appropriate solutions.It's noteworthy that the majority of existing studies on reservoir damage primarily focus on conventional reservoirs,with limited attention given to deep reservoirs and a lack of systematic summaries.In light of this,our approach entails initially summarizing the current knowledge pertaining to the types of fracturing fluids employed in deep and ultra-deep reservoirs.Subsequently,we delve into a systematic examination of the damage processes and mechanisms caused by fracturing fluids within the context of hydraulic fracturing in deep reservoirs,taking into account the unique reservoir characteristics of high temperature,high pressure,and high in-situ stress.In addition,we provide an overview of research progress related to high-temperature deep reservoir fracturing fluid and the damage of aqueous fracturing fluids to rock matrix,both artificial and natural fractures,and sand-packed fractures.We conclude by offering a summary of current research advancements and future directions,which hold significant potential for facilitating the efficient development of deep oil and gas reservoirs while effectively mitigating reservoir damage.

Toward a Learnable Climate Model in the Artificial Intelligence Era

Artificial intelligence(AI)models have significantly impacted various areas of the atmospheric sciences,reshaping our approach to climate-related challenges.Amid this AI-driven transformation,the foundational role of physics in climate science has occasionally been overlooked.Our perspective suggests that the future of climate modeling involves a synergistic partnership between AI and physics,rather than an“either/or”scenario.Scrutinizing controversies around current physical inconsistencies in large AI models,we stress the critical need for detailed dynamic diagnostics and physical constraints.Furthermore,we provide illustrative examples to guide future assessments and constraints for AI models.Regarding AI integration with numerical models,we argue that offline AI parameterization schemes may fall short of achieving global optimality,emphasizing the importance of constructing online schemes.Additionally,we highlight the significance of fostering a community culture and propose the OCR(Open,Comparable,Reproducible)principles.Through a better community culture and a deep integration of physics and AI,we contend that developing a learnable climate model,balancing AI and physics,is an achievable goal.

Correcting Climate Model Sea Surface Temperature Simulations with Generative Adversarial Networks:Climatology,Interannual Variability,and Extremes

Climate models are vital for understanding and projecting global climate change and its associated impacts.However,these models suffer from biases that limit their accuracy in historical simulations and the trustworthiness of future projections.Addressing these challenges requires addressing internal variability,hindering the direct alignment between model simulations and observations,and thwarting conventional supervised learning methods.Here,we employ an unsupervised Cycle-consistent Generative Adversarial Network(CycleGAN),to correct daily Sea Surface Temperature(SST)simulations from the Community Earth System Model 2(CESM2).Our results reveal that the CycleGAN not only corrects climatological biases but also improves the simulation of major dynamic modes including the El Niño-Southern Oscillation(ENSO)and the Indian Ocean Dipole mode,as well as SST extremes.Notably,it substantially corrects climatological SST biases,decreasing the globally averaged Root-Mean-Square Error(RMSE)by 58%.Intriguingly,the CycleGAN effectively addresses the well-known excessive westward bias in ENSO SST anomalies,a common issue in climate models that traditional methods,like quantile mapping,struggle to rectify.Additionally,it substantially improves the simulation of SST extremes,raising the pattern correlation coefficient(PCC)from 0.56 to 0.88 and lowering the RMSE from 0.5 to 0.32.This enhancement is attributed to better representations of interannual,intraseasonal,and synoptic scales variabilities.Our study offers a novel approach to correct global SST simulations and underscores its effectiveness across different time scales and primary dynamical modes.

Artificial intelligence-assisted repair of peripheral nerve injury: a new research hotspot and associated challenges

Artificial intelligence can be indirectly applied to the repair of peripheral nerve injury.Specifically,it can be used to analyze and process data regarding peripheral nerve injury and repair,while study findings on peripheral nerve injury and repair can provide valuable data to enrich artificial intelligence algorithms.To investigate advances in the use of artificial intelligence in the diagnosis,rehabilitation,and scientific examination of peripheral nerve injury,we used CiteSpace and VOSviewer software to analyze the relevant literature included in the Web of Science from 1994–2023.We identified the following research hotspots in peripheral nerve injury and repair:(1)diagnosis,classification,and prognostic assessment of peripheral nerve injury using neuroimaging and artificial intelligence techniques,such as corneal confocal microscopy and coherent anti-Stokes Raman spectroscopy;(2)motion control and rehabilitation following peripheral nerve injury using artificial neural networks and machine learning algorithms,such as wearable devices and assisted wheelchair systems;(3)improving the accuracy and effectiveness of peripheral nerve electrical stimulation therapy using artificial intelligence techniques combined with deep learning,such as implantable peripheral nerve interfaces;(4)the application of artificial intelligence technology to brain-machine interfaces for disabled patients and those with reduced mobility,enabling them to control devices such as networked hand prostheses;(5)artificial intelligence robots that can replace doctors in certain procedures during surgery or rehabilitation,thereby reducing surgical risk and complications,and facilitating postoperative recovery.Although artificial intelligence has shown many benefits and potential applications in peripheral nerve injury and repair,there are some limitations to this technology,such as the consequences of missing or imbalanced data,low data accuracy and reproducibility,and ethical issues(e.g.,privacy,data security,research transparency).Future research should address the issue of data collection,as large-scale,high-quality clinical datasets are required to establish effective artificial intelligence models.Multimodal data processing is also necessary,along with interdisciplinary collaboration,medical-industrial integration,and multicenter,large-sample clinical studies.

ResoNet:Robust and Explainable ENSO Forecasts with Hybrid Convolution and Transformer Networks

Recent studies have shown that deep learning(DL)models can skillfully forecast El Niño–Southern Oscillation(ENSO)events more than 1.5 years in advance.However,concerns regarding the reliability of predictions made by DL methods persist,including potential overfitting issues and lack of interpretability.Here,we propose ResoNet,a DL model that combines CNN(convolutional neural network)and transformer architectures.This hybrid architecture enables our model to adequately capture local sea surface temperature anomalies as well as long-range inter-basin interactions across oceans.We show that ResoNet can robustly predict ENSO at lead times of 19 months,thus outperforming existing approaches in terms of the forecast horizon.According to an explainability method applied to ResoNet predictions of El Niño and La Niña from 1-to 18-month leads,we find that it predicts the Niño-3.4 index based on multiple physically reasonable mechanisms,such as the recharge oscillator concept,seasonal footprint mechanism,and Indian Ocean capacitor effect.Moreover,we demonstrate for the first time that the asymmetry between El Niño and La Niña development can be captured by ResoNet.Our results could help to alleviate skepticism about applying DL models for ENSO prediction and encourage more attempts to discover and predict climate phenomena using AI methods.

Artificial Intelligence Meets Flexible Sensors:Emerging Smart Flexible Sensing Systems Driven by Machine Learning and Artificial Synapses

The recent wave of the artificial intelligence(AI)revolution has aroused unprecedented interest in the intelligentialize of human society.As an essential component that bridges the physical world and digital signals,flexible sensors are evolving from a single sensing element to a smarter system,which is capable of highly efficient acquisition,analysis,and even perception of vast,multifaceted data.While challenging from a manual perspective,the development of intelligent flexible sensing has been remarkably facilitated owing to the rapid advances of brain-inspired AI innovations from both the algorithm(machine learning)and the framework(artificial synapses)level.This review presents the recent progress of the emerging AI-driven,intelligent flexible sensing systems.The basic concept of machine learning and artificial synapses are introduced.The new enabling features induced by the fusion of AI and flexible sensing are comprehensively reviewed,which significantly advances the applications such as flexible sensory systems,soft/humanoid robotics,and human activity monitoring.As two of the most profound innovations in the twenty-first century,the deep incorporation of flexible sensing and AI technology holds tremendous potential for creating a smarter world for human beings.

UAV-Assisted Dynamic Avatar Task Migration for Vehicular Metaverse Services: A Multi-Agent Deep Reinforcement Learning Approach

Avatars, as promising digital representations and service assistants of users in Metaverses, can enable drivers and passengers to immerse themselves in 3D virtual services and spaces of UAV-assisted vehicular Metaverses. However, avatar tasks include a multitude of human-to-avatar and avatar-to-avatar interactive applications, e.g., augmented reality navigation,which consumes intensive computing resources. It is inefficient and impractical for vehicles to process avatar tasks locally. Fortunately, migrating avatar tasks to the nearest roadside units(RSU)or unmanned aerial vehicles(UAV) for execution is a promising solution to decrease computation overhead and reduce task processing latency, while the high mobility of vehicles brings challenges for vehicles to independently perform avatar migration decisions depending on current and future vehicle status. To address these challenges, in this paper, we propose a novel avatar task migration system based on multi-agent deep reinforcement learning(MADRL) to execute immersive vehicular avatar tasks dynamically. Specifically, we first formulate the problem of avatar task migration from vehicles to RSUs/UAVs as a partially observable Markov decision process that can be solved by MADRL algorithms. We then design the multi-agent proximal policy optimization(MAPPO) approach as the MADRL algorithm for the avatar task migration problem. To overcome slow convergence resulting from the curse of dimensionality and non-stationary issues caused by shared parameters in MAPPO, we further propose a transformer-based MAPPO approach via sequential decision-making models for the efficient representation of relationships among agents. Finally, to motivate terrestrial or non-terrestrial edge servers(e.g., RSUs or UAVs) to share computation resources and ensure traceability of the sharing records, we apply smart contracts and blockchain technologies to achieve secure sharing management. Numerical results demonstrate that the proposed approach outperforms the MAPPO approach by around 2% and effectively reduces approximately 20% of the latency of avatar task execution in UAV-assisted vehicular Metaverses.

Dendritic Deep Learning for Medical Segmentation

Dear Editor,This letter presents a novel segmentation approach that leverages dendritic neurons to tackle the challenges of medical imaging segmentation.In this study,we enhance the segmentation accuracy based on a SegNet variant including an encoder-decoder structure,an upsampling index,and a deep supervision method.Furthermore,we introduce a dendritic neuron-based convolutional block to enable nonlinear feature mapping,thereby further improving the effectiveness of our approach.

ST-LSTM-SA:A New Ocean Sound Velocity Field Prediction Model Based on Deep Learning

The scarcity of in-situ ocean observations poses a challenge for real-time information acquisition in the ocean.Among the crucial hydroacoustic environmental parameters,ocean sound velocity exhibits significant spatial and temporal variability and it is highly relevant to oceanic research.In this study,we propose a new data-driven approach,leveraging deep learning techniques,for the prediction of sound velocity fields(SVFs).Our novel spatiotemporal prediction model,STLSTM-SA,combines Spatiotemporal Long Short-Term Memory(ST-LSTM) with a self-attention mechanism to enable accurate and real-time prediction of SVFs.To circumvent the limited amount of observational data,we employ transfer learning by first training the model using reanalysis datasets,followed by fine-tuning it using in-situ analysis data to obtain the final prediction model.By utilizing the historical 12-month SVFs as input,our model predicts the SVFs for the subsequent three months.We compare the performance of five models:Artificial Neural Networks(ANN),Long ShortTerm Memory(LSTM),Convolutional LSTM(ConvLSTM),ST-LSTM,and our proposed ST-LSTM-SA model in a test experiment spanning 2019 to 2022.Our results demonstrate that the ST-LSTM-SA model significantly improves the prediction accuracy and stability of sound velocity in both temporal and spatial dimensions.The ST-LSTM-SA model not only accurately predicts the ocean sound velocity field(SVF),but also provides valuable insights for spatiotemporal prediction of other oceanic environmental variables.