A Comprehensive Survey of Recent Transformers in Image,Video and Diffusion Models

Transformer models have emerged as dominant networks for various tasks in computer vision compared to Convolutional Neural Networks(CNNs).The transformers demonstrate the ability to model long-range dependencies by utilizing a self-attention mechanism.This study aims to provide a comprehensive survey of recent transformerbased approaches in image and video applications,as well as diffusion models.We begin by discussing existing surveys of vision transformers and comparing them to this work.Then,we review the main components of a vanilla transformer network,including the self-attention mechanism,feed-forward network,position encoding,etc.In the main part of this survey,we review recent transformer-based models in three categories:Transformer for downstream tasks,Vision Transformer for Generation,and Vision Transformer for Segmentation.We also provide a comprehensive overview of recent transformer models for video tasks and diffusion models.We compare the performance of various hierarchical transformer networks for multiple tasks on popular benchmark datasets.Finally,we explore some future research directions to further improve the field.

A multilayer network diffusion-based model for reviewer recommendation

With the rapid growth of manuscript submissions,finding eligible reviewers for every submission has become a heavy task.Recommender systems are powerful tools developed in computer science and information science to deal with this problem.However,most existing approaches resort to text mining techniques to match manuscripts with potential reviewers,which require high-quality textual information to perform well.In this paper,we propose a reviewer recommendation algorithm based on a network diffusion process on a scholar-paper multilayer network,with no requirement for textual information.The network incorporates the relationship of scholar-paper pairs,the collaboration among scholars,and the bibliographic coupling among papers.Experimental results show that our proposed algorithm outperforms other state-of-the-art recommendation methods that use graph random walk and matrix factorization and methods that use machine learning and natural language processing,with improvements of over 7.62%in recall,5.66%in hit rate,and 47.53%in ranking score.Our work sheds light on the effectiveness of multilayer network diffusion-based methods in the reviewer recommendation problem,which will help to facilitate the peer-review process and promote information retrieval research in other practical scenes.

Research on Restoration of Murals Based on Diffusion Model and Transformer

Due to the limitations of a priori knowledge and convolution operation,the existing image restoration techniques cannot be directly applied to the cultural relics mural restoration,in order to more accurately restore the original appearance of the cultural relics mural images,an image restoration based on the denoising diffusion probability model(Denoising Diffusion Probability Model(DDPM))and the Transformer method.The process involves two steps:in the first step,the damaged mural image is firstly utilized as the condition to generate the noise image,using the time,condition and noise image patch as the inputs to the noise prediction network,capturing the global dependencies in the input sequence through the multi-attentionmechanismof the input sequence and feedforward neural network processing,and designing a long skip connection between the shallow and deep layers in the Transformer blocks between the shallow and deep layers using long skip connections to fuse the feature information of global and local outputs to maintain the overall consistency of the restoration results;In the second step,taking the noisy image as a condition to direct the diffusion model to back sample to generate the restored image.Experiment results show that the PSNR and SSIM of the proposedmethod are improved by 2%to 9%and 1%to 3.3%,respectively,which are compared to the comparison methods.This study proposed synthesizes the advantages of the diffusionmodel and deep learningmodel to make themural restoration results more accurate.

Dynamics modeling and optimal control for multi-information diffusion in Social Internet of Things

As an ingenious convergence between the Internet of Things and social networks,the Social Internet of Things(SIoT)can provide effective and intelligent information services and has become one of the main platforms for people to spread and share information.Nevertheless,SIoT is characterized by high openness and autonomy,multiple kinds of information can spread rapidly,freely and cooperatively in SIoT,which makes it challenging to accurately reveal the characteristics of the information diffusion process and effectively control its diffusion.To this end,with the aim of exploring multi-information cooperative diffusion processes in SIoT,we first develop a dynamics model for multi-information cooperative diffusion based on the system dynamics theory in this paper.Subsequently,the characteristics and laws of the dynamical evolution process of multi-information cooperative diffusion are theoretically investigated,and the diffusion trend is predicted.On this basis,to further control the multi-information cooperative diffusion process efficiently,we propose two control strategies for information diffusion with control objectives,develop an optimal control system for the multi-information cooperative diffusion process,and propose the corresponding optimal control method.The optimal solution distribution of the control strategy satisfying the control system constraints and the control budget constraints is solved using the optimal control theory.Finally,extensive simulation experiments based on real dataset from Twitter validate the correctness and effectiveness of the proposed model,strategy and method.

Spatial patterns of the Brusselator model with asymmetric Lévy diffusion

The formation of spatial patterns is an important issue in reaction–diffusion systems.Previous studies have mainly focused on the spatial patterns in reaction–diffusion models equipped with symmetric diffusion(such as normal or fractional Laplace diffusion),namely,assuming that spatial environments of the systems are homogeneous.However,the complexity and heterogeneity of spatial environments of biochemical reactions in vivo can lead to asymmetric diffusion of reactants.Naturally,there arises an open question of how the asymmetric diffusion affects dynamical behaviors of biochemical reaction systems.To answer this,we build a general asymmetric L´evy diffusion model based on the theory of a continuous time random walk.In addition,we investigate the two-species Brusselator model with asymmetric L´evy diffusion,and obtain a general condition for the formation of Turing and wave patterns.More interestingly,we find that even though the Brusselator model with symmetric diffusion cannot produce steady spatial patterns for some parameters,the asymmetry of L´evy diffusion for this model can produce wave patterns.This is different from the previous result that wave instability requires at least a three-species model.In addition,the asymmetry of L´evy diffusion can significantly affect the amplitude and frequency of the spatial patterns.Our results enrich our knowledge of the mechanisms of pattern formation.

2-D Modeling and Calculations of Stratospheric Ozone and Influences of Convection, Diffusion, and Time

An engineering system approach of 2-D cylindrical model of transient mass balance calculations of ozone and other concerned chemicals along with fourteen photolysis, ozone-generating and ozone-depleting chemical reaction equations was developed, validated, and used for studying the ozone concentrations, distribution and peak of the layer, ozone depletion and total ozone abundance in the stratosphere. The calculated ozone concentrations and profile at both the Equator and a 60˚N location were found to follow closely with the measured data. The calculated average ozone concentration was within 1% of the measured average, and the deviation of ozone profiles was within 14%. The monthly evolution of stratospheric ozone concentrations and distribution above the Equator was studied with results discussed in details. The influences of slow air movement in both altitudinal and radial directions on ozone concentrations and profile in the stratosphere were explored and discussed. Parametric studies of the influences of gas diffusivities of ozone DO3 and active atomic oxygen DO on ozone concentrations and distributions were also studied and delineated. Having both influences through physical diffusion and chemical reactions, the diffusivity (and diffusion) of atomic oxygen DO was found to be more sensitive and important than that of ozone DO3 on ozone concentrations and distribution. The 2-D ozone model present in this paper for stratospheric ozone and its layer and depletion is shown to be robust, convenient, efficient, and executable for analyzing the complex ozone phenomena in the stratosphere. .

BeFOI: A Novel Method Based on Conditional Diffusion Model for Medical Image Denoising

The progress in medical imaging technology highlights the importance of image quality for effective diagnosis and treatment.Yet,noise during capture and transmission can compromise image accuracy and reliability,complicating clinical decisions.The rising interest in diffusion models has led to their exploration of denoising images.We present Be-FOI(Better Fluoro Images),a weakly supervised model that uses cine images to denoise fluoroscopic images,both DR types.Trained through precise noise estimation and simulation,BeFOI employs Markov chains to denoise using only the fluoroscopic image as guidance.Our tests show that BeFOI outperforms other methods,reducing noise and enhancing clar-ity and diagnostic utility,making it an effective post-processing tool for medical images.

Analytical evaluation of steady-state solute distribution in through- diffusion and membrane behavior test under non-perfectly flushing boundary conditions

The through-diffusion and membrane behavior testing procedure using a closed-system apparatus has been widely used for concurrent measurement of diffusion and membrane efficiency coefficients of low-permeability clay-based barrier materials.However,the common assumption of perfectly flushing conditions at the specimen boundaries could induce errors in analyses of the diffusion coefficients and membrane efficiencies.In this study,an innovative pseudo three-dimensional(3D)analytical method was proposed to evaluate solute distribution along the boundary surfaces of the soil-porous disks system,considering the non-perfectly flushing conditions.The results were consistent with numerical models under two scenarios considering different inflow/outflow positions.The proposed model has been demonstrated to be an accurate and reliable method to estimate solute distributions along the bound-aries.The calculated membrane efficiency coefficient and diffusion coefficient based on the proposed analytical method are more accurate,resulting in up to 50%less relative error than the traditional approach that adopts the arithmetic mean value of the influent and effluent concentrations.The retar-dation factor of the clay specimen also can be calculated with a revised cumulative mass approach.Finally,the simulated transient solute transport matched with experimental data from a multi-stage through-diffusion and membrane behavior test,validating the accuracy of the proposed method.

Unveiling the Re,Cr,and I diffusion in saturated compacted bentonite using machine-learning methods

The safety assessment of high-level radioactive waste repositories requires a high predictive accuracy for radionuclide diffusion and a comprehensive understanding of the diffusion mechanism.In this study,a through-diffusion method and six machine-learning methods were employed to investigate the diffusion of ReO_(4)^(−),HCrO_(4)^(−),and I−in saturated compacted bentonite under different salinities and compacted dry densities.The machine-learning models were trained using two datasets.One dataset contained six input features and 293 instances obtained from the diffusion database system of the Japan Atomic Energy Agency(JAEA-DDB)and 15 publications.The other dataset,comprising 15,000 pseudo-instances,was produced using a multi-porosity model and contained eight input features.The results indicate that the former dataset yielded a higher predictive accuracy than the latter.Light gradient-boosting exhibited a higher prediction accuracy(R2=0.92)and lower error(MSE=0.01)than the other machine-learning algorithms.In addition,Shapley Additive Explanations,Feature Importance,and Partial Dependence Plot analysis results indicate that the rock capacity factor and compacted dry density had the two most significant effects on predicting the effective diffusion coefficient,thereby offering valuable insights.

LoRA模型微调Stable Diffusion的设计师风格服装生成方法

为控制Stable Diffusion模型准确生成特定设计师风格的服装图像,应用LoRA模型微调Stable Diffusion生成过程的方法实现风格化服装图像生成。以中国十佳时装服装设计师孙林的设计风格为例,首先对设计师的发布会服装进行系统分析,提取设计师的风格特征因子,根据设计师风格因子分别训练3个风格化LoRA模型,分别验证每个风格因子的表达效果。最后,通过3个LoRA模型叠加使用的方式,生成具有设计师风格特征的服装设计图像。结果表明:同时调用多个风格化LoRA模型微调Stable Diffusion的生成过程,能够使生成的服装符合设计师的服装风格,实现设计师风格的迁移设计,提高设计师进行服装系列设计和产品开发的效率。

Sulfate diffusion in coal pillar:experimental data and prediction model

The stability of coal pillar dams is crucial for the long-term service of underground reservoirs storing water or heat.Chemi-cal damage of coal dams induced by ions-atttacking in coal is one of the main reasons for the premature failure of coal dams.However,the diffusion process of harmful ions in coal is far from clear,limiting the reliability and durability of coal dam designs.This paper investigates sulfate diffusion in coal pillar through experimental and analytical methods.Coal specimens are prepared and exposed to sulfate solutions with different concentrations.The sulfate concentrations at different locations and time are measured.Based on experimental data and Fick's law,the time-dependent surface concentration of sulfate and diffusion coefficient are determined and formulated.Further,an analytical model for predicting sulfate diffusion in coal pillar is developed by considering dual time-dependent characteristics and Laplace transformations.Through comparisons with experimental data,the accuracy of the analytical model for predicting sulfate diffusion is verified.Further,sulfate diffusions in coal dams for different concentrations of sulfate in mine water are investigated.It has been found that the sulfate concen-tration of exposure surface and diffusion coefficient in coal are both time-dependent and increase with time.Conventional Fick's law is not able to predict the sulfate diffusion in coal pillar due to the dual time-dependent characteristics.The sulfate attacking makes the coal dam a typical heterogeneous gradient structure.For sulfate concentrations 0.01-0.20 mol/L in mine water,it takes almost 1.5 and 4 years for sulfate ions to diffuse 9.46 and 18.92 m,respectively.The experimental data and developed model provide a practical method for predicting sulfate diffusion in coal pillar,which helps the service life design of coal dams.

Molecular Dynamics, Physical Properties, Diffusion Coefficients and Activation Energy of the Lithium Oxide (Li-O) and Sodium Oxide (Na-O) Electrolyte (Cathode)

This work is a simulation model with the LAMMPS calculation code of an electrode based on alkali metal oxides (lithium, sodium and potassium) using the Lennard Jones potential. For a multiplicity of 8*8*8, we studied a gap-free model using molecular dynamics. Physical quantities such as volume and pressure of the Na-O and Li-O systems exhibit similar behaviors around the thermodynamic ensembles NPT and NVE. However, for the Na2O system, at a minimum temperature value, we observe a range of total energy values;in contrast, for the Li2O system, a minimum energy corresponds to a range of temperatures. Finally, for physicochemical properties, we studied the diffusion coefficient and activation energy of lithium and potassium oxides around their melting temperatures. The order of magnitude of the diffusion coefficients is given by the relation Dli-O >DNa-O for the multiplicity 8*8*8, while for the activation energy, the order is well reversed EaNa-O > EaLi-O.

Identification of the Diffusion Coefficients of Pollutants ( N O 3 - and Mn2+) through the Walls of Concrete Tank and Vulnerability of the Stored Water Quality

The populations of urban centers in Congo-Brazzaville have decided to develop various methods of water storage (concrete or masonry underground tanks) for domestic use, due to shortages in the distribution of water through the public network. However, questions remain as to the physico-chemical quality of the water stored in these tanks, when these structures are built in wet and relatively polluted areas. This paper presents a model of pollutant diffusion through the cementitious matrix (concrete) of tank walls simulated at a buried reservoir. The results of the experimental and numerical simulations show that certain concrete parameters, such as porosity, permeability and diffusivity, have a significant influence on the transfer of pollutants through the concrete walls, thus altering the physico-chemical quality of the stored water. The numerical models (1D) used to predict pollutant transfer and the quality of the stored water are consistent with those of the optimal control for identifying the diffusion coefficient. Major ion concentrations appear to be correlated with system porosity and diffusion coefficient. Nevertheless, the identification of the diffusion coefficient from the optimal control method, based on an explicit numerical resolution of a finite volume PDE for the approximation of the experiment, is not consistent with that of the optimal control method.

Determination of Natural Logarithm of Diffusion Coefficient and Activation Energy of Thin Layer Drying Process of Ginger Rhizome Slices

This study is an extension of the previous work done with ARS-680 Environmental Chamber. Drying is a complex operation that demands much energy and time. Drying is essentially important for preservation of ginger rhizome. Drying of ginger was modeled, and then the effective diffusion coefficient and activation energy were determined. For this purpose, the experiments were done at six levels of varied temperatures: 10°C, 20°C, 30°C, 40°C, 50°C and 60°C. The values of effective diffusion coefficients obtained in this work for the variously treated ginger rhizomes closely agreed with the average effective diffusion coefficients of other notable authors who determined the drying kinetics and convective heat transfer coefficients of ginger slices.

Numerical Simulation of Diffusion Type Traffic Flow Model Using Second-Order Lax-Wendroff Scheme Based on Exponential Velocity Density Function

In order to control traffic congestion, many mathematical models have been used for several decades. In this paper, we study diffusion-type traffic flow model based on exponential velocity density relation, which provides a non-linear second-order parabolic partial differential equation. The analytical solution of the diffusion-type traffic flow model is very complicated to approximate the initial density of the Cauchy problem as a function of x from given data and it may cause a huge error. For the complexity of the analytical solution, the numerical solution is performed by implementing an explicit upwind, explicitly centered, and second-order Lax-Wendroff scheme for the numerical solution. From the comparison of relative error among these three schemes, it is observed that Lax-Wendroff scheme gives less error than the explicit upwind and explicit centered difference scheme. The numerical, analytical analysis and comparative result discussion bring out the fact that the Lax-Wendroff scheme with exponential velocity-density relation of diffusion type traffic flow model is suitable for the congested area and shows a better fit in traffic-congested regions.

Study on the Mathematical Model of Technology Diffusion in Technical Economic System

In the technical and economic system,there is a certain interaction between various technologies,mainly in the form of technology diffusion.By constructing the mathematical model of technology diffusion,the trend of technology diffusion can be predicted,the reasonable distribution of science and technology productivity can be realized,and the policy optimization measures can be formulated on the basis of evaluating the existing policies.This paper first points out the basic idea of digital model of technology diffusion,and then focuses on the analysis of various parameters of mathematical model of technology diffusion,and puts forward the application conclusion of mathematical model of technology diffusion,in order to provide reference for relevant industry personnel.

基于Diffusion-Model的PM_(2.5)检测中数据样本扩充研究

在PM_(2.5)检测领域,生成对抗网络(GAN)是最为常见的数据集扩充方案,通常在GAN的训练中通过扩大生成器和鉴别器分布的支持度以防止鉴别器过拟合。然而,因难以找到合适的噪声分布,采用这种方法难以达到预期的数据集扩充目标。而结合Diffusion-Model框架构建的Diffusion-GAN数据生成模型可以缓解丢失问题来稳定训练梯度,并通过创建同一样本的不同噪声版本来增强数据,从而达到提高生成样本质量和生成器的多样性的目的。实验表明,Diffusion-GAN相较于基线算法,其训练性能和生成样本的保真度分别提高了15.1%和32.7%,能够满足PM_(2.5)空气质量检测数据扩充的需求。

基于CiteSpace对diffusion model的研究趋势分析

文章利用citespace以扩散模型(diffusion model)为标题,日期为2012—2022年的文献进行数据分析,分别对扩散模型文献的发文量、国家、机构、关键词、文献的引用等这几个方面进行了客观的分析,得到扩散模型的发文量一直呈上升的趋势,说明对扩散模型的研究一直是近几年学者研究的热点。通过对国家和机构的分布,以及发文量的分析,发现中国和美国对扩散模型的研究较多,且各个国家之间线条连接较紧密,说明国家之间对扩散模型的学术交流较广泛,机构之间的合作也较密切,其中中国科学院(Chinese Acad Sci)发文量最多,美国的主要是一些大学机构发表较多。通过对关键词的共现、聚类、突显词、引用文献的分析,可以知道2012—2022年扩散模型在医学类研究领域是一个热点主题,结合其它聚类的标签词,可以知道,扩散模型在生态领域的研究也较广泛,在传染病,机器学习类也受到学者广泛的青睐。

Dynamic vaccine distribution model based on epidemic diffusion rule and clustering approach

Due to the fact that the emergency medicine distribution is vital to the quick response to urgent demand when an epidemic occurs, the optimal vaccine distribution approach is explored according to the epidemic diffusion rule and different urgency degrees of affected areas with the background of the epidemic outbreak in a given region. First, the SIQR （susceptible, infected, quarantined,recovered） epidemic model with pulse vaccination is introduced to describe the epidemic diffusion rule and obtain the demanded vaccine in each pulse. Based on the SIQR model, the affected areas are clustered by using the self-organizing map （SOM） neutral network to qualify the results. Then, a dynamic vaccine distribution model is formulated, incorporating the results of clustering the affected areas with the goals of both reducing the transportation cost and decreasing the unsatisfied demand for the emergency logistics network. Numerical study with twenty affected areas and four distribution centers is carried out. The corresponding numerical results indicate that the proposed approach can make an outstanding contribution to controlling the affected areas with a relatively high degree of urgency, and the comparison results prove that the performance of the clustering method is superior to that of the non-clustering method on controlling epidemic diffusion.

Multi-scale simulation of diffusion behavior of deterrent in propellant

Concentration distribution of the deterrent in single-base propellant during the process of firing plays an important role in the ballistic properties of gun propellant in weapons. However, the diffusion coefficient calculated by molecular dynamics(MD) simulation is 6 orders of magnitude larger than the experimental values. Meanwhile, few simple and comprehensive theoretical models can explain the phenomenon and accurately predict the concentration distribution of the propellant. Herein, an onion model combining with MD simulation and finite element method of diffusion in propellants is introduced to bridge the gap between the experiments and simulations, and correctly predict the concentration distribution of deterrent. Furthermore, a new time scale is found to characterize the diffusion process. Finally, the time-and position-depended concentration distributions of dibutyl phthalate in nitrocellulose are measured by Raman spectroscopy to verify the correctness of the onion model. This work not only provides guidance for the design of the deterrent, but could be also extended to the diffusion of small molecules in polymer with different crystallinity.