Diffusion Model Research Articles

Diffusion model of multi-agent collaborative behavior in public crisis governance network based on complex network evolutionary game

In order to explore the phenomenon of diffusion of group decision making formed by the emergence of decision-making behaviors of governance agents in public crisis governance systems, this research uses a complex network evolutionary game approach, considers BA scale-free networks as network vectors of public crisis governance systems, and develops a diffusion model of collaborative governance decision making behaviors. Simulation experiments are also conducted to show the macro-level impact of micro-subjects' decision-making behavior on group “Emergence-Diffusion”. The results of this study show that the cost of collaborative governance has the most significant effect on the depth and breadth of the spread of collaborative behavior in governance networks. The size of the network determines the speed of network diffusion. The smaller the network size, the more sensitive it is to the spillover benefits of collaborative governance, and the larger the network size, the more sensitive it is to the penalties of non-compliance. The findings of the study have implications for the collaborative behavior of multiple agents in public crisis governance. The main findings are that (1) in order to ensure the stability of the collaborative governance system, decision making options should be selected according to the size of the network. (2) A reasonable penalty mechanism for breach of contract should be set up to avoid the phenomenon of "free-riding" in collaborative governance. (3) Reasonable allocation of collaboration benefits and maintenance of cooperative relationships between nodes in the neighborhood. (4) External regulators should ensure that information in the network is disseminated without barriers and reduce the phenomenon of information asymmetry.

Geodetic Evidence for Cascading Landslide Motion Triggered by Extreme Rain Events at Joshimath, NW Himalaya

AbstractSlope instability due to tectonic, hydrological and anthropogenic activities cause severe landslides in Himalaya. Joshimath, a densely populated Himalayan town witnessed a catastrophic landslide event during December 2022 and January 2023 causing damages to ∼700 buildings. We use Interferometric synthetic aperture radar, Global Positioning System and rainfall measurements to probe the kinematics of the Joshimath landslide. We separate the seasonal and episodic deformation components using singular spectrum analysis. While the low amplitude annual landslide motions are modulated by seasonal precipitation, acceleration phases are triggered by extreme rain events. Our analysis revealed episodes of cascading motions triggered by extreme rain events resulting an overall increase in landslide velocity from −22 mm/yr during 2004–2010 to −325 mm/yr during 2022–2023. We estimate the landslide depth (∼30 m) and hydraulic diffusivity (∼3 × 10−5 m2/s) using a 1‐D pore‐water pressure diffusion model. Our study reveals the importance of systematic monitoring of ground deformation and weather parameters for landslide hazard mitigation.

Influence of salinity gradients on the diffusion of water and ionic species in dual porosity clay samples

Most of the available data on diffusion in natural clayey rocks consider tracer diffusion in the absence of a salinity gradient despite the fact that such gradients are frequently found in natural and engineered subsurface environments. To assess the role of such gradients on the diffusion properties of clayey materials, through-diffusion experiments were carried out in the presence and absence of a salinity gradient using salt-diffusion and radioisotope tracer techniques. The experiments were carried out with vermiculite samples that contained equal proportions of interparticle and interlayer porosities so as to assess also the role played by the two types of porosities on the diffusion of water and ions. Data were interpreted using both a classical Fickian diffusion model and with a reactive transport code, CrunchClay that can handle multi-porosity diffusion processes in the presence of charged surfaces. By combining experimental and simulated data, we demonstrated that (i) the flux of water diffusing through vermiculite interlayer porosity was minor compared to that diffusing through the interparticle porosity, and (ii) a model considering at least three types of porous volumes (interlayer, interparticle diffuse layer, and bulk interparticle) was necessary to reproduce consistently the variations of neutral and charged species diffusion as a function of salinity gradient conditions.

Breaking Through the Noisy Correspondence: A Robust Model for Image-Text Matching

Unleashing the power of image-text matching in real-world applications is hampered by noisy correspondence. Manually curating high-quality datasets is expensive and time-consuming, and datasets generated using diffusion models are not adequately well-aligned. The most promising way is to collect image-text pairs from the Internet, but it will inevitably introduce noisy correspondence. To reduce the negative impact of noisy correspondence, we propose a novel model that first transforms the noisy correspondence filtering problem into a similarity distribution modeling problem by exploiting the powerful capabilities of pre-trained models. Specifically, we use the Gaussian Mixture model to model the similarity obtained by CLIP as clean distribution and noisy distribution, to filter out most of the noisy correspondence in the dataset. Afterward, we used relatively clean data to fine-tune the model. To further reduce the negative impact of unfiltered noisy correspondence, i.e., a minimal part where two distributions intersect during the fine-tuning process, we propose a distribution-sensitive dynamic margin ranking loss, further increasing the distance between the two distributions. Through continuous iteration, the noisy correspondence gradually decreases and the model performance gradually improves. Our extensive experiments demonstrate the effectiveness and robustness of our model even under high noise rates.

In Search of Dispersed Memories: Generative Diffusion Models Are Associative Memory Networks.

Uncovering the mechanisms behind long-term memory is one of the most fascinating open problems in neuroscience and artificial intelligence. Artificial associative memory networks have been used to formalize important aspects of biological memory. Generative diffusion models are a type of generative machine learning techniques that have shown great performance in many tasks. Similar to associative memory systems, these networks define a dynamical system that converges to a set of target states. In this work, we show that generative diffusion models can be interpreted as energy-based models and that, when trained on discrete patterns, their energy function is (asymptotically) identical to that of modern Hopfield networks. This equivalence allows us to interpret the supervised training of diffusion models as a synaptic learning process that encodes the associative dynamics of a modern Hopfield network in the weight structure of a deep neural network. Leveraging this connection, we formulate a generalized framework for understanding the formation of long-term memory, where creative generation and memory recall can be seen as parts of a unified continuum.

Generalized Linear Driving Force Formulas for Diffusion and Reaction in Porous Catalysts

Approximate models are a fast and most often precise tool for determining the effectiveness factor for heterogeneous catalysis processes that are realized in the real world. They are also frequently applied as robust transient models describing the work of a single catalyst pellet or as a part of a more complex model, for example, a reactor model, where mass balances for the gas phase and solid phase are necessary. So far, approximate models for diffusion and reaction processes have been presented for processes described by a single balance equation. In the present work, approximate models without the mentioned limitation are presented and discussed. In addition, simple rules are shown for the development of other complex approximate models without tedious derivation in the complex domain. The formulas considered in this work are typical long-time approximations of the transient process. The accuracy is good, especially in the range of small and intermediate Thiele modulus values.

Evolutionary Game Analysis of Green Supply Chain Management Diffusion under Environmental Regulation

The continuous deterioration of the ecological environment and the increasing scarcity of resources have posed a serious challenge to the survival and development of human beings, and the implementation of green supply chain management (GSCM) in this context is an effective means to ensure the sustainable development of society and the economy. In order to seek the optimal strategy of evolutionary game in the implementation of green supply chain management and explore the influence of environmental regulation intensity and public preference degree on the evolution process of green supply chain management diffusion development, this paper takes the study of green supply chain management diffusion as the core innovation point, and under the premise of environmental regulation, selects the government, the core enterprise, and the public as the participating bodies of green supply chain management diffusion, and uses the theory of evolutionary game to construct a diffusion model of green supply chain management. Using evolutionary game theory to construct a diffusion model, and with the help of MATLAB and other mathematical tools for numerical simulation analysis, we discuss the diffusion of the green supply chain and derive the optimal combination strategy. The results of the study show that: (1) there are four evolutionary stable states in the process of green supply chain management diffusion: preliminary diffusion, extinction, semi-diffusion, and full diffusion; (2) it will be beneficial for the government to promote the evolutionary diffusion of green supply chain management by implementing a higher intensity of pollution tax policy while implementing green supply chain incentive strategies; (3) the government, while implementing environmental regulation policies, should also pay attention to the guidance of the public’s awareness of environmental friendliness and greenness, and focus on the role of the comprehensive strategy selection of the three parties of the game in reaching the optimal state. The conclusions of the study provide theoretical guidance and decision support for the implementation and diffusion of green supply chain management under environmental regulation.

Evolution of submarine canyons and hanging-wall fans: insights from geomorphic experiments and morphodynamic models

Abstract. Tectonics play a significant role in shaping the morphology of submarine canyons, which form essential links in source-to-sink (S2S) systems. It is difficult, however, to investigate the resulting morphodynamics over the long term. For this purpose, we propose a novel experimental approach that can generate submarine canyons and hanging-wall fans on continuously evolving active faults. We utilize morphometric analysis and morphodynamic models to understand the response of these systems to fault slip rate (Vr) and inflow discharge (Q). Our research reveals several key findings. Firstly, the fault slip rate controls the merging speed of submarine canyons and hanging-wall fans, which in turn affects their quantity and spacing. Additionally, the long profile shapes of submarine canyons and hanging-wall fans can be decoupled into a gravity-dominated breaching process and an underflow-dominated diffusion process, which can be described using a constant-slope relationship and a morphodynamic diffusion model, respectively. Furthermore, both experimental and simulated submarine canyon–hanging-wall fan long profiles exhibit strong self-similarity, indicating that the long profiles are scale independent. The Hack's scaling relationship established through morphometric analyses serves as an important link between different scales in S2S systems, bridging laboratory-scale data to field-scale data and submarine-to-terrestrial relationships. Lastly, for deep-water sedimentary systems, we propose an empirical formula to estimate fan volume using canyon length, and the data from 26 worldwide S2S systems utilized for comparison show a strong agreement. Our geomorphic experiments provide a novel perspective for better understanding of the influence of tectonics on deep-water sedimentary processes. The scaling relationships and empirical formulas we have established aim to assist in estimating volume information that is difficult to obtain during long-term landscape evolution processes.

By-product Eucalyptus leaves valorization in the basic dye adsorption: kinetic equilibrium and thermodynamic study

Abstract Algerian Eucalyptus Leaves (AEL), a natural biodegradable adsorbent abundantly available, was used for the removal of methylene blue (MB) dye. The AEL properties for the removal of MB were investigated under different conditions by varying the AEL amount, MB concentration, pH of the solution and the reaction temperature. Scanning electron microscopy (SEM) and infrared spectroscopy (FTIR) techniques have been used to characterize AEL biosorbent. Experimental results showed that the adsorption of MB dye at the concentration of 50 mg L−1 reached to 91 % at pH 10 with a stirring speed of 200 rpm and after 180 min of reaction time. The experimental data were analyzed using the linear forms of different kinetic models (pseudo-first order kinetic model, pseudo-second order kinetic model, and intra-particle diffusion models). The results demonstrated that the adsorption kinetics of MB was consistent with the pseudo-second order model with R 2 value of 0.9969. The isotherm models Langmuir, Freundlich, Dubinin, Elovich, Brunaut Emmet Teller and Temkin models were also investigated to describe the adsorption equilibrium. The results show that the AEL adsorption is in accordance with Temkin isotherm. The thermodynamic study revealed that the adsorption is spontaneous and exothermic. Therefore, as a cheap green adsorbent with high MB adsorption performance, AEL is expected to become one of the best candidate materials for future industrial wastewater treatment.

A New Robust Iterative Scheme Applied in Solving a Fractional Diffusion Model for Oxygen Delivery via a Capillary of Tissues

In this paper, we constructed a new and robust fixed point iterative scheme called the UO iterative scheme for the approximation of a contraction mapping. The scheme converges strongly to the fixed point of a contraction mapping. A rate of convergence result is shown with an example, and our scheme, when compared, converges faster than some existing iterative schemes in the literature. Furthermore, the stability and data dependence results are shown. Our new scheme is applied in the approximation of the solution to the oxygen diffusion model. Finally, our results are applied in the approximation of the solution to the boundary value problems using Green’s functions with an example.

A computer-vision-based deep learning model of smoke diffusion

Smoke is a lethal hazard in fire accidents and the biggest obstacle to emergency rescue. Predicting its diffusion range is important in fire command and emergency rescue. This article utilizes computer-vision-based deep learning methods to learn the characteristics of smoke diffusion and achieve real-time prediction of smoke-temperature fields. Firstly, a tunnel fire was simulated to verify the reliability of the deep learning model in predicting smoke-temperature fields. Thirty-six sets of fire simulation conditions were established with various values of the fire source location and heat release rate (HRR) and longitudinal ventilation airflow velocity. This created a database of 17,280 smoke-temperature fields. Based on windless scenario with the fire source located in the middle of the tunnel, it was determined that the SiLU activation function provides the best nonlinear modelling ability and stability in training neural networks. Based on the smoke-temperature-field database, a U-Net network was constructed to predict the smoke diffusion process of the tunnel fire. Under different scenarios, the U-Net model achieved 90 % accuracy in smoke-temperature-field prediction. The proposed U-Net model was then used to predict smoke dispersion in a petroleum storage depot scenario. The results show that this approach can help to delineate safe areas during fire emergencies.

Conditioning diffusion models via attributes and semantic masks for face generation

Deep generative models have shown impressive results in generating realistic images of faces. GANs managed to generate high-quality, high-fidelity images when conditioned on semantic masks, but they still lack the ability to diversify their output. Diffusion models partially solve this problem and are able to generate diverse samples given the same condition. This paper introduces a novel strategy for enhancing diffusion models through multi-conditioning, harnessing cross-attention mechanisms to utilize multiple feature sets, ultimately enabling the generation of high-quality and controllable images. The proposed method extends previous approaches by introducing conditioning on both attributes and semantic masks, ensuring finer control over the generated face images. In order to improve the training time and the generation quality, the impact of applying perceptual-focused loss weighting into the latent space instead of the pixel space is also investigated. The proposed solution has been evaluated on the CelebA-HQ dataset, and it can generate realistic and diverse samples while allowing for fine-grained control over multiple attributes and semantic regions. Experiments on the DeepFashion dataset have also been performed in order to analyze the capability of the proposed model to generalize to different domains. In addition, an ablation study has been conducted to evaluate the impact of different conditioning strategies on the quality and diversity of the generated images.

A data-driven model for damage evolution of bridge stay cable

Chloride ion concentration is crucial for the reliability of stay cables, particularly when their protective covers are compromised. Previous studies have focused on the corrosion of single wires or cables over time, proposing estimation models that primarily consider corrosion duration while neglecting other significant factors. Addressing this gap, this paper establishes a corrosion damage model for stay cables considering time, temperature, humidity and inclination angle based on the diffusion characteristics of chloride ions. Firstly, We conducted accelerated corrosion experiments on stay cables in different environments to assess the characteristics of the spatial distribution of chloride ion concentrations within the cables and the mass loss characteristics of the wires. Secondly, a three-dimensional diffusion model for chloride ions in stay cables is proposed based on Fick's second law. A pivotal innovation of our research is the application of diverse machine learning techniques to explore how the diffusion coefficient of chloride ions is influenced by factors like time and environmental temperature, leading to the formulation of a sophisticated chloride ion diffusion model. Finally, by amalgamating this diffusion model with the established corrosion law for steel reinforcement, we devised a comprehensive corrosion damage evolution model for the steel wires of stay cables. The result demonstrates the anisotropic diffusion of chloride ions and its substantial correlation with the corrosion damage observed in steel wires. Notably, the PSO-BPNN method emerged as a highly accurate predictor for the diffusion coefficient of chloride ions. The model established in this paper considers a variety of influencing factors and accurately predicts the corrosion damage characteristics of stay cables, which makes an important contribution to improving the reliability of stay cables.

Cinematographic Camera Diffusion Model

AbstractDesigning effective camera trajectories in virtual 3D environments is a challenging task even for experienced animators. Despite an elaborate film grammar, forged through years of experience, that enables the specification of camera motions through cinematographic properties (framing, shots sizes, angles, motions), there are endless possibilities in deciding how to place and move cameras with characters. Dealing with these possibilities is part of the complexity of the problem. While numerous techniques have been proposed in the literature (optimization‐based solving, encoding of empirical rules, learning from real examples,…), the results either lack variety or ease of control.In this paper, we propose a cinematographic camera diffusion model using a transformer‐based architecture to handle temporality and exploit the stochasticity of diffusion models to generate diverse and qualitative trajectories conditioned by high‐level textual descriptions. We extend the work by integrating keyframing constraints and the ability to blend naturally between motions using latent interpolation, in a way to augment the degree of control of the designers. We demonstrate the strengths of this text‐to‐camera motion approach through qualitative and quantitative experiments and gather feedback from professional artists. The code and data are available at https://github.com/jianghd1996/Camera-control.

On wavefront patterns in a fractional reaction–diffusion model for predator–prey system with anti-predator behavior

AbstractIn recent studies, much attention has been paid to reaction–diffusion systems with anomalous diffusion. In this paper, we investigate the formation of wavefront patterns in a predator–prey model with anti-predator behavior under the influence of anomalous subdiffusion. We use methods of traveling wave analysis and numerical integration to establish the existence of traveling wavefront solutions. Further, obtained traveling wavefront solutions are validated through direct computer simulations of time-dependent solutions for fractional partial differential equation system. It is found that wavefronts exist in a range of system parameters, which travel faster in the subdiffusive system than in the normal diffusive one.

Corrosion protection failure test analysis of the initial damaged cable ICCP mechanism

Given the corrosion protection failure of the cable-stayed bridge cable in the actual project, the impressed current cathodic protection (ICCP) technology was used to carry out the anti-corrosion tests from steel wire scale to cable scale, analyze the corrosion protection characteristics of steel wire ICCP, deduce the corrosion factor ICCP diffusion model, and reveal the space ICCP mechanism of the initial damaged cable. The analysis shows that ICCP can effectively inhibit the corrosive effect of steel wires and cables, and reduce the loss of mechanical properties of steel wires. Its protective efficiency increases with an increase in the absolute value of its protective potential and decreases with an increase in the dead load level. ICCP has a blocking effect on the diffusion of corrosion factors in the cable space, and the corrosion spatial distribution range, corrosion factor concentration, and diffusion coefficient of the initial damaged cable decrease significantly with the increase in the absolute value of the protection potential. The open circuit potential moves in a positive direction with the development of the corrosion process, and the blocking effect of corrosive factor diffusion is transformed from the joint participation by sacrificial anode cathodic protection and ICCP to the dominance of ICCP.

COSMIC: Molecular Conformation Space Modeling in Internal Coordinates with an Adversarial Framework.

The fast and accurate conformation space modeling is an essential part of computational approaches for solving ligand and structure-based drug discovery problems. Recent state-of-the-art diffusion models for molecular conformation generation show promising distribution coverage and physical plausibility metrics but suffer from a slow sampling procedure. We propose a novel adversarial generative framework, COSMIC, that shows comparable generative performance but provides a time-efficient sampling and training procedure. Given a molecular graph and random noise, the generator produces a conformation in two stages. First, it constructs a conformation in a rotation and translation invariant representation─internal coordinates. In the second step, the model predicts the distances between neighboring atoms and performs a few fast optimization steps to refine the initial conformation. The proposed model considers conformation energy, achieving comparable space coverage, and diversity metrics results.

Diffusion prediction of competitive information with time-varying attractiveness in social networks

The ubiquity of social media has facilitated the simultaneous dissemination of large-scale information within online social networks. By assuming that information attractiveness is static, numerous studies have been devoted to the analysis of multiple information dissemination. However, real-world information attractiveness often exhibits variations throughout the dissemination process. This paper delves into the study of how time-varying information attractiveness influences the diffusion process, particularly in the context of competitive information dissemination regarding multiple products. First, we propose a Markov multi-information diffusion model, incorporating three critical parameters: the attractiveness degree, the information-boom time, and the information prosperity index, to address the dynamic nature of attractiveness. The basic reproduction number is derived, and the accuracy of the proposed model is verified. Furthermore, our numerical simulation results illustrate that both the maximal attractiveness value and the information prosperity duration significantly enhance information competitiveness, while delayed information boom time may undermine this competitiveness. In addition, it is indicated that improving the peak attractiveness is the key for time-varying information to achieve a better spreading effect. Moreover, we find that the growth of information attractiveness can even mitigate the impact of blocking the spread caused by information discarding behavior, highlighting intrinsic quality as the paramount determinant of information promotion outcomes.

Performance of removing aqueous contaminant by zirconium based adsorbents: a critical review

The studies on materials for decontamination in aqueous solutions have increasingly received greater attentions. Such contaminants as heavy metals, arsenic, fluoride and phosphate are harmful to humans and aqueous species due to higher toxicity. Zirconium based adsorbents have become more attractive due to outstanding performance in decontamination. This article provides a comprehensive review of the performance and mechanisms of five types adsorbents: zirconium (hydro)oxides, zirconium hydrogen sulfate, zirconium based multiple metal typed adsorbents and zirconium impregnated complexes. The pseudo-first order and pseudo-second order equations and the intraparticle diffusion model can be applied in describing the adsorption kinetics, while Langmuir and Freundlich equations are the most commonly used adsorption isotherms. The important mechanisms for uptake of contaminants are: ligand exchange between adsorbate and adsorbent, surface complexation formation, and Lewis acid–base and electrostatic interactions. A series of successful studies demonstrate that the adsorbents are promising for removing aqueous contaminants.

Enhancing human–robot communication with a comprehensive language-conditioned imitation policy for embodied robots in smart cities

Integrating Embodied Robots into a smart city’s networked system can significantly enhance the city’s operational efficiency. These robots can be connected to the city’s network, receiving and transmitting data in real time for enhancing human–robot communications. Language-conditioned robot behavior plays a vital role in executing complex tasks by associating human commands or instructions with perception and actions. However, most language-conditioned policy-related research is limited to specific datasets and cannot generalize across different environments. In this study, we propose a novel imitation learning framework tailored for language-conditioned robotic tasks. Our framework includes specialized encoders designed for various benchmarks and utilizes two distinct models: the Transformer and Diffusion models. We rigorously evaluate this framework in three different robotic environments. Our findings indicate that the framework consistently delivers superior performance across multiple domains. Notably, we observe that the Transformer model is particularly effective in managing tasks with long trajectories, whereas the Diffusion model demonstrates enhanced proficiency in generating trajectories from limited training datasets. Our approach showcases remarkable generalization capabilities across a range of tasks and achieves significantly higher success rates in task completion.