Diffusion Model Research Articles

Measurement Guidance in Diffusion Models: Insight from Medical Image Synthesis.

In the field of healthcare, the acquisition of sample is usually restricted by multiple considerations, including cost, labor- intensive annotation, privacy concerns, and radiation hazards, therefore, synthesizing images-of-interest is an important tool to data augmentation. Diffusion models have recently attained state-of-the-art results in various synthesis tasks, and embedding energy functions has been proved that can effectively guide the pre-trained model to synthesize target samples. However, we notice that current method development and validation are still limited to improving indicators, such as Fréchet Inception Distance score (FID) and Inception Score (IS), and have not provided deeper investigations on downstream tasks, like disease grading and diagnosis. Moreover, existing classifier guidance which can be regarded as a special case of energy function can only has a singular effect on altering the distribution of the synthetic dataset. This may contribute to in-distribution synthetic sample that has limited help to downstream model optimization. All these limitations remind that we still have a long way to go to achieve controllable generation. In this work, we first conducted an analysis on previous guidance as well as its contributions on further applications from the perspective of data distribution. To synthesize samples which can help downstream applications, we then introduce uncertainty guidance in each sampling step and design an uncertainty-guided diffusion models. Extensive experiments on four medical datasets, with ten classic networks trained on the augmented sample sets provided a comprehensive evaluation on the practical contributions of our methodology. Furthermore, we provide a theoretical guarantee for general gradient guidance in diffusion models, which would benefit future research on investigating other forms of measurement guidance for specific generative tasks. Codes and models are available at: https://github.com/yangqy1110/MGDM.

LigandDiff: de Novo Ligand Design for 3D Transition Metal Complexes with Diffusion Models.

Transition metal complexes are a class of compounds with varied and versatile properties, making them of great technological importance. Their applications cover a wide range of fields, either as metallodrugs in medicine or as materials, catalysts, batteries, solar cells, etc. The demand for the novel design of transition metal complexes with new properties remains of great interest. However, the traditional high-throughput screening approach is inherently expensive and laborious since it depends on human expertise. Here, we present LigandDiff, a generative model for the de novo design of novel transition metal complexes. Unlike the existing methods that simply extract and combine ligands with the metal to get new complexes, LigandDiff aims at designing configurationally novel ligands from scratch, which opens new pathways for the discovery of organometallic complexes. Moreover, it overcomes the limitations of current methods, where the diversity of new complexes highly relies on the diversity of available ligands, while LigandDiff can design numerous novel ligands without human intervention. Our results indicate that LigandDiff designs unique and novel ligands under different contexts, and these generated ligands are synthetically accessible. Moreover, LigandDiff shows good transferability by generating successful ligands for any transition metal complex.

Analysis of the recording of Fibonacci lenses on photopolymers with 3-D diffusion model

In this work a three-dimensional diffusion model is used to model photopolymers as a recording media. This model allows us to predict the properties of the Diffractive Optical Elements (DOEs) once we recorded into the photopolymer. This model had never been tested with more complex elements, such as multifocal diffractive lenses, as presented in the following in this work. In addition, the model includes; the estimation of the refractive index modulation, the low-pass filtering effect due to the experimental optical setup, and the evolution of the transverse intensity distribution. In this way, the selection of the appropriate material characteristics depending on the intended DOE application is made possible. Specifically, an acrylamide-based PVA/AA photopolymer is simulated using the proposed model. Moreover, coverplating and index matching systems are considered together to avoid the effects of thickness variation. Furthermore, in order to compare their properties using the proposed model, we focus on Fibonacci lenses (FL), a type of bifocal lenses. This allows us to evaluate the dependence of the focii intensity on the polymerisation rate, the diffusivity parameter, low-pass filtering effect and the use of the index matching system for these lenses. This enables us to know the recording parameters in order to produce this type of multifocal diffractive lenses with higher quality and precision.

Sequestration of divalent heavy metal ions from aqueous environment by adsorption using biomass-bentonite composites as potential adsorbent: Equilibrium and kinetic studies

Hybrid clay adsorbents have been investigated in recent studies as cheap and effective adsorbents for eliminating micropollutants, such as heavy metal ions from aqueous solution. Here, we modified bentonite (BEN) with palm kernel shell and groundnut shell by calcination to obtain palm kernel shell-modified bentonite (BPKS) and groundnut shell-modified bentonite (BGS). We studied the batch equilibrium adsorption of Pb2+, Cd2+ and Ni2+ ions under the influence of dosage, solution pH, contact time and concentration. Under these conditions, we found that the modified adsorbents (BGS and BPKS) had a higher equilibrium adsorption (qe) for the metal ions than unmodified BEN. BEN exhibited Langmuir adsorption capacities (Qo) of 32.47, 14.04 and 14.16 mg/g for Pb2+, Cd2+ and Ni2+, respectively. BGS and BPKS had Qo values of 29.15, 14.27, 16.61 and 31.75, 17.67, 15.625 mg/g for Pb2+, Cd2+, Ni2+, respectively. Investigations revealed that the rate of adsorption on the three adsorbents is best described by a pseudo-second-order kinetic model, with R2 values ≥ 0.9; the intra-particle diffusion model establishes a surface interaction mechanism involving the exchange of electrons between the surfaces of the adsorbents and the adsorbate. Scanning electron microscopy results suggest porous and heterogeneous adsorbent surfaces, indicating a physio-sorption phenomenon. The modifications increased the surface area of BEN from 78.435 to 194.850 and 140.650 m2/g in BGS and BPKS, respectively, using BET surface area analysis. However, the average pore width of BEN reduces from 3.8 to 3.3 nm in both BGS and BPKS. Also, the modification enhanced the cation exchange capacity in BPKS only. The findings from this study offered invaluable insights into how biomass can enhance the physicochemical properties of BEN, as needed for practical environmental application and/or optimization, during the removal of Pb2+, Cd2+ and Ni2+ ions from aqueous media.

The Role of Convection in the Limit Shape of the Critical Front Profile for Born-Infeld Diffusion Models

AbstractIn this paper, we deal with models with Born-Infeld (or relativistic) type diffusion and monostable reaction, investigating the effect of the introduction of a convection term on the limit shape of the critical front profile for vanishing diffusion. We first provide an estimate of the critical speed and then, through a careful analysis of an equivalent first-order problem, we show that different convection terms may lead either to a complete sharpening of the limit profile or to its complete regularization, presenting some related numerical simulations.

Layered Patterns in Reaction–Diffusion Models with Perona–Malik Diffusions

AbstractIn this paper we deal with a reaction–diffusion equation in a bounded interval of the real line with a nonlinear diffusion of Perona–Malik’s type and a balanced bistable reaction term. Under very general assumptions, we study the persistence of layered solutions, showing that it strongly depends on the behavior of the reaction term close to the stable equilibria $$\pm 1$$ ± 1 , described by a parameter $$\theta >1$$ θ > 1 . If $$\theta \in (1,2)$$ θ ∈ ( 1 , 2 ) , we prove existence of steady states oscillating (and touching) $$\pm 1$$ ± 1 , called compactons, while in the case $$\theta =2$$ θ = 2 we prove the presence of metastable solutions, namely solutions with a transition layer structure which is maintained for an exponentially long time. Finally, for $$\theta >2$$ θ > 2 , solutions with an unstable transition layer structure persist only for an algebraically long time.

Persistence of a competition model of plankton allelopathy in time–space periodic environment

This work is devoted to the study of a competition model of plankton allelopathy imposed in time-space periodic environment. We prove that the system admits positive periodic solutions under certain conditions. We further obtain some sufficient conditions for the uniqueness and global stability of the positive periodic solution, which shows that the model is persistent. The main tools for our arguments are comparison theorems based on the maximum principle, sub- and supersolutions method, and an iteration method, which also permit the treatment of some more general reaction–diffusion models in periodic environment.

Analysis of pipeline leakage in unsaturated stratum: A new seepage-diffusion model

Accurate prediction of the spread of fluids leaking from underground pipelines is crucial for risk assessment. In this study, we developed a method for predicting the diffusion range of pipeline seepage fluids within an unsaturated stratum. First, we derived a seepage–diffusion equation based on the generalized Darcy’s law and subsequently analyzed the mechanism of fluid diffusion in the unsaturated stratum. We then constructed a model for the seepage–diffusion of a pipeline leakage fluid, incorporating variables such as the saturation, permeability coefficient, diffusion pressure, and diffusion distance of soil microelements across various time intervals, using a stepwise algorithm in tandem with the Green–Ampt model and VG–Mualen permeability coefficient function. We investigated the influence of fluid self-gravity, saturated permeability coefficient, initial saturation, and intra-pipe pressure on the diffusion distance of the pipeline leakage fluid in an unsaturated stratum, focusing on specific cases. The results indicate that an increase in the saturated permeability coefficient, initial saturation, and intra-pipe pressure leads to an increase in the fluid diffusion distance. A simulation test was conducted to validate the proposed seepage–diffusion model. The findings of this study can be employed to predict the diffusion range of pipeline leakage fluids in various formation types, providing a vital foundation for pipeline leakage accident management.

DCDIMB: Dynamic Community-based Diversified Influence Maximization using Bridge Nodes

Influence maximization (IM) is the fundamental study of social network analysis. The IM problem finds the top k nodes that have maximum influence in the network. Most of the studies in IM focus on maximizing the number of activated nodes in the static social network. But in real life, social networks are dynamic in nature. This work addresses the diversification of activated nodes in the dynamic social network. This work proposes an objective function that maximizes the number of communities by utilizing bridge nodes. We also propose a diffusion model that considers the role of inactive nodes in influencing a node. We prove the submodularity, and monotonicity of the objective function under the proposed diffusion model. This work analyzes the impact of different ratios of bridge nodes in the seed set on real-world and synthetic datasets. Further, we prove the NP-Hardness of the objective function under the proposed diffusion model. The experiments are conducted on various real-world and synthetic datasets with known and unknown community information. The proposed work experimentally shows that the objective function gives the maximum number of communities considering bridge nodes compared to the benchmark algorithms.

Diffusion behavior of rhodamine 6G in single octadecylsilyl-functionalized silica particle revealed by fluorescence correlation spectroscopy.

We investigated the diffusion behavior of rhodamine 6G (Rh6G) within single octadecylsilyl-functionalized (ODS) silica particle in an acetonitrile (ACN)/water system using fluorescence correlation spectroscopy (FCS). FCS measurements were conducted at the center of the particle to exclusively determine the intraparticle diffusion coefficient (D). The obtained D values were analyzed based on a pore and surface diffusion model, the results of which indicate that surface diffusion primarily governs the intraparticle diffusion of Rh6G. Furthermore, an increase in the concentration of ACN (CACN) resulted in a corresponding increase in the surface diffusion coefficient (Ds), whereas the addition of NaCl did not significantly affect the Ds values. We attributed this dependence of Ds to the dielectric constant change in the interfacial liquid phase formed on the ODS layer. Specially, Ds values of (4.0 ± 0.5) × 10-7, (7.7 ± 1.1) × 10-7, (1.0 ± 0.3) × 10-6, and (1.1 ± 0.2) × 10-6cm2s-1 were obtained for CACN = 20, 30, 40, and 50vol%, respectively. We anticipate that this approach will contribute to advancing research on intraparticle mass transfer mechanisms.

Leveraging diffusion models for unsupervised out-of-distribution detection on image manifold

Out-of-distribution (OOD) detection is crucial for enhancing the reliability of machine learning models when confronted with data that differ from their training distribution. In the image domain, we hypothesize that images inhabit manifolds defined by latent properties such as color, position, and shape. Leveraging this intuition, we propose a novel approach to OOD detection using a diffusion model to discern images that deviate from the in-domain distribution. Our method involves training a diffusion model using in-domain images. At inference time, we lift an image from its original manifold using a masking process, and then apply a diffusion model to map it towards the in-domain manifold. We measure the distance between the original and mapped images, and identify those with a large distance as OOD. Our experiments encompass comprehensive evaluation across various datasets characterized by differences in color, semantics, and resolution. Our method demonstrates strong and consistent performance in detecting OOD images across the tested datasets, highlighting its effectiveness in handling images with diverse characteristics. Additionally, ablation studies confirm the significant contribution of each component in our framework to the overall performance.

Plastic deformation of dry fine-grained olivine aggregates under high pressures

Abstract This study investigates the effect of pressure on diffusion creep of dry San Carlos and synthetic (prepared by sol-gel method) olivine. We prepared dry (water content < 9 ppm wt) fine-grained (< 1 μm grain-size) olivine and deformed the samples (both San Carlos and solgel olivine in the same assembly) in the same sample assembly under high-pressure (P = 2.9–8.8 GPa) and modest temperatures (T = 980–1250 K) at a fixed strain-rate. Evolution of strength was studied using the radial X-ray diffraction from various diffraction planes. We found that San Carlos and sol-gel olivine show similar rheological behaviour (when normalized to the same grain-size). Stress estimated by the radial X-ray diffraction increases with time and initially shows similar values for all diffraction planes. In many cases, stress values start to depend on the diffraction planes in the later stage and time dependence becomes minor. The micro-structural observations show that grain-size increases during an experiment. The results are interpreted using a theory of radial X-ray diffraction and the theoretical models of diffusion and dislocation creep. We conclude that the initial stage of deformation is by diffusion creep, but deformation in the later stage is by dislocation creep. For dislocation creep, our results are in reasonable agreement with previous low temperature dislocation creep results after a correction of temperature effect. For diffusion creep, we obtain an activation volume of 7.0 ± 2.4 cm3/mol that is substantially smaller than the values reported on dislocation creep but agrees well with the results on grain-growth. By comparing the present results on dry olivine with the previous results on wet (water-saturated) olivine, we found that water enhances diffusion creep but only modestly in comparison to dislocation creep. The difference in the pressure and water content dependence between diffusion and dislocation creep has an important influence on the dominant deformation mechanisms of olivine in the upper mantle.

Diclofenac sodium adsorption in aqueous media by activated carbon obtained from einkorn (Triticum monococcum L.) husk

AbstractThe uncontrolled release of waste diclofenac with low biodegradability is considered to be a potential threat for the environment and creatures. To find effective solution for this issue, this study reports the adsorption performance of diclofenac sodium salt (DCF) by using activated carbon (EHAC) obtained from einkorn (Triticum monococcum L.) husk in aqueous solution under various circumstances. It was found that DCF adsorption on EHAC was highly solution pH dependent, and DCF adsorption by EHAC decreased with increasing adsorption temperature. Equilibrium data showed that fitted isotherm model with the experiment results of DCF adsorption on EHAC followed the order of Langmuir > Temkin > Freundlich > Dubinin-Radushkevich. Adsorption capacity of EHAC for DCF adsorption in aqueous solution was calculated to be 147.06 mg/g at 25 °C. The adsorption kinetic of DCF adsorption on EHAC was determined to obey the pseudo-second-order kinetic model. By utilizing FTIR and pH data obtained from DCF adsorption on EHAC, DCF adsorption mechanisms with some interactions such as π-π stacking, electrostatic interactions, and hydrogen bonding were suggested at diverse pH values. Additionally, intraparticle diffusion model was applied to kinetic results to further recognize the kinetic mechanism of DCF adsorption on EHAC. Furthermore, thermodynamic parameters for DCF adsorption on EHAC were calculated and evaluated, in which DCF adsorption process by EHAC was determined to be exothermic, spontaneous, and feasible.

Learning to Choose: Behavioral Dynamics Underlying the Initial Acquisition of Decision Making

Current theories of decision making propose that decisions arise through competition between choice options. Computational models of the decision process estimate how quickly information about choice options is integrated and how much information is needed to trigger a choice. Experiments using this approach typically report data from well-trained participants. As such, we do not know how the decision process evolves as a decision-making task is learned for the first time. To address this gap, we used a behavioral design separating learning the value of choice options from learning to make choices. We trained male rats to respond to single visual stimuli with different reward values. Then, we trained them to make choices between pairs of stimuli. Initially, the rats responded more slowly when presented with choices. However, as they gained experience in making choices, this slowing reduced. Response slowing on choice trials persisted throughout the testing period. We found that it was specifically associated with increased exponential variability when the rats chose the higher value stimulus. Additionally, our analysis using drift diffusion modeling revealed that the rats required less information to make choices over time. These reductions in the decision threshold occurred after just a single session of choice learning. These findings provide new insights into the learning process of decision-making tasks. They suggest that the value of choice options and the ability to make choices are learned separately, and that experience plays a crucial role in improving decision-making performance.Significance StatementWe investigated the dynamics of decision-making as rats initially learned to choose between visual stimuli associated with different rewards. Unlike prior research focusing on well-trained participants, we explored the initial stages of learning to make decisions. We used a behavioral design that separated value learning from choice learning. Initially, rats exhibited slower responses when making choices, but this slowing diminished with experience. Response slowing persisted throughout the period of early choice learning. Drift diffusion modeling found reduced evidence of reduced information requirements for making choices over the period of early learning, with decision thresholds decreasing after just one choice-learning session. These studies revealed that experience significantly enhances decision-making, and shed light on the learning mechanisms that underlie decision-making tasks.

Dynamics Behavior of a Predatorg-Prey Diffusion Model Incorporating Hunting Cooperation and Predator-Taxis

In this paper, we consider a predator-prey diffusion model incorporating hunting cooperation and predator-taxis. Firstly, we establish the global existence of a classical solution for the model in any spatial dimension. Secondly, we analyze the stability/instability caused by predator-taxis, and we observe that predator-taxis play a key role in inducing stability changes. Specifically, if the positive equilibrium is stable for the corresponding reaction-diffusion model, the attractive predator-taxis can further stabilize the system, while the repulsive predator-taxis may lead to a change in spatial stability, if the positive equilibrium is unstable for the corresponding reaction-diffusion model, the attractive predator-taxis makes the model remain unstable, while the repulsive predator-taxis has a stabilizing effect. Finally, numerical simulations are employed to validate the obtained results.

Time series diffusion method: A denoising diffusion probabilistic model for vibration signal generation

Diffusion models have demonstrated powerful data generation capabilities in various research fields such as image generation. However, in the field of vibration signal generation, the criteria for evaluating the quality of the generated signal are different from that of image generation and there is a fundamental difference between them. At present, there is no research on the ability of diffusion model to generate vibration signal. In this paper, a Time Series Diffusion Method (TSDM) is proposed for vibration signal generation, leveraging the foundational principles of diffusion models. The TSDM uses an improved U-net architecture with attention block, ResBlock and TimeEmbedding to effectively segment and extract features from one-dimensional time series data. It operates based on forward diffusion and reverse denoising processes for time-series generation. Experimental validation is conducted using single-frequency, multi-frequency datasets, and bearing fault datasets. The results show that TSDM can accurately generate the single-frequency and multi-frequency features in the time series and retain the basic frequency features for the diffusion generation results of the bearing fault series. It is also found that the original DDPM could not generate high quality vibration signals, but the improved U-net in TSDM, which applied the combination of attention block and ResBlock, could effectively improve the quality of vibration signal generation. Finally, TSDM is applied to the small sample fault diagnosis of three public bearing fault datasets, and the results show that the accuracy of small sample fault diagnosis of the three datasets is improved by 32.380%, 18.355% and 9.298% at most, respectively.

A New Framework Combining Diffusion Models and the Convolution Classifier for Generating Images from EEG Signals

The generation of images from electroencephalography (EEG) signals has become a popular research topic in recent research because it can bridge the gap between brain signals and visual stimuli and has wide application prospects in neuroscience and computer vision. However, due to the high complexity of EEG signals, the reconstruction of visual stimuli through EEG signals continues to pose a challenge. In this work, we propose an EEG-ConDiffusion framework that involves three stages: feature extraction, fine-tuning of the pretrained model, and image generation. In the EEG-ConDiffusion framework, classification features of EEG signals are first obtained through the feature extraction block. Then, the classification features are taken as conditions to fine-tune the stable diffusion model in the image generation block to generate images with corresponding semantics. This framework combines EEG classification and image generation means to enhance the quality of generated images. Our proposed framework was tested on an EEG-based visual classification dataset. The performance of our framework is measured by classification accuracy, 50-way top-k accuracy, and inception score. The results indicate that the proposed EEG-Condiffusion framework can extract effective classification features and generate high-quality images from EEG signals to realize EEG-to-image conversion.

Influence, inertia, and independence: a diffusion model for temporal social networks

ABSTRACT In this work, we propose a diffusion model for temporal social networks and relate it to other well-known models of social influence by investigating its formal properties. The model establishes dyadic influence weights based on two antagonistic components: the susceptibility to be influenced (or, conversely, inertia with respect to the status quo) and becoming independent of prior influence. The proposed model generalizes the Friedkin-Johnsen model by the inertia with respect to the current influence relationships. We show that this generalization is an over-parameterization for static but not for dynamic influence networks. These findings suggest that the model at hand expands the set of existing social influence models in a non-trivial way.

Diffusion models as stochastic quantization in lattice field theory

In this work, we establish a direct connection between generative diffusion models (DMs) and stochastic quantization (SQ). The DM is realized by approximating the reversal of a stochastic process dictated by the Langevin equation, generating samples from a prior distribution to effectively mimic the target distribution. Using numerical simulations, we demonstrate that the DM can serve as a global sampler for generating quantum lattice field configurations in two-dimensional ϕ4 theory. We demonstrate that DMs can notably reduce autocorrelation times in the Markov chain, especially in the critical region where standard Markov Chain Monte-Carlo (MCMC) algorithms experience critical slowing down. The findings can potentially inspire further advancements in lattice field theory simulations, in particular in cases where it is expensive to generate large ensembles.

Innovations for Building Implementation Science Capacity among Researchers and Policymakers: The Depth and Diffusion Model

Innovations for Building Implementation Science Capacity among Researchers and Policymakers: The Depth and Diffusion Model