Diffusion Model Research Articles

Bayesian confidence in optimal decisions.

The optimal way to make decisions in many circumstances is to track the difference in evidence collected in favor of the options. The drift diffusion model (DDM) implements this approach and provides an excellent account of decisions and response times. However, existing DDM-based models of confidence exhibit certain deficits, and many theories of confidence have used alternative, nonoptimal models of decisions. Motivated by the historical success of the DDM, we ask whether simple extensions to this framework might allow it to better account for confidence. Motivated by the idea that the brain will not duplicate representations of evidence, in all model variants decisions and confidence are based on the same evidence accumulation process. We compare the models to benchmark results, and successfully apply four qualitative tests concerning the relationships between confidence, evidence, and time, in a new preregistered study. Using computationally cheap expressions to model confidence on a trial-by-trial basis, we find that a subset of model variants also provide a very good to excellent account of precise quantitative effects observed in confidence data. Specifically, our results favor the hypothesis that confidence reflects the strength of accumulated evidence penalized by the time taken to reach the decision (Bayesian readout), with the penalty applied not perfectly calibrated to the specific task context. These results suggest there is no need to abandon the DDM or single accumulator models to successfully account for confidence reports. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Structure-aware diffusion for low-dose CT imaging

Reducing the radiation dose leads to the x-ray computed tomography (CT) images suffering from heavy noise and artifacts, which inevitably interferes with the subsequent clinic diagnostic and analysis. Leading works have explored diffusion models for low-dose CT imaging to avoid the structure degeneration and blurring effects of previous deep denoising models. However, most of them always begin their generative processes with Gaussian noise, which has little or no structure priors of the clean data distribution, thereby leading to long-time inference and unpleasant reconstruction quality. To alleviate these problems, this paper presents a Structure-Aware Diffusion model (SAD), an end-to-end self-guided learning framework for high-fidelity CT image reconstruction. First, SAD builds a nonlinear diffusion bridge between clean and degraded data distributions, which could directly learn the implicit physical degradation prior from observed measurements. Second, SAD integrates the prompt learning mechanism and implicit neural representation into the diffusion process, where rich and diverse structure representations extracted by degraded inputs are exploited as prompts, which provides global and local structure priors, to guide CT image reconstruction. Finally, we devise an efficient self-guided diffusion architecture using an iterative updated strategy, which further refines structural prompts during each generative step to drive finer image reconstruction. Extensive experiments on AAPM-Mayo and LoDoPaB-CT datasets demonstrate that our SAD could achieve superior performance in terms of noise removal, structure preservation, and blind-dose generalization, with few generative steps, even one step only.

Analytical and Numerical Conceptions of Flame Propagation in Porous Media with a Generalization in Diffusion and Reaction Modeling

ABSTRACT In this paper, we expand on previous studies to further describe the flame propagation dynamics in porous media. We develop a model that incorporates a p-Laplacian operator to represent both fast and slow diffusion effects. Additionally, we include an advection term and a modified reaction term to address a broader range of chemical reactions related to flame spread. Based on analytical assessments, we first ascertain fundamental properties of the model’s solutions, such as regularity, boundedness, and uniqueness. We also examine stationary solutions, deriving their approximate analytical forms using regular perturbation methods and analyzing them within a Hamiltonian framework. Finally, we compare these analytical solutions with numerical ones to determine the spatial region where the approximate analytical solutions are valid within acceptable levels of error.

Head and neck tumor segmentation from [18F]F-FDG PET/CT images based on 3D diffusion model

Objective. Head and neck (H&N) cancers are among the most prevalent types of cancer worldwide, and [18F]F-FDG PET/CT is widely used for H&N cancer management. Recently, the diffusion model has demonstrated remarkable performance in various image-generation tasks. In this work, we proposed a 3D diffusion model to accurately perform H&N tumor segmentation from 3D PET and CT volumes. Approach. The 3D diffusion model was developed considering the 3D nature of PET and CT images acquired. During the reverse process, the model utilized a 3D UNet structure and took the concatenation of 3D PET, CT, and Gaussian noise volumes as the network input to generate the tumor mask. Experiments based on the HECKTOR challenge dataset were conducted to evaluate the effectiveness of the proposed diffusion model. Several state-of-the-art techniques based on U-Net and Transformer structures were adopted as the reference methods. Benefits of employing both PET and CT as the network input, as well as further extending the diffusion model from 2D to 3D, were investigated based on various quantitative metrics and qualitative results. Main results. Results showed that the proposed 3D diffusion model could generate more accurate segmentation results compared with other methods (mean Dice of 0.739 compared to less than 0.726 for other methods). Compared to the diffusion model in 2D form, the proposed 3D model yielded superior results (mean Dice of 0.739 compared to 0.669). Our experiments also highlighted the advantage of utilizing dual-modality PET and CT data over only single-modality data for H&N tumor segmentation (with mean Dice less than 0.570). Significance. This work demonstrated the effectiveness of the proposed 3D diffusion model in generating more accurate H&N tumor segmentation masks compared to the other reference methods.

Image inpainting algorithm based on double curvature-driven diffusion model with P-Laplace operator.

The method of partial differential equations for image inpainting achieves better repair results and is economically feasible with fast repair time. Addresses the inability of Curvature-Driven Diffusion (CDD) models to repair complex textures or edges when the input image is affected by severe noise or distortion, resulting in discontinuous repair features, blurred detail textures, and an inability to deal with the consistency of global image content, In this paper, we have the CDD model of P-Laplace operator term to image inpainting. In this method, the P-Laplace operator is firstly introduced into the diffusion term of CDD model to regulate the diffusion speed; then the improved CDD model is discretized, and the known information around the broken region is divided into two weighted average iterations to get the inpainting image; finally, the final inpainting image is obtained by weighted averaging the two image inpainting images according to the distancing. Experiments show that the model restoration results in this paper are more rational in terms of texture structure and outperform other models in terms of visualization and objective data. Comparing the inpainting images with 150, 1000 and 100 iterations respectively, Total Variation(TV) model and the CDD model inpainting algorithm always has inpainting traces in details, and TV model can't meet the visual connectivity, but the algorithm in this paper can remove the inpainting traces well, TV model and the CDD model inpainting algorithm always have inpainting traces in details, and TV model can't meet the visual connectivity, but the algorithm in this paper can remove the inpainting traces well. Of the images used for testing, the highest PSNR reached 38.7982, SSIM reached 0.9407, and FSIM reached 0.9781, the algorithm not only inpainting the effect and, but also has fewer iterations.

Parallel diffusion models promote high detail-fidelity photoacoustic microscopy in sparse sampling

Reconstructing sparsely sampled data is fundamental for achieving high spatiotemporal resolution photoacoustic microscopy (PAM) of microvascular morphology in vivo. Convolutional networks (CNN) and generative adversarial networks (GAN) have been introduced to high-speed PAM, but due to the use of upsampling in CNN-based networks to restore details and the instability in GAN training, they struggle to learn the entangled microvascular network structure and vascular texture features, resulting in only achieving low detail-fidelity imaging of microvascular. The diffusion models is richly sampled and can generate high-quality images, which is very helpful for the complex vascular features in PAM. Here, we propose an approach named parallel diffusion models (PDM) with parallel learning of Noise task and Image task, where the Noise task optimizes through variational lower bounds to generate microvascular structures that are visually realistic, and the Image task improves the fidelity of the generated microvascular details through image-based loss. With only 1.56% of fully sampled pixels from photoacoustic human oral data, PDM achieves an LPIPS of 0.199. Additionally, using PDM in high-speed 16x PAM prevents breathing artifacts and image distortion issues caused by low-speed sampling, reduces the standard deviation of the Row-wise Self-Correlation Coefficient, and maintains high image quality. It achieves high confidence in reconstructing detailed information from sparsely sampled data and will promote the application of reconstructed sparsely sampled data in realizing high spatiotemporal resolution PAM.

The modelling error in multi-dimensional time-dependent solute transport models

Starting from full-dimensional models of solute transport, we derive and analyze multi-dimensional models of time-dependent convection, diffusion, and exchange in and around pulsating vascular and perivascular networks. These models are widely applicable for modelling transport in vascularized tissue, brain perivascular spaces, vascular plants and similar environments. We show the existence and uniqueness of solutions to both the full- and the multi-dimensional equations under suitable assumptions on the domain velocity. Moreover, we quantify the associated modelling errors by establishing a-priori estimates in evolving Bochner spaces. In particular, we show that the modelling error decreases with the characteristic vessel diameter and thus vanishes for infinitely slender vessels. Numerical tests in idealized geometries corroborate and extend upon our theoretical findings.

Identification of the coefficient in the diffusion model of hydrodynamic flow in a chemical reactor

This article analyzes in detail the actual problem of the spread of disinformation through various media resources and their impact on society. The main aspects of information consumption by Ukrainian society, in particular on the Internet, are highlighted, and the potential target audience is determined. Based on the analysis, a new intelligent system based on Natural Language Processing technology is proposed, which provides users with a comprehensive overview of the activity of each media resource. Available analogues with their advantages and limitations are examined in detail, emphasizing the significant advantages of the proposed intelligent system for increasing information literacy and countering disinformation. Refs.: 19 titles.

Image classification and reconstruction from low-density EEG

Recent advances in visual decoding have enabled the classification and reconstruction of perceived images from the brain. However, previous approaches have predominantly relied on stationary, costly equipment like fMRI or high-density EEG, limiting the real-world availability and applicability of such projects. Additionally, several EEG-based paradigms have utilized artifactual, rather than stimulus-related information yielding flawed classification and reconstruction results. Our goal was to reduce the cost of the decoding paradigm, while increasing its flexibility. Therefore, we investigated whether the classification of an image category and the reconstruction of the image itself is possible from the visually evoked brain activity measured by a portable, 8-channel EEG. To compensate for the low electrode count and to avoid flawed predictions, we designed a theory-guided EEG setup and created a new experiment to obtain a dataset from 9 subjects. We compared five contemporary classification models with our setup reaching an average accuracy of 34.4% for 20 image classes on hold-out test recordings. For the reconstruction, the top-performing model was used as an EEG-encoder which was combined with a pretrained latent diffusion model via double-conditioning. After fine-tuning, we reconstructed images from the test set with a 1000 trial 50-class top-1 accuracy of 35.3%. While not reaching the same performance as MRI-based paradigms on unseen stimuli, our approach greatly improved the affordability and mobility of the visual decoding technology.

Quantum‐Noise‐Driven Generative Diffusion Models

AbstractGenerative models realized with Machine Learning (ML) techniques are powerful tools to infer complex and unknown data distributions from a finite number of training samples in order to produce new synthetic data. Diffusion Models (DMs) are an emerging framework that have recently overcome Generative Adversarial Networks (GANs) in creating high‐quality images. Here, is proposed and discussed the quantum generalization of DMs, i.e., three Quantum‐Noise‐Driven Generative Diffusion Models (QNDGDMs) that could be experimentally tested on real quantum systems. The idea is to harness unique quantum features, in particular the non‐trivial interplay among coherence, entanglement, and noise that the currently available noisy quantum processors do unavoidably suffer from, in order to overcome the main computational burdens of classical diffusion models during inference. Hence, the suggestion is to exploit quantum noise not as an issue to be detected and solved but instead as a beneficial key ingredient to generate complex probability distributions from which a quantum processor might sample more efficiently than a classical one. Three examples of the numerical simulations are also included for the proposed approaches. The results are expected to pave the way for new quantum‐inspired or quantum‐based generative diffusion algorithms addressing tasks as data generation with widespread real‐world applications.

Enhance Ethanol Sensing Performance of Fe-Doped Tetragonal SnO2 Films on Glass Substrate with a Proposed Mathematical Model for Diffusion in Porous Media

This research enhances ethanol sensing with Fe-doped tetragonal SnO2 films on glass, improving gas sensor reliability and sensitivity. The primary objective was to improve the sensitivity and operational efficiency of SnO2 sensors through Fe doping. The SnO2 sensors were synthesized using a flexible and adaptable method that allows for precise doping control, with energy-dispersive X-ray spectroscopy (EDX) confirming homogeneous Fe distribution within the SnO2 matrix. A morphological analysis showed a surface structure ideal for gas sensing. The results demonstrated significant improvement in ethanol response (1 to 20 ppm) and lower temperatures compared to undoped SnO2 sensors. The Fe-doped sensors exhibited higher sensitivity, enabling the detection of low ethanol concentrations and showing rapid response and recovery times. These findings suggest that Fe doping enhances the interaction between ethanol molecules and the sensor surface, improving performance. A mathematical model based on diffusion in porous media was employed to further analyze and optimize sensor performance. The model considers the diffusion of ethanol molecules through the porous SnO2 matrix, considering factors such as surface morphology and doping concentration. Additionally, the choice of electrode material plays a crucial role in extending the sensor’s lifespan, highlighting the importance of material selection in sensor design.

Multi-period urban hydrogen refueling stations site selection and capacity planning with many-objective optimization for hydrogen supply chain

The location and quantity of hydrogen refueling stations (HRSs) play a crucial role in the development and promotion of hydrogen fuel cell vehicles (HFCVs). This study proposes a multi-period urban HRSs site selection and capacity planning with many-objective optimization framework for hydrogen supply chain (HSC). Firstly, the city's multi-period hydrogen requirement is predicted based on a generalized Bass diffusion model. Using publicly available data, including gas station network data, geographic information system (GIS) data, population data, and regional economic data, a spatially aggregated demand model is established to allocate hydrogen requirement at candidate sites in the city of the geographic grid model with a 1 km resolution. On this basis, four interrelated objective functions (total investment cost, hydrogen requirement coverage, risk coefficient, and environmental factors) are developed. The third-generation non-dominated sorting genetic algorithm (NSGA-III) and the technique for order preference by similarity to an ideal solution (TOPSIS) are employed to achieve multi-period urban HRSs site selection and capacity planning with many-objective optimization for HSC. By comparing the results of single-objective optimization focusing on minimum cost and maximum hydrogen requirement coverage, it is observed that many-objective optimization achieves a better balance among the four conflicting objectives. After comprehensive analysis, the distribution of HRSs exhibits a clustered structure, influenced by the population and economic structure of the city. In the initial stages of HFCV development, most stations serve the peripheral areas around the city center; while n the later stages of the optimization period, a few high-capacity HRSs begin to concentrate in the city center.

Whole-tumor histogram analysis of multiple non-Gaussian diffusion models at high b values for assessing cervical cancer.

To assess the diagnostic potential of whole-tumor histogram analysis of multiple non-Gaussian diffusion models for differentiating cervical cancer (CC) aggressive status regarding of pathological types, differentiation degree, stage, and p16 expression. Patients were enrolled in this prospective single-center study from March 2022 to July 2023. Diffusion-weighted images (DWI) were obtained including 15 b-values (0 ~ 4000s/mm2). Diffusion parameters derived from four non-Gaussian diffusion models including continuous-time random-walk (CTRW), diffusion-kurtosis imaging (DKI), fractional order calculus (FROC), and intravoxel incoherent motion (IVIM) were calculated, and their histogram features were analyzed. To select the most significant features and establish predictive models, univariate analysis and multivariate logistic regression were performed. Finally, we evaluated the diagnostic performance of our models by using receiver operating characteristic (ROC) analyses. 89 women (mean age, 55 ± 11 years) with CC were enrolled in our study. The combined model, which incorporated the CTRW, DKI, FROC, and IVIM diffusion models, offered a significantly higher AUC than that from any individual models (0.836 vs. 0.664, 0.642, 0.651, 0.649, respectively; p < 0.05) in distinguishing cervical squamous cell cancer from cervical adenocarcinoma. To distinguish tumor differentiation degree, except the combined model showed a better predictive performance compared to the DKI model (AUC, 0.839 vs. 0.697, respectively; p < 0.05), no significant differences in AUCs were found among other individual models and combined model. To predict the International Federation of Gynecology and Obstetrics (FIGO) stage, only DKI and FROC model were established and there was no significant difference in predictive performance among different models. In terms of predicting p16 expression, the predictive ability of DKI model is significantly lower than that of FROC and combined model (AUC, 0.693 vs. 0.850, 0.859, respectively; p < 0.05). Multiple non-Gaussian diffusion models with whole-tumor histogram analysis show great promise to assess the aggressive status of CC.

Parameter sharing fault data generation method based on diffusion model under imbalance data

Abstract Rotating machinery will inevitably fail under long-term heavy load working conditions. Obtaining enough data to train the deep learning model can enable managers to detect and deal with related failures in time, which greatly improves the safety of equipment operation. Mechanical fault samples are often much smaller than healthy samples. Traditional data enhancement methods mostly change the original data, but cannot improve the diversity of its features, so that the number of fault samples becomes larger, but the features remain unchanged. Aiming at the above problems, a diffusion model based on parameter sharing and inverted bottleneck residual structure (DDPM) is proposed. Firstly, the diffusion process gradually covers the original data with Gaussian noise, to learn the corresponding fault characteristics of the original data. In the diffusion process, the parameter sharing attention mechanism is embedded in the learning process of the diffusion process. Then, the feature extraction module is constructed by using the inverted bottleneck residual structure to enhance the learning ability of the network. After obtaining the fault characteristics of the original data, the reverse process of the results restores the Gaussian noise to data with different fault characteristics through the same steps as the diffusion process. By comparing the results of various generation models and analysing the characteristics of the generated data, the feasibility and universality of the proposed method in data augmentation tasks are verified.

Non-decision time: The Higgs Boson of decision.

Generative models of decision now permeate all subfields of psychology, cognitive, and clinical neuroscience. To successfully investigate decision mechanisms from behavior, it is necessary to assume the presence of delays prior and after the decision process itself. However, directly observing this "non-decision time (NDT)" from behavior long appeared beyond reach, the field mainly relying on models to estimate it. Here, we propose a biological definition of decision that includes perceptual discrimination and action selection, and in turn, explicitly equates NDT with the minimum sensorimotor delay, or "deadtime." We show how this delay is directly observable in behavioral data, without modeling assumptions, using the visual interference approach. We apply this approach to 11 novel and archival data sets from humans and monkeys gathered from multiple labs. We validate the method by showing that visual properties (brightness, color, size) consistently affect empirically measured visuomotor deadtime (VMDT), as predicted by neurophysiology. We then show that endogenous factors (strategic slowing, attention) do not affect VMDT. Therefore, VMDT consistently satisfies widespread selective influence assumptions, in contrast to NDT parameters from model fits. Last, contrasting empirically observed VMDT with NDT estimates from the EZ, drift diffusion, and linear ballistic accumulator models, we conclude that NDT parameters from these models are unlikely to consistently reflect visuomotor delays, neither at a group level nor for individual differences, in contrast to a widely held assumption. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Diffusion distance variations in coal pulverization based on equivalent matrix size: Implications for coal and gas outburst indicators

Coal matrix is rich in complex pores, constituting channels for gas diffusion. Diffusion distance plays a crucial role in diffusion resistance, directly influencing the intermediate processes of “desorption-diffusion-seepage” in coal. However, current research often quantifies diffusion distance using average particle size, failing to represent the true diffusion distance. In this study, we introduce a novel concept of equivalent matrix size based on gas desorption experiments. Diffusion coefficients were obtained by fitting the time-varying diffusion model, and the mass exchange of interporosity flow was estimated. Additionally, temporal matrix shape factors were analyzed to determine the matrix size with the aid of numerical simulation. Results demonstrate a temporal evolution in matrix shape factors, namely a decrease in the initial stage, stability in the medium stage, and inconsistency in the later stage. During pulverization, fracture and matrix damage occur simultaneously, with fractures undergoing more severe fragmentation. Lost time significantly impacts the accuracy of accessing lost gas amount ML and outburst indicator K1. Upon incorporating the matrix size, ML for XT and QD samples were overestimated by 0.83 cm3/g and 0.14 cm3/g, and K1 was overestimated by 21.2 % and 2.7 %, which could help establish a theoretical foundation for predicting coal and gas outbursts.

Amplitude equations for wave bifurcations in reaction–diffusion systems

A wave bifurcation is the counterpart to a Turing instability in reaction–diffusion systems, but where the critical wavenumber corresponds to a pure imaginary pair rather than a zero temporal eigenvalue. Such bifurcations require at least three components and give rise to patterns that are periodic in both space and time. Depending on boundary conditions, these patterns can comprise either rotating or standing waves. Restricting to systems in one spatial dimension, complete formulae are derived for the evaluation of the coefficients of the weakly nonlinear normal form of the bifurcation up to order five, including those that determine the criticality of both rotating and standing waves. The formulae apply to arbitrary n-component systems ( n⩾3 ) and their evaluation is implemented in software which is made available as supplementary material. The theory is illustrated on two different versions of three-component reaction–diffusion models of excitable media that were previously shown to feature super- and subcritical wave instabilities and on a five-component model of two-layer chemical reaction. In each case, two-parameter bifurcation diagrams are produced to illustrate the connection between complex dispersion relations and different types of Hopf, Turing, and wave bifurcations, including the existence of several codimension-two bifurcations.

Experimental and DFT calculation study on the efficient removal of high fluoride wastewater from metallurgical wastewater by kaolinite

Fluoride pollution and water scarcity are urgent issues. Reducing fluoride concentration in water is crucial. Kaolinite has been used to study adsorption and fluoride removal in water and to characterize material properties. The experimental results showed that the adsorption capacity of kaolinite decreased with increasing pH. The highest adsorption of fluoride occurred at pH 2, with a capacity of 11.1 mg/g. The fluoride removal efficiency remained high after four regeneration cycles. The fitting results with the Freundlich isotherm model and the external diffusion model showed that the non-homogeneous adsorption of kaolinite fit the adsorption behavior better. Finally, the adsorption mechanism was analyzed by FT-IR and XPS. The binding energies of various adsorption sites and the chemical adsorption properties of atomic states were discussed in relation to DFT calculations. The results showed that Al and H sites were the main binding sites, and the bonding stability for different forms of fluoride varies, with the size of Al–F (−7.498 eV) > H–F (−6.04 eV) > H–HF (−3.439 eV) > Al–HF (−3.283 eV). Furthermore, the density of states and Mulliken charge distribution revealed that the 2p orbital of F was found to be active in the adsorption process and was the main orbital for charge transfer.

Detecting images generated by diffusers

In recent years, the field of artificial intelligence has witnessed a remarkable surge in the generation of synthetic images, driven by advancements in deep learning techniques. These synthetic images, often created through complex algorithms, closely mimic real photographs, blurring the lines between reality and artificiality. This proliferation of synthetic visuals presents a pressing challenge: how to accurately and reliably distinguish between genuine and generated images. This article, in particular, explores the task of detecting images generated by text-to-image diffusion models, highlighting the challenges and peculiarities of this field. To evaluate this, we consider images generated from captions in the MSCOCO and Wikimedia datasets using two state-of-the-art models: Stable Diffusion and GLIDE. Our experiments show that it is possible to detect the generated images using simple multi-layer perceptrons (MLPs), starting from features extracted by CLIP or RoBERTa, or using traditional convolutional neural networks (CNNs). These latter models achieve remarkable performances in particular when pretrained on large datasets. We also observe that models trained on images generated by Stable Diffusion can occasionally detect images generated by GLIDE, but only on the MSCOCO dataset. However, the reverse is not true. Lastly, we find that incorporating the associated textual information with the images in some cases can lead to a better generalization capability, especially if textual features are closely related to visual ones. We also discovered that the type of subject depicted in the image can significantly impact performance. This work provides insights into the feasibility of detecting generated images and has implications for security and privacy concerns in real-world applications. The code to reproduce our results is available at: https://github.com/davide-coccomini/Detecting-Images-Generated-by-Diffusers.

Unraveling the enigmatic role of the subthalamic nucleus.

Subpopulations of neurons in the subthalamic nucleus have distinct activity patterns that relate to the three hypotheses of the Drift Diffusion Model.