Low Computational Cost Research Articles

Salt-wedge estuary's response to rising sea level, reduced discharge, and Nature-Based Solution

Vulnerable estuaries face resilience challenges against climate-induced salinization. This study examines the Po di Goro estuary in the Northern Adriatic Sea using an innovative modeling approach. It assesses the effectiveness of a Nature-Based Solution in reducing the threat of salt-wedge intrusion. An intermediate-complexity numerical model is considered, leveraging its low computational cost, which is suitable for climate projections, along with robust physics encompassing the main estuarine processes. Two centennial climate experiments covering 1991–2100 are proposed following a mechanistic modeling approach to understand the compound effects of sea level rise and river discharge changes. The first experiment is a full forcing experiment. The second experiment uses the same model but removes the sea level rise as an input forcing. A third experiment, referred to as the Digital Twin Experiment, assesses the effectiveness of a location-specific Nature-Based Solution. This experiment specifically examines the impact of reducing salt levels in the water by using a halophyte plant along the estuary. The results show that, in a future climate change scenario, the salt-wedge intrusion increases. This response is due to the non-linear combination of reduced river discharge and the local sea level rise. The discharge decrease acts as the main driver in the mid-term future (i.e., 2050–2080). In the long-term future (i.e., 2080–2100), the local sea level rise becomes more relevant as the discharge trend is expected to be null. The salt-wedge intrusion in the Po di Goro is projected to increase up to 63% annually (120% in summertime). Additionally, the river mouth salinity could rise by 27% annually (69% in summertime) in the long-term future (2081–2100). The halophyte plant, Atriplex portulaciodes, proposed as Nature-Based Solution, could reduce salt-wedge intrusion in the Po di Goro by up to 16% annually (22% in summer) in the long-term future. In the short-term future, this Nature-Based Solution may be effective enough to counteract the salt increase.

The actuator farm model for large eddy simulation (LES) of wind-farm-induced atmospheric gravity waves and farm–farm interaction

Abstract. This study introduces the actuator farm model (AFM), a novel parameterization for simulating wind turbines within large eddy simulations (LESs) of wind farms. Unlike conventional models like the actuator disk (AD) or actuator line (AL), the AFM utilizes a single actuator point at the rotor center and only requires two to three mesh cells across the rotor diameter. Turbine force is distributed to the surrounding cells using a new projection function characterized by an axisymmetric spatial support in the rotor plane and Gaussian decay in the streamwise direction. The spatial support's size is controlled by three parameters: the half-decay radius r1/2, smoothness s, and streamwise standard deviation σ. Numerical experiments on an isolated National Renewable Energy Laboratory (NREL) 5MW wind turbine demonstrate that selecting r1/2=R (where R is the turbine radius), s between 6 and 10, and σ≈Δx/1.6 (where Δx is the grid size in the streamwise direction) yields wake deficit profiles, turbine thrust, and power predictions similar to those obtained using the actuator disk model (ADM), irrespective of horizontal grid spacing down to the order of the rotor radius. Using these parameters, LESs of a small cluster of 25 turbines in both staggered and aligned layouts are conducted at different horizontal grid resolutions using the AFM. Results are compared against ADM simulations employing a spatial resolution that places at least 10 grid points across the rotor diameter. The wind farm is placed in a neutral atmospheric boundary layer (ABL) with turbulent inflow conditions interpolated from a previous simulation without turbines. At horizontal resolutions finer than or equal to R/2, the AFM yields similar velocity, shear stress, turbine thrust, and power as the ADM. Coarser resolutions reveal the AFM's ability to accurately capture power at the non-waked wind farm rows, although it underestimates the power of waked turbines. However, the far wake of the cluster can be predicted well even when the cell size is of the order of the turbine radius. Finally, combining the AFM with a domain nesting method allows us to conduct simulations of two aligned wind farms in a fully neutral ABL and of wind-farm-induced atmospheric gravity waves under a conventionally neutral ABL, obtaining excellent agreement with ADM simulations but with much lower computational cost. The simulations highlight the AFM's ability to investigate the mutual interactions between large turbine arrays and the thermally stratified atmosphere.

Optimal design of experiments in the context of machine-learning inter-atomic potentials: improving the efficiency and transferability of kernel based methods.

Abstract Data-driven machine learning (ML) models of atomistic interactions are often based on flexible and non-physical functions that can relate nuanced aspects of atomic arrangements to predictions of energies and forces. As a result, these potentials are only as good as the training data (usually the results of so-called ab initio simulations), and we need to ensure that we have enough information to make a model sufficiently accurate, reliable and transferable. The main challenge stems from the fact that descriptors of chemical environments are often sparse, high-dimensional objects without a well-defined continuous metric. Therefore, it is rather unlikely that any ad hoc method for selecting training examples will be indiscriminate, and it is easy to fall into the trap of confirmation bias, where the same narrow and biased sampling is used to generate training and test sets. We will show that an approach derived from classical concepts of statistical planning of experiments and optimal design can help to mitigate such problems at a relatively low computational cost. The key feature of the method we will investigate is that it allows us to assess the quality of the data without obtaining reference energies and forces - a so-called offline approach. In other words, we are focusing on an approach that is easy to implement and doesn't require sophisticated frameworks that involve automated access to high performance computing (HPC).

Development of a Deep Learning-Based Flooding Region Segmentation Model for Recognizing Urban Flooding Situations

Urban flooding has become increasingly frequent due to the rising intensity of rainfall driven by urban development and climate change. Effective prevention measures are crucial to mitigate the significant human and material damages caused by such events. Rapid and accurate pre-detection techniques can help to reduce the impacts of urban flooding. With the advancement of deep learning, deep neural networks (DNNs) have been successfully applied across various domains, including computer vision and speech recognition. In particular, DNNs for computer vision demonstrate high performance with relatively low computational costs. In this paper, we propose a flooding region segmentation model for urban underpasses based on the U-Net architecture. To train and evaluate the model, we collected datasets from the Mannyeon, Oryang, and Daedong underpasses in Daejeon. The proposed method achieved Dice coefficients of 98.8%, 94.03%, and 93.85%, respectively. This model demonstrates high segmentation performance in detecting flooded regions and can be integrated into continuous flood monitoring systems.

Acceleration of Reaction Space Projector Analysis Using Combinatorial Optimization: Application to Organic Chemical Reactions.

In recent years, automated reaction path search methods have established the concept of a reaction route network. The Reaction Space Projector (ReSPer) visualizes the potential energy hypersurface into a lower-dimensional subspace using principal coordinates. The main time-consuming process in ReSPer is calculating the structural distance matrix, making it impractical for complex organic reaction route networks. We implemented the Alternate Optimization (AO) algorithm, one of the combinatorial optimizations, in ReSPer to reduce computational costs. Evaluations using gold clusters and the Au5 several reaction route networks showed that ReSPer-AO accurately computes distances with lower computational costs. Applying ReSPer-AO to the C5H8O reaction route network clarified dynamic conformation changes in its potential energy landscape. The ReSPer-AO method enables analysis of chemical reactions and dynamic conformations in a low-dimensional reaction space that accurately represents hydrocarbon reaction route networks.

EFC-Evolving Fuzzy Classifier with Incremental Clustering Algorithm Based on Samples Mean Value

This paper introduces a new multiclass classifier called the evolving Fuzzy Classifier (eFC). Starting its knowledge base from scratch, the eFC structure evolves based on a clustering algorithm that can add, merge, delete, or update clusters (= rules) simultaneously while providing class predictions. The procedure to add clusters uses the procrastination idea to prevent outliers from affecting the quality of learning. Two pruning mechanisms are used to maintain a concise and compact structure. In the first, redundant clusters are merged based on a similarity measure, and in the second, obsolete and unrepresentative clusters are excluded based on an inactivity strategy. The center of the clusters is adjusted based on the mean value of the attributes. The eFC model was evaluated and compared with state-of-the-art evolving fuzzy systems on 8 randomly selected data streams from the UCI and Kaggle repositories. The experimental results indicate that the eFC outperforms or is at least comparable to alternative state-of-the-art models. Specifically, the eFC achieved an average accuracy of 7% to 37% higher than the competing classifiers. The results and comparisons demonstrate that the eFC is a promising alternative for classification tasks in non-stationary environments, offering good accuracy, a compact structure, low computational cost, and efficient processing time.

Spin-symmetry-enforced solution of the many-body Schrödinger equation with a deep neural network.

The integration of deep neural networks with the variational Monte Carlo (VMC) method has marked a substantial advancement in solving the Schrödinger equation. In this work we enforce spin symmetry in the neural-network-based VMC calculation using a modified optimization target. Our method is designed to solve for the ground state and multiple excited states with target spin symmetry at a low computational cost. It predicts accurate energies while maintaining the correct symmetry in strongly correlated systems, even in cases in which different spin states are nearly degenerate. Our approach also excels at spin-gap calculations, including the singlet-triplet gap in biradical systems, which is of high interest in photochemistry. Overall, this work establishes a robust framework for efficiently calculating various quantum states with specific spin symmetry in correlated systems.

Future-proofing class-incremental learning

Exemplar-free class incremental learning is a highly challenging setting where replay memory is unavailable. Methods relying on frozen feature extractors have drawn attention recently in this setting due to their impressive performances and lower computational costs. However, those methods are highly dependent on the data used to train the feature extractor and may struggle when an insufficient amount of classes are available during the first incremental step. To overcome this limitation, we propose to use a pre-trained text-to-image diffusion model in order to generate synthetic images of future classes and use them to train the feature extractor. Experiments on the standard benchmarks CIFAR100 and ImageNet-Subset demonstrate that our proposed method can be used to improve state-of-the-art methods for exemplar-free class incremental learning, especially in the most difficult settings where the first incremental step only contains few classes. Moreover, we show that using synthetic samples of future classes achieves higher performance than using real data from different classes, paving the way for better and less costly pre-training methods for incremental learning.

ADS-B Communication Security: Assessing the Performance of Format-Preserving Lightweight Block Ciphers

In this work, we document an investigation regarding the vulnerabilities of ADS-B (automatic dependent surveillance-broadcast) communication, considering the risks to operational security. The ADS-B technology is a communication protocol to transmit surveillance information, that is, data related to position, altimetry, and speed, from the aircraft's avionics, an evolution to radars. We present an encryption solution for this communication using format-preserving encryption implemented in a microcontroller-embedded system. We also evaluate the use of lightweight symmetric block ciphers for better computational performance. When used as a pseudorandom function, we observe that such ciphers maintain a high entropy output value with low computational cost --- up to sixteen times faster for similar entropy values. Finally, the computational performance obtained with the proposed solution is analyzed by processing real-time data from aircraft in the landing and takeoff phase at the international airport of Recife/Guararapes - Gilberto Freyre, based in Recife/PE, Brazil.

Machine Learning With Tree Tensor Networks, CP Rank Constraints, and Tensor Dropout.

Tensor networks developed in the context of condensed matter physics try to approximate order- N tensors with a reduced number of degrees of freedom that is only polynomial in N and arranged as a network of partially contracted smaller tensors. As we have recently demonstrated in the context of quantum many-body physics, computation costs can be further substantially reduced by imposing constraints on the canonical polyadic (CP) rank of the tensors in such networks. Here, we demonstrate how tree tensor networks (TTN) with CP rank constraints and tensor dropout can be used in machine learning. The approach is found to outperform other tensor-network-based methods in Fashion-MNIST image classification. A low-rank TTN classifier with branching ratio b=4 reaches a test set accuracy of 90.3% with low computation costs. Consisting of mostly linear elements, tensor network classifiers avoid the vanishing gradient problem of deep neural networks. The CP rank constraints have additional advantages: The number of parameters can be decreased and tuned more freely to control overfitting, improve generalization properties, and reduce computation costs. They allow us to employ trees with large branching ratios, substantially improving the representation power.

A fast differential network with adaptive reference sample for gaze estimation

Most non-invasive gaze estimation methods do not consider the inter-individual differences in anatomical structure, but directly regress the gaze direction from the appearance image information, which limits the accuracy of individual-independent gaze estimation networks. In addition, existing gaze estimation methods tend to consider only how to improve the model’s generalization performance, ignoring the crucial issue of efficiency, which leads to bulky models that are difficult to deploy and have questionable cost-effectiveness in practical use. This paper makes the following contributions: (1) A differential network for gaze estimation using adaptive reference samples is proposed, which can adaptively select reference samples based on scene and individual characteristics. (2) The knowledge distillation is used to transfer the knowledge structure of robust teacher networks into lightweight networks so that our networks can execute quickly and at low computational cost, dramatically increasing the prospect and value of applying gaze estimation. (3) Integrating the above innovations, a novel fast differential neural network (Diff-Net) named FDAR-Net is constructed and achieved excellent results on MPIIGaze, UTMultiview and EyeDiap.

Insight into uranyl binding by cyclic peptides from molecular dynamics and density functional theory

It is a challenging task to develop uranyl-chelating agents based on peptide chemistry. A recently developed cationic dummy atom model of uranyl in conjunction with the classical molecular dynamics simulation presents a helpful utility to study the chelation of uranyl by peptides with a low computational cost. In the present study, it was used to describe the chelation of uranyl by the cyclic decapeptide with 4 Glu residues cyc-GluArgGluProGlyGluTrpGluProGly and its derivatives containing two phosphorylated serines in place of two Glu, termed pS16, pS18, pS38, and pS68. The obtained structures were further studied by density functional theory (DFT) and subsequent density analysis. We show that a combination of steered molecular dynamics and simulated annealing, using standard forcefields for peptide with the cationic dummy atom model of uranyl, can quickly and reliably obtain binding modes of uranyl-peptide complexes. Classical molecular dynamics simulation in explicit water produces geometry very close to the DFT-optimized structure. The presence of uranyl completely changes the conformation of these cyclic peptides from unstructured to organised. The simulation of a peptide with two uranyl units explained why only the 1: 1 ratio of peptide and chelated-uranyl is observed experimentally in most cases, by the insufficiency of the anionic residues for the chelation of two UO22+ units, but that pS16 can accommodate two such units.

A universal method for seizure onset zone localization in focal epilepsy using standard deviation of spike amplitude

BackgroundPrecisely localizing the seizure onset zone (SOZ) is critical for focal epilepsy surgery. Existing methods mainly focus on high-frequency activities in stereo-electroencephalography, but often fail when seizures are not driven by high-frequency activities. Recognized as biomarkers of epileptic seizures, ictal spikes in SOZ induce epileptiform discharges in other brain regions. Based on this understanding, we aim to develop a universal algorithm to localize SOZ and investigate how ictal spikes within the SOZ induce seizures. MethodsWe proposed a novel metric called standard deviation of spike amplitude (SDSA) and utilized channel-averaged SDSA to describe seizure processes and detect seizures. By integrating SDSA values in specific intervals, the score for each channel located within SOZ was calculated. Channels with high SOZ scores were clustered as SOZ. The localization accuracy was asserted using area under the receiver operating characteristic (ROC) curve. Further, we analyzed early ictal signals from SOZ channels and investigated factors influencing their duration to reveal the seizure inducing conditions. ResultsWe analyzed data from 15 patients with focal epilepsy. The channel-averaged SDSA successfully detected all 28 seizures without false alarms. Using SDSA integration, we achieved precise SOZ localization with an average area under ROC curve (AUC) of 0.96, significantly outperforming previous methods based on high-frequency activities. Further, we discovered that energy of ictal spikes in SOZ was concentrated at a specific frequency distributed in [6, 12Hz]. Additionally, we found that the higher the energy per second in this frequency band, the faster ictal spikes could induce seizures. ConclusionThe SDSA metric offered precise SOZ localization with robustness and low computational cost, making it suitable for clinical practice. By studying the propagation patterns of ictal spikes between the SOZ and non-SOZ, we suggest that ictal spikes from SOZ need to accumulate energy at a specific central frequency to induce epileptic spikes in non-SOZ, which may have significant implications for understanding the seizure onset pattern.

Simulating concrete cracking and failure under impact loading using a novel constitutive integration paradigm in non-ordinary state-based Peridynamics

Concrete cracking and failure under external loads are common in infrastructure, particularly under impact conditions where the variability of loads and the complexity of the cracking process present significant challenges. While the Bond-Based Peridynamic (BBPD) and Ordinary State-Based Peridynamic (OSBPD) models are widely used in concrete damage analysis due to their low computational cost, their simple constitutive laws limit their effectiveness in simulating concrete’s response to impact loading. To overcome this limitation, we propose a novel approach that integrates classical constitutive models into the Peridynamics (PD) framework. Specifically, we reformulate the Concrete Damage Plasticity Model 2 (CDPM2) within a non-ordinary state-based PD (NOSBPD) framework, resulting in the CDPM2-PD model. This model incorporates factors such as strain rate effects and strain hardening and introduces a bond damage criterion for both tensile and compressive damage. Through three numerical examples, the CDPM2-PD model demonstrates its ability to accurately capture the cracking and failure process of concrete under impact, showing strong agreement with experimental observations in areas such as failure mode transitions and crack patterns. These results validate the model’s effectiveness and offer a versatile method for integrating classical constitutive models into the NOSBPD framework for complex material analysis.

3D visualization diagnostics for lung cancer detection

<span>Lung cancer is a leading cause of cancer deaths worldwide with an estimated 2 million new cases and 1·76 million deaths yearly. Early detection can improve survival, and CT scans are a precise imaging technique to diagnose lung cancer. However, analyzing hundreds of 2D CT slices is challenging and can cause false alarms. 3D visualization of lung nodules can aid clinicians in detection and diagnosis. The MobileNet model integrates multi-view and multi-scale nodule features using depthwise separable convolutional layers. These layers split standard convolutions into depthwise and pointwise convolutions to reduce computational cost. Finally, the 3D pulmonary nodular models were created using a ray-casting volume rendering approach. Compared to other state-of-the-art deep neural networks, this factorization enables MobileNet to achieve a much lower computational cost while maintaining a decent degree of accuracy. The proposed approach was tested on an LIDC dataset of 986 nodules. Experiment findings reveal that MobileNet provides exceptional segmentation performance on the LIDC dataset, with an accuracy of 93.3%. The study demonstrates that the MobileNet detects and segments lung nodules somewhat better than other older technologies. As a result, the proposed system proposes an automated 3D lung cancer tumor visualization.</span>

Uncertainty quantification analysis using multi-fidelity deep neural network

Abstract In order to solve the highly-dimensional and highly-nonlinear problems in uncertainty quantification of CFD, the paper develops a multi-fidelity neural network model and proposes a residual connection-based multi-fidelity deep neural network fusion model. Based on the small sample size and low computational cost of multi-fidelity models, the residual structure is further utilized to accelerate the convergence rate of the model and reduce the possibility of inducing systemic risks. Finally, the feasibility and accuracy of the algorithm are verified by six typical numerical examples, and the uncertainty quantification analysis of the NACA0012 airfoil’s yaw moment coefficients is completed.

Dilated convolution neural operator for multiscale partial differential equations

This paper introduces a data-driven operator learning method for multiscale partial differential equations, with a particular emphasis on preserving high-frequency information. Drawing inspiration from the representation of multiscale parameterized solutions as a combination of low-rank global bases (such as low-frequency Fourier modes) and localized bases over coarse patches (analogous to dilated convolution), we propose the Dilated Convolutional Neural Operator (DCNO). The DCNO architecture effectively captures both high-frequency and low-frequency features while maintaining a low computational cost through a combination of convolution and Fourier layers. We conduct experiments to evaluate the performance of DCNO on various datasets, including the multiscale elliptic equation, its inverse problem, Navier–Stokes equation, and Helmholtz equation. Our results demonstrate that DCNO achieves an optimal balance between accuracy and computational efficiency, making it a promising approach for multiscale operator learning.

Not Another Dual Attention UNet Transformer (NNDA-UNETR): a plug-and-play parallel dual attention block in U-Net with enhanced residual blocks for medical image segmentation.

Medical image segmentation is crucial for clinical diagnostics and treatment planning. Recently, hybrid models often neglect the local modeling capabilities of Transformers for medical image segmentation, despite the complementary nature of local information from both convolutional neural networks (CNNs) and transformers. This limitation is particularly problematic in multi-organ segmentation, where organs are closely adhered, and accurate delineation is essential. This study aims to develop a novel method that leverages the strengths of both CNNs and transformers to address the challenges of segmenting multiple closely adhered organs, improving accuracy and robustness in multi-organ segmentation. In this study, we present the Not Another Dual Attention block (NNDA-block), a versatile, plug-and-play module that seamlessly integrates channel and spatial attention mechanisms. This block can be incorporated at any stage within the Not Another Dual Attention UNet Transformer (NNDA-UNETR) framework. Our novel approach to spatial attention uniquely combines local modeling, significantly reducing detail and texture loss during the up- and down-sampling processes typical in U-shaped architectures. Evaluations on the Beyond the Cranial Vault (BTCV) and Automatic Cardiac Diagnosis Challenge benchmark datasets demonstrate our method's effectiveness in balancing model complexity with accuracy. On the BTCV dataset, our model sets new state-of-the-art benchmarks, achieving a Dice similarity coefficient of 84.13%, normalized surface dice of 86.96%, mean average surface distance of 3.46 mm, and an Hausdorff distance at 95% of 17.76 mm. Moreover, it reduces the parameter count by 34.2% compared to leading contemporary methods, all while maintaining low computational costs [measured in floating point operations per second (FLOPs)]. NNDA-UNETR offers a robust solution for accurate segmentation in multi-organ tasks, particularly where organ adhesion poses challenges. Its lightweight design also makes it well-suited for deployment in real-world medical environments with limited computational resources.

Nonlinear subspace clustering by functional link neural networks

Nonlinear subspace clustering based on a feed-forward neural network has been demonstrated to provide better clustering accuracy than some advanced subspace clustering algorithms. While this approach demonstrates impressive outcomes, it involves a balance between effectiveness and computational cost. In this study, we employ a functional link neural network to transform data samples into a nonlinear domain. Subsequently, we acquire a self-representation matrix through a learning mechanism that builds upon the mapped samples. As the functional link neural network is a single-layer neural network, our proposed method achieves high computational efficiency while ensuring desirable clustering performance. By incorporating the local similarity regularization to enhance the grouping effect, our proposed method further improves the quality of the clustering results. We name our method as Functional Link Neural Network Subspace Clustering (FLNNSC). Furthermore, we propose a convex combination subspace clustering scheme that combines a linear subspace clustering method with the functional link neural network subspace clustering approach. This combination method is named as Convex Combination Subspace Clustering (CCSC), which allows for a dynamic balance between linear and nonlinear representations. Extensive experiments conducted on four widely used datasets, including Extended Yale B, USPS, COIL20, and ORL, demonstrate that both FLNNSC and CCSC outperform several state-of-art subspace clustering methods in terms of clustering accuracy. Our affinity graph experiments reveal that FLNNSC exhibits clear block diagonal structures. We provide recommendations for hyperparameters in FLNNSC by performing a parameter sensitivity analysis, and empirically verify the convergence of FLNNSC. Additionally, we show that FLNNSC has a lower computational cost compared to two high-performing methods.

A pseudo-density conjugate heat transfer method and its application to simulating the quasi-steady temperature field of a compressor cylinder

The substantial disparity in time scales between heat conduction and convection requires extremely tiny time steps in simulations, leading to significant computational demands when using current-tightly coupled methods to solve the temperature field of cylinders at the quasi-steady stage. A new pseudo-density method is proposed to accelerate the heat conduction rate. The influence of solid density on transient temperature under the periodic third-type boundary conditions is analyzed in a one-dimensional model by analytical methods and verified in a three-dimensional structure by numerical methods. It is found that reducing solid density can accelerate the heat conduction rate and increase the fluctuation amplitude of solid temperature. Therefore, a variable pseudo density is adopted for solids in the simulation of the transient heat transfer characteristics between the compressed gas and the cylinder. The results show that the pseudo-density method provides similar computational accuracy to the current tightly coupled methods, but at significantly lower computational cost. The fluctuation amplitude of the cylinder's temperature decreases rapidly in the wall thickness direction. The distribution of the convective heat transfer coefficient on the inner wall of the cylinder is extremely transient and highly non-uniform, which is consistent with the distribution of gas velocity near the cylinder wall.