High Computational Efficiency Research Articles

A flexible 2.5D medical image segmentation approach with in-slice and cross-slice attention

Deep learning has become the de facto method for medical image segmentation, with 3D segmentation models excelling in capturing complex 3D structures and 2D models offering high computational efficiency. However, segmenting 2.5D images, characterized by high in-plane resolution but lower through-plane resolution, presents significant challenges. While applying 2D models to individual slices of a 2.5D image is feasible, it fails to capture the spatial relationships between slices. On the other hand, 3D models face challenges such as resolution inconsistencies in 2.5D images, along with computational complexity and susceptibility to overfitting when trained with limited data. In this context, 2.5D models, which capture inter-slice correlations using only 2D neural networks, emerge as a promising solution due to their reduced computational demand and simplicity in implementation. In this paper, we introduce CSA-Net, a flexible 2.5D segmentation model capable of processing 2.5D images with an arbitrary number of slices. CSA-Net features an innovative Cross-Slice Attention (CSA) module that effectively captures 3D spatial information by learning long-range dependencies between the center slice (for segmentation) and its neighboring slices. Moreover, CSA-Net utilizes the self-attention mechanism to learn correlations among pixels within the center slice. We evaluated CSA-Net on three 2.5D segmentation tasks: (1) multi-class brain MR image segmentation, (2) binary prostate MR image segmentation, and (3) multi-class prostate MR image segmentation. CSA-Net outperformed leading 2D, 2.5D, and 3D segmentation methods across all three tasks, achieving average Dice coefficients and HD95 values of 0.897 and 1.40 mm for the brain dataset, 0.921 and 1.06 mm for the prostate dataset, and 0.659 and 2.70 mm for the ProstateX dataset, demonstrating its efficacy and superiority. Our code is publicly available at: https://github.com/mirthAI/CSA-Net.

CLAP. I. Resolving miscalibration for deep learning-based galaxy photometric redshift estimation

Obtaining well-calibrated photometric redshift probability densities for galaxies without a spectroscopic measurement remains a challenge. Deep learning discriminative models, typically fed with multi-band galaxy images, can produce outputs that mimic probability densities and achieve state-of-the-art accuracy. However, several previous studies have found that such models may be affected by miscalibration, an issue that would result in discrepancies between the model outputs and the actual distributions of true redshifts. Our work develops a novel method called the C ontrastive L earning and A daptive KNN for P hotometric Redshift (CLAP) that resolves this issue. It leverages supervised contrastive learning (SCL) and $k$-nearest neighbours (KNN) to construct and calibrate raw probability density estimates, and implements a refitting procedure to resume end-to-end discriminative models ready to produce final estimates for large-scale imaging data, bypassing the intensive computation required for KNN. The harmonic mean is adopted to combine an ensemble of estimates from multiple realisations for improving accuracy. Our experiments demonstrate that CLAP takes advantage of both deep learning and KNN, outperforming benchmark methods on the calibration of probability density estimates and retaining high accuracy and computational efficiency. With reference to CLAP, a deeper investigation on miscalibration for conventional deep learning is presented. We point out that miscalibration is particularly sensitive to the method-induced excessive correlations among data instances in addition to the unaccounted-for epistemic uncertainties. Reducing the uncertainties may not guarantee the removal of miscalibration due to the presence of such excessive correlations, yet this is a problem for conventional methods rather than CLAP. These discussions underscore the robustness of CLAP for obtaining photometric redshift probability densities required by astrophysical and cosmological applications. This is the first paper in our series on CLAP.

Multicolor wavefront sensing using Talbot effect for high-order harmonic generation

High-order harmonic generation (HHG) using femtosecond lasers is an efficient and cost-effective method for producing attosecond and femtosecond light in the extreme ultraviolet and soft x-ray regions. This technique generates high-brightness harmonic beams with stable spectra, intensity, and wavefronts, making it a valuable tool for applications such as nanoimaging, material metrology, and studies of ultrafast dynamics. Accurate wavefront measurement of HHG beams is essential for analyzing complex incident fields, understanding atomic responses, and enhancing applications such as attosecond-driven ultrafast processes. However, quantifying the wavefronts of each harmonic component remains a challenge. Here, we present a multicolor wavefront sensing for HHG beams using the Talbot effect, validated theoretically and experimentally. Our method reconstructs individual high-order harmonic wavefronts with a spectral resolution of 4.8 eV, capturing multiple harmonic wavelengths in a single scan along the propagation direction, without the need for additional spectrometers or complex calibrations. This approach surpasses conventional Talbot effect wavefront sensing, which typically relies on quasimonochromatic waves or wavelength-averaging. Additionally, we introduce a high-accuracy centroid detection strategy that reduces tracking errors to the thousandths of a pixel while maintaining high computational efficiency, paving the way for broader applications in imaging and metrology. Published by the American Physical Society 2024

Automatically Differentiable Higher-Order Parabolic Equation for Real-Time Underwater Sound Speed Profile Sensing

This paper is dedicated to the acoustic inversion of the vertical sound speed profiles (SSPs) in the underwater marine environment. The method of automatic differentiation is applied for the first time in this context. Representing the finite-difference Padé approximation of the propagation operator as a computational graph allows for the analytical computation of the gradient with respect to the SSP directly within the numerical scheme. The availability of the gradient, along with the high computational efficiency of the numerical method used, enables rapid inversion of the SSP based on acoustic measurements from a hydrophone array. It is demonstrated that local optimization methods can be effectively used for real-time sound speed inversion. Comparative analysis with existing methods shows the significant superiority of the proposed method in terms of computation speed.

Spectral element method for random vibration analysis of the vehicle-track coupling system

In this paper, we propose a new random vibration analysis model for the vertical vehicle-track coupling system, which combines the symplectic method and the spectral element method (SEM) to enhance calculation efficiency and accuracy. The model assumed that the vehicle is composed of a multi-rigid body with 10 DOFs, and its motion equation is derived using the traditional frequency analysis method. Tracks are considered periodic structures, and the part between sleeper support points is considered a substructure. The motion equation of the periodic track-substructure is established by spectral element method (SEM), and the propagation of vibration wave between the substructures is calculated by the symplectic method. The frequency responses of the vehicle-track coupling system are calculated by using the German low-interference irregularity spectrum and short-wave irregularity spectrum as excitation. Then, the rationality of this method is verified by comparing it with the results from the traditional symplectic method and simulations using Universal Mechanism software. The comparison shows that this method surpasses the traditional symplectic method in high-frequency accuracy while maintaining high computational efficiency. Although the finite element method also has the capability for high-frequency analysis, this method can calculate the high-frequency vibration of an infinitely long track using just one element, making it more suitable for the efficient analysis of vehicle-track coupling systems. Finally, the model is applied to analyze the distribution characteristics of the vibration response of the coupled system in the frequency domain.

Leveraging Machine Learning for Photometric Redshift Estimation of JWST Galaxies

With the launch of JWST, the volume and complexity of astronomical data are increasing, a trend that will continue with future instruments such as SKA and Euclid. It is inevitable that data-driven methods will become more prominent alongside model-driven analysis. This research utilises machine learning, specifically a kernelised local linear regression model, in photometric redshift predictions of galaxies observed with JWST. With a spectroscopic dataset of 4605 galaxies, we trained the model and achieved a deviation of σdz = 0.018 and a catastrophic outlier rate of foutlier = 3.5%. These results demonstrate high accuracy and computational efficiency, highlighting the potential of machine learning for astronomical data analysis, in particular for large-scale surveys.

A fast temporal multiple sparse Bayesian learning-based channel estimation method for time-varying underwater acoustic OFDM systems

In this paper, a fast temporal multiple sparse Bayesian learning (FTMSBL)-based channel estimation method for underwater acoustic (UWA) orthogonal frequency division multiplexing (OFDM) systems is proposed, which is optimized using the fast marginalized likelihood maximization method. The algorithm fully uses the consistent sparse structure and time-domain correlation properties of channels to improve the reconstruction performance and computational efficiency, offering better performance and higher computational efficiency than the traditional Bayesian learning algorithms. At the same time, the FTMSBL algorithm does not require computing the inverse of large matrices and consumes very little storage resources in the operation, making it suitable for hardware implementation. Simulation and sea trial results show that the FTMSBL-based underwater channel estimation algorithm achieves higher channel estimation accuracy than the orthogonal matching tracking algorithm, and the system bit error rate (BER) is significantly reduced; specifically, the FTMSBL algorithm can achieve optimal performance in strong time-dependent channels.

Drop Tests and Numerical Analyses of Composite Beams with Corrugated Web

To evaluate the energy absorption ability of composite beams with corrugated web under impact, the drop tests were conducted to demonstrate the fracture modes and obtain the load and energy history curves. Based on the drop tests, a finite element (FE) model was proposed with a weak link designed to trigger crushing, developed using Abaqus/Explicit, which integrates cohesive elements and employs the improved Hashin criterion along with the Choi–Chang criterion to accurately predict the onset and progression of delamination. The model accounts for the anisotropy and progressive damage in composite laminates. The periodic boundary conditions (PBCs) were integrated into the single-wave corrugated web beam model, the inherent limitations of which were corrected, enabling it to simulate the crushing process with the same effectiveness as the multi-wave model. As a result, the proposed model can accurately depict the progressive crushing behavior of corrugated web beams under impact loads while maintaining high computational efficiency.

LT-DeepLab: an improved DeepLabV3+ cross-scale segmentation algorithm for Zanthoxylum bungeanum Maxim leaf-trunk diseases in real-world environments

IntroductionZanthoxylum bungeanum Maxim is an economically significant crop in Asia, but large-scale cultivation is often threatened by frequent diseases, leading to significant yield declines. Deep learning-based methods for crop disease recognition have emerged as a vital research area in agriculture.MethodsThis paper presents a novel model, LT-DeepLab, for the semantic segmentation of leaf spot (folium macula), rust, frost damage (gelu damnum), and diseased leaves and trunks in complex field environments. The proposed model enhances DeepLabV3+ with an innovative Fission Depth Separable with CRCC Atrous Spatial Pyramid Pooling module, which reduces the structural parameters of Atrous Spatial Pyramid Pooling module and improves cross-scale extraction capability. Incorporating Criss-Cross Attention with the Convolutional Block Attention Module provides a complementary boost to channel feature extraction. Additionally, deformable convolution enhances low-dimensional features, and a Fully Convolutional Network auxiliary header is integrated to optimize the network and enhance model accuracy without increasing parameter count.ResultsLT-DeepLab improves the mean Intersection over Union (mIoU) by 3.59%, the mean Pixel Accuracy (mPA) by 2.16%, and the Overall Accuracy (OA) by 0.94% compared to the baseline DeepLabV3+. It also reduces computational demands by 11.11% and decreases the parameter count by 16.82%.DiscussionThese results indicate that LT-DeepLab demonstrates excellent disease segmentation capabilities in complex field environments while maintaining high computational efficiency, offering a promising solution for improving crop disease management efficiency.

Position and orientation estimation method based on 3D digital morphology contour registration

Abstract Accurately and quickly obtaining the positions and orientations of mechanical parts based on the digital morphologies of mechanical parts are the key to achieving efficient and accurate assembly of mechanical parts. However, due to poor robustness and compactness in extracting digital morphology contours of mechanical parts, the accuracy of assembly positions and orientations obtained by using digital morphology contours cannot meet the requirements of high-precision assembly. Therefore, this paper proposes a position and orientation estimation method based on 3D digital morphology contour registration. This method extracts and optimizes digital morphology contours of mechanical parts and obtains the assembly positions and orientations by using an improved iterative closest point method to register the extracted digital morphology contours with those of the mechanical parts in assembly targets with desired positions and orientations. Experiments are conducted using mechanical parts from the ABC dataset and inertial confinement fusion micro-target. From the experimental results, when using the assembly positions and orientations obtained through the method proposed in this paper to assemble mechanical parts, it can achieve translation absolute errors of 8 μm, 5 μm, and 9 μm along the X-, Y-, and Z-axes, respectively. Similarly, the angular absolute errors in rotations around the Z-, Y-, and X-axes can be less than or equal to 0.16°, 0.15°, and 0.11°, respectively. The results prove that the proposed method in this paper exhibits high computational efficiency and accuracy, providing an effective approach for digital assembly of mechanical parts.

Deep-Learning-Based Amplitude Variation with Angle Inversion with Multi-Input Neural Networks

Deep-learning-based (DL-based) seismic inversion has emerged as one of the state-of-the-art research areas in exploration geophysics with the development of artificial intelligence technology. Due to its good portability and high computational efficiency, this method has emerged as a data-driven approach for estimating subsurface properties. However, most of the current DL-based methods rely solely on seismic data, lacking the incorporation of prior information. In addition, these methods are usually performed trace-by-trace, resulting in insufficient horizontal constraints. These limitations make traditional methods less robust, particularly when dealing with high noise levels or limited data. To address these challenges, we propose a multi-input deep learning network for pre-stack inversion, which combines data-driven and model-driven approaches for optimization. The proposed method separately extracts features from the model and data, merging them to improve feature utilization. Moreover, by adopting a 2-D training unit, rather than a trace-by-trace approach, the method improves the horizontal continuity of the results. Tests on synthetic and real seismic data confirmed the robustness and improved stability of the proposed method, even under challenging conditions. This dual-driven approach significantly enhances the reliability of seismic inversion.

A Review of the Homogenized Lattice Boltzmann Method for Particulate Flow Simulations: From Fundamentals to Applications

The homogenized lattice Boltzmann method (HLBM) has emerged as a flexible computational framework for studying particulate flows, providing a monolithic approach to modeling pure fluid flows and flows through porous media, including moving solid and porous particles, within a unified framework. This paper presents a thorough review of HLBM, elucidating its underlying principles and highlighting its diverse applications to particle-laden flows in various fields as reported in literature. These include studies leading to new fundamental knowledge on the settling of single arbitrarily shaped particles as well as application-oriented research on wall-flow filters, hindered settling, and evaluation of the damage potential during particle transport. Among the strengths of HLBM are its monolithic approach, which allows seamless simulation of different fluid-solid interactions, and its ability to handle arbitrary particle shapes, including irregular and concave geometries, while resolving surface interactions to capture local forces. In addition, its parallel scheme based on the lattice Boltzmann method (LBM) results in high computational efficiency, making it suitable for large-scale simulations, even though LBM requires small time steps. Important future development needs are identified, including the addition of a lubrication force correction model, performance enhancements, such as support for hybrid parallelization and GPU, and the extension of compatible contact models to accommodate concave shapes. These advances promise expanded capabilities for HLBM and broader applicability for solving complex real-world problems.

Sequential bias-corrected weighted least squares solution of mixed additive and multiplicative error models

In the era of big data, the number of observations in adjustment calculations may reach tens or even hundreds of thousands. When dealing with these large dataset problems, many matrix operations are often required. At this time, the dimensions of the matrix will be large, which will generate a great computational burden. At present, no research results have been published on the computational efficiency of bias-corrected weighted least squares (bcWLS) for mixed additive and multiplicative error models (MAMEM). Sequential adjustment (SEA) groups the observations for calculation and can provide the same computational precision while greatly improving computational efficiency. This paper applies the idea of SEA to the calculation of bcWLS and proposes an iterative solution for sequential bcWLS (SEbcWLS). Using three simulation experiments to verify the effectiveness of our method, it was found that when the number of observations is 10000, the effect is better when the number of groups does not exceed 100, achieving the same precision as the original method while having high computational efficiency. The calculation results of line fitting and plane fitting are not affected by the number of grouping groups. For DEM (Digital elevation model) experiments with strong nonlinearity, when the number of grouping groups is too large, the effect is not very good, but the calculation efficiency is also higher than the original method, and the difference in calculation results is not significant.

H²CT: Asynchronous Distributed Key Generation With High-Computational Efficiency and Threshold Security in Blockchain Network

H²CT: Asynchronous Distributed Key Generation With High-Computational Efficiency and Threshold Security in Blockchain Network

A hybrid algorithm based on Bayesian optimization and Interior Point OPTimizer for optimal operation of energy conversion systems

Optimization methods are essential to improve the operation of energy conversion systems including energy storage equipment and fluctuating renewable energy. Modern systems consist of many components, operating in a wide range of conditions and governed by nonlinear balance equations. Consequently, identifying their optimal operation (e.g. minimizing operational costs) requires solving challenging optimization problems, with the global optimum often hidden behind many local ones. In this work, we propose a hybrid method that advantageously combines Bayesian optimization (BO) and Interior Point OPTimizer (IPOPT). The BO is a global approach exploiting Gaussian process regression to build a surrogate model of the cost function to be optimized, while IPOPT is a local approach using quasi-Newton updates. The proposed BO-IPOPT combination allows leveraging the parameter space exploration of the BO with the quasi-Newton convergence of IPOPT once solution candidates are in the neighborhood of an optimum. Using a challenging constrained test function, we test BO-IPOPT in accuracy and computational efficiency. Finally, we showcase the proposed method in the optimal operation of a renewable steam generation system. The results show that BO-IPOPT combines high accuracy and computational efficiency, achieving up to 50% better objective function values at the same CPU time than other state-of-the-art methods.

Cyclic shift short-distance attention for defect detection of fine etching mesh

Abstract Based on the deficiency of the existing Swin Transformer in the detection of surface defects in metal etched grids, i.e. the isolation of information transmission between different windows, and the limitations of image matrix size and network structure,which make it not suitable for long sequence data processing and dense prediction scenarios, a defect detection method for metal etching nets based on Cyclic shift Long Short Distance Attention (C-LSDA) is proposed. In the proposed method, cross-scale features are used to capture long and short distance relationships,and enhance the feature transfer between windows. Tracking position bias (TPB) is also used to improve relative position bias (RPB), in order to help the model to better understand the spatial relationships in the input data. In addition, the image is downsampled at various scales, so that the proposed method can be suitable for long sequence data processing and dense prediction scenarios. In order to improve the detection accuracy of small targets. a multi-feature fusion pyramid network (MFFPN) for feature fusion is also proposed. The experimental results show that the proposed method is significantly superior to the traditional CNN recognition method and the transformer recognition method. The proposed methodhas good identification accuracy and low computational cost.

A real-world battery state of charge prediction method based on a lightweight mixer architecture

This paper proposes a lightweight model based on the improved Mixer architecture—TrendFlow-Mixer, designed for efficient prediction of the state of charge (SOC) of lithium-ion batteries. SOC accuracy is crucial for the safe operation, usage, and maintenance of power batteries. When predicting the SOC of real-world batteries, the algorithm's generalization, robustness, and computational efficiency across different battery types must be considered. To this end, TrendFlow-Mixer introduces targeted improvements to battery characteristics, significantly alleviating the Mixer architecture's lag in SOC prediction. The model's core structure integrates the Fast Fourier Transform and arrangement transformation convolutional structures. This allows for efficient extraction of domain-specific spatial features and enables cross-channel correlation interactions. This reduces the model's parameter size while enhancing its adaptability to new domains. The model quickly captures data trend changes and identifies data flow patterns, improving computational efficiency by 70 % compared to Transformer-based architectures. TrendFlow-Mixer employs a dual-prediction head, combining rapid feature change analysis with a nonlinear prediction output to generate SOC predictions. This design facilitates fine-tuning of the model's structure and parameters while enhancing overall flexibility. Experimental results show that in a laboratory dynamic condition dataset, the model achieves a minimum mean absolute error (MAE) of 0.11 %, with the maximum prediction error under extreme temperatures controlled within 4.66 %. In validation using real-world data from the BMW i3, the mean absolute percentage error remains around 1 %, with a minimum MAE of 0.465 %. The results demonstrate that TrendFlow-Mixer possesses high computational efficiency, excellent generalization, and adaptability, effectively predicting SOC trends.

Optimal operation of three-phase unbalanced active distribution system based on semidefinite relaxation and convex-concave procedure

The three-phase distribution optimal power flow (D-OPF) problem has attracted much attention in recent years due to practical requirements for managing distributed energy resources and multi-energy customers in unbalanced active distribution systems. To deal with the strong nonconvexity of the three-phase OPF problem, this paper develops the semidefinite programming (SDP) formulation of three-phase OPF problems considering the mutual coupling in multi-phase networks. Instead of directly dropping the rank-one constraints, a sequential convex optimization algorithm based on the convex-concave procedure (CCP) is proposed for recovering a feasible and optimal solution of three-phase power flow. The proposed iterative algorithm not only has high computational efficiency but also guarantees the optimality when the semidefinite relaxation or second-order cone relaxation method is inexact. Case studies on various IEEE testing distribution systems verify that the proposed algorithm recovers the feasible and optimal power flow solution. It also exhibits a relatively low number of iterations and short solving time in large-scale IEEE testing systems.

A Minibatch Stochastic Gradient Descent-Based Learning Metapolicy for Inventory Systems with Myopic Optimal Policy

Stochastic gradient descent (SGD) has proven effective in solving many inventory control problems with demand learning. However, it often faces the pitfall of an infeasible target inventory level that is lower than the current inventory level. Several recent works have been successful in resolving this issue in various inventory systems. However, their techniques are rather sophisticated and difficult to apply to more complicated scenarios, such as multiproduct and multiconstraint inventory systems. In this paper, we address the infeasible target inventory-level issue from a new technical perspective; we propose a novel minibatch SGD-based metapolicy. Our metapolicy is flexible enough to be applied to a general inventory systems framework covering a wide range of inventory management problems with myopic clairvoyant optimal policy. By devising the optimal minibatch scheme, our metapolicy achieves a regret bound of [Formula: see text] for the general convex case and [Formula: see text] for the strongly convex case. To demonstrate the power and flexibility of our metapolicy, we apply it to three important inventory control problems, multiproduct and multiconstraint systems, multiechelon serial systems, and one-warehouse and multistore systems, by carefully designing application-specific subroutines. We also conduct extensive numerical experiments to demonstrate that our metapolicy enjoys competitive regret performance, high computational efficiency, and low variances among a wide range of applications. This paper was accepted by J. George Shanthikumar, data science. Funding: J. Xie and S. Yuan are supported by the National Natural Science Foundation of China (NSFC) [Grant 72331011]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/mnsc.2023.00920 .

Aerodynamic Analysis of Blade Stall Flutter Prediction for Transonic Compressor Using Energy Method

In this study, stall flutter onset prediction in a transonic compressor is carried out using the (uncoupled) energy method with Fourier transform. As the study is conducted computationally using computational fluid dynamics (CFD)-based simulations, the energy method was employed due to its higher computational efficiency by implementing the one-way FSI (Fluid Structure Interaction) model. The energy method is relatively uncommon for determining the aerodynamic damping and flutter prediction, specifically in blade stall conditions for the 3D blade passages. The NASA Rotor 67 was chosen for the validation of the study due to the availability of a wide range of experimental data. A flutter prediction analysis was performed computationally using CFD for the two-blade passages of the rotor in the peak efficiency and stall regions. Prior to this, the modal analysis on the prestressed blade was conducted, considering the centrifugal effects. The modal analysis provided accurate blade frequency and amplitude, which were the inputs of the flutter analysis. The first three modes of blade resonance were studied with a range of nodal diameters within near-peak efficiency and stall regions. The energy method implemented in this study for the flutter analysis was successfully able to predict the aerodynamic damping coefficients of the first three modes for a range of nodal diameters from the periodic-unsteady solution of the defined blade oscillation within the regions of interest (peak efficiency and stall point). The results of the study confirm the rotor blade’s stability within the near-peak region and, most importantly, the prediction of the flutter onset in the stall region. The study concluded that the computationally inexpensive and time-efficient energy method is capable of predicting the stall flutter onset. In the future, further validations of the energy method and investigations related to flow mechanism of stall flutter onset are planned.