Accuracy Degradation Research Articles

Accuracy degradation aware bit rate allocation for layer-wise uniform quantization of weights in neural network

Abstract Motivated by the fact that optimizing quantization error fails to minimize the accuracy degradation caused by the quantization of Neural Network (NN) weights, in this paper, we highlight the need for a comprehensive analysis of post-training quantization, covering average bit rate, accuracy degradation and SQNR. One such analysis, for the layer-wise uniform quantization and its application is NN weight compression is presented in this paper. We introduce an additional aspect of Uniform Quantizer (UQ) choice, and we allow the selection of one of two UQs (2-bit or 3-bit UQs) for each layer of NN. However, choosing a particular bit rate allocation to minimize average bit rate and accuracy degradation requires finding a trade-off solution. To solve the issue underpinning our motivation, for a post-training layer-wise uniform quantization framework with two UQs, we propose an algorithm for generating bit rate allocation patterns and corresponding pairs of average bit rate and accuracy degradation. We analyze the output of this algorithm and determine a solution of the elaborated two-objective optimization problem. To confirm the validity of the results we also apply the Pareto dominance method. This work can be considered a solid foundation for further analyses of more complex and deeper NNs.

Extending higher-order model for non-conservative perturbing forces acting on Galileo satellites during eclipse periods

For precise orbit determination (POD) and precise applications with POD products, one of the critical issues is the modeling of non-conservative forces acting on satellites. Since the official publication of Galileo satellite metadata in 2017, analytical models including the box-wing model and thermal thrust models have been established to absorb a substantial amount of solar radiation pressure (SRP) and thermal thrust. These models serve as the foundation for the best overall modeling approach, combining the analytical box-wing model and thermal thrust model with parameterization of the remaining non-conservative perturbing forces using various optimized Empirical CODE Orbit Models (ECOMs) of the Center for Orbit Determination in Europe (CODE). Firstly, we have demonstrated the significance of the second-order signals in the D direction and the first-order signals in the B direction through spectral analyses of the pure box-wing model, which are consistent with the currently recommended 7-parameter Empirical CODE Orbit Model 2 (ECOM2). In spite of this, we still found that degradation in orbit accuracy frequently occurs during deep eclipse seasons when using the ECOM2 model. We confirm a high-frequency signal existing in the fluctuating orbit overlap differences through the spectral analysis. Considering this, the ECOM2 force model should be extended to higher order and adapted to absorb the remaining effects of potential perturbing forces. After extending the ECOM2 force model to the sixth order in the Sun direction, we demonstrated the significance of fourth- and sixth-order sine terms for deep eclipses. Due to the higher-order periodic terms, the averaged RMS values of orbit overlap difference over deep eclipses can be reduced from 5.3, 10.8, and 23.8 cm to 3.2, 3.9, and 9.9 cm for in-orbit validation (IOV) satellites, from 5.0, 8.6, and 17.7 cm to 3.0, 3.0, and 7.1 cm for the first generation of full operational capability (FOC-1) satellites, and from 5.4, 8.6, and 19.0 cm to 3.6, 3.6, and 7.4 cm for the second generation of FOC (FOC-2) satellites, in the radial, cross-track, and along-track directions, respectively. Fluctuations with a peak amplitude of approximately 0.4 nm/s2 in the bias in the solar panel axis (Y) direction (Y-bias) are effectively mitigated by the higher-order terms. Due to the higher-order terms, the vertical positioning errors during kinematic precise point positioning (PPP) convergence can be improved from 42.3 to 37.1 cm at the 95.5% confidence level. Meanwhile, a low correlation level of up to 0.02 is found between the newly introduced higher-order parameters and earth rotation parameters (ERPs).

Advancing graph neural network architecture for fluid flow and heat transfer surrogate modeling: Variable boundary conditions and geometry

Graph neural networks (GNNs) represent a promising instrument for surrogate modeling, capable of handling unstructured computational meshes naturally. We address a typical issue of the accuracy degradation for larger computational domains due to the limited receptive field of GNN models and long-range global interactions between nodes of the mesh. We propose a modification of the GNN architecture allowing to improve the accuracy by a factor of 3 without significant increase in computational costs. The validation tests of the model concentrate on the two-dimensional stationary fluid flow around a bluff body in a channel and corresponding heat transfer. The problem formulation includes bluff bodies of randomly generated shapes and various boundary conditions. The model shows a robust performance for the out-of-domain data, i.e., the flow over an airfoil for different angles of attack.

An automatic eddy detection algorithm based on quadrant angle of velocity vector and its application

ABSTRACT Eddies assume a pivotal role in both the oceanic heat cycle and marine dynamic processes. The development of real-time satellite observations, coupled with advancements in computer intelligence, has propelled automatic eddy detection algorithms to the forefront of ocean remote sensing research. Nevertheless, target omissions and accuracy degradations are inevitable for multiple detection algorithms due to strong morphological variations in ocean eddies. This paper proposes an automatic detection algorithm based on the quadrant angle of the velocity vector in the flow field. Firstly, a rectangular search box is established, and the corresponding quadrant angles at the four vertices are calculated. Secondly, the centres and types of eddies are determined according to the cumulative sum and variety rules of quadrant angles. Then the outermost boundary is identified by regularity of quadrant angles in eight directions expanding outward from the centre of the eddy using the stream function equation. The new algorithm is assessed and verified using geostrophic flow data derived from the CMEMS standard gridded sea level anomaly product. Furthermore, its detection capabilities are demonstrated through a comparative analysis with several other algorithms. All methods exhibit consistent efficacy in detecting the majority of eddies with similar spatial distributions. In cases where the new algorithm identifies a specific eddy not detected by the FF15 and ND10 algorithms, sea surface temperature data is employed for verification. The sea surface temperature distribution map, along with the obtained results, illustrates the superiority of the new algorithm and its adaptability to products with varying resolutions. Additionally, the results undergo verification through a manual detection method, revealing that the new algorithm achieves SDR of 91.73% and an EDR of 3.54%. This percentage significantly surpasses the lower acceptable limit of 80% for the SDR parameter. The new algorithm is applied to study eddies with lifetimes exceeding ten days, spanning from March 2022 to February 2023, within the East China Sea and the South China Sea. The radius of SCS eddies is generally larger than that of ECS eddies, both cyclonic and anticyclonic. This further confirms the significant influence of regional ocean dynamics on the structure of eddies in different regions. Finally, the algorithm is applied to SWOT data to test the adaptability of the algorithm to non-regular grid data. The results show that the algorithm can effectively analyse SWOT data within 60° latitudes.

Research on Non-Intrusive Load Disaggregation Technology Based on VMD–Nyströmformer–BiTCN

Non-intrusive load disaggregation is a technique that monitors the total electrical load of an entire building or household. It uses a single power metering device to measure the total load. Then, it employs algorithms to break it down into the individual usage of different electrical devices. To address issues in load disaggregation models such as long training times, feature interference caused by the activation of other loads, and accuracy deficiencies caused by behavioral interference from users’ electricity usage habits, this paper proposes a VMD–Nyströmformer–BiTCN network architecture. The variational mode decomposition (VMD) filters the raw power data, reducing errors caused by noise and enhancing the accuracy of decomposing the load. A deep learning network utilizes a modified attention model, Nyströmformer, to reduce feature entanglement and accuracy degradation caused by habitual behavior interference during load disaggregation, while ensuring precise accuracy and improving network operational speed. The training network uses a bidirectional temporal convolutional network (BiTCN) and incorporates a residual network to expand the receptive field, allowing it to receive longer load sequence data and acquire more effective load information, thereby improving the disaggregation effectiveness for target appliances.

One-Dimensional Deep Residual Network with Aggregated Transformations for Internet of Things (IoT)-Enabled Human Activity Recognition in an Uncontrolled Environment

Human activity recognition (HAR) in real-world settings has gained significance due to the growth of Internet of Things (IoT) devices such as smartphones and smartwatches. Nonetheless, limitations such as fluctuating environmental conditions and intricate behavioral patterns have impacted the accuracy of the current procedures. This research introduces an innovative methodology employing a modified deep residual network, called 1D-ResNeXt, for IoT-enabled HAR in uncontrolled environments. We developed a comprehensive network that utilizes feature fusion and a multi-kernel block approach. The residual connections and the split–transform–merge technique mitigate the accuracy degradation and reduce the parameter number. We assessed our suggested model on three available datasets, mHealth, MotionSense, and Wild-SHARD, utilizing accuracy metrics, cross-entropy loss, and F1 score. The findings indicated substantial enhancements in proficiency in recognition, attaining 99.97% on mHealth, 98.77% on MotionSense, and 97.59% on Wild-SHARD, surpassing contemporary methodologies. Significantly, our model attained these outcomes with considerably fewer parameters (24,130–26,118) than other models, several of which exceeded 700,000 parameters. The 1D-ResNeXt model demonstrated outstanding effectiveness under various ambient circumstances, tackling a significant obstacle in practical HAR applications. The findings indicate that our modified deep residual network presents a viable approach for improving the dependability and usability of IoT-based HAR systems in dynamic, uncontrolled situations while preserving the computational effectiveness essential for IoT devices. The results significantly impact multiple sectors, including healthcare surveillance, intelligent residences, and customized assistive devices.

Novelty-aware concept drift detection for neural networks

Neural network models are widely adopted in real-world applications for processing streaming data. However, these applications often face challenges in terms of accuracy degradation, caused by changes in the data distribution of the stream data compared to the training data. Two underlying reasons contribute to these changes. The first, known as the concept drift problem, occurs when there is a change in the correlation between the input data and the prediction output, making the models trained on the training data no longer suitable for the new data. The second reason, known as the novelty problem, arises when real-world data contains unexpected data categories that were not present in the training data, resulting in incorrect predictions. The research community has divided into different groups and each developed various methods to detect either concept drift or novelty distribution changes. However, these methods only address one aspect of the problem and are unable to distinguish between them. This leads to an inappropriate allocation of model maintenance resources, including the high cost of model retraining and the acquisition of true label data. In this study, we aim to address this gap by proposing a novel concept drift detection method that is capable of distinguishing between known labeled concept drift and novelty. Our method is also more efficient than existing drift detection methods, making it suitable for applications on neural networks.

Optimizing Machine Learning Models with Data-level Approximate Computing: The Role of Diverse Sampling, Precision Scaling, Quantization and Feature Selection Strategies

Efficiency, low-power consumption, and real-time processing in embedded machine learning implementations are critical, particularly for models deployed in environments with large-scale data processing and resource-constrained environments. This paper investigates the application of approximate computing techniques as a viable solution to reduce computational complexity and optimize machine learning models, focusing on two widely used supervised machine learning models: k-nearest neighbors (KNN) and support vector machines (SVM). Although many studies compare machine learning classification techniques, the combined use of optimization strategies remains underexplored. Specifically, the combined utilization of feature selection, sampling, quantization, precision scaling, and relaxation methods for the purpose of optimizing and acquiring training and validation data is underexplored, particularly within the context of medical diagnosis datasets. In this paper, we propose a framework that uses data-level approximate computing techniques, including by diverse sampling strategies, precision scaling, quantization, and feature selection methods, to evaluate the impact of these techniques on the computational efficiency and accuracy of KNN and SVM models. Experimental results demonstrate that with careful application of approximate computing strategies, especially in critical applications such as medical diagnosis, it is possible to achieve considerable gains in efficiency while maintaining acceptable levels of accuracy. The combined application of these methods by selecting 3 features and quantizing the data values to 8 levels, then applying random sampling with 30% reductions and scaling the precision at 5 bits resulted in reductions of 87.5% in computation, 76.9% in memory usage, and 17% in delay, without any degradation in accuracy, as validated by tenfold cross-validation, Training Data validation, and full dataset validation. This study confirms the potential of approximate computing to optimize machine learning workflows, making it particularly suitable for applications with limited computational resources. The source code is publicly available online https://github.com/AyadMDalloo/DatalvlAxC.

End-to-End Deployment of Winograd-Based DNNs on Edge GPU

The Winograd algorithm reduces the computational complexity of convolutional neural networks (CNNs) by minimizing the number of multiplications required for convolutions, making it particularly suitable for resource-constrained edge devices. Concurrently, most edge hardware accelerators utilize 8-bit integer arithmetic to enhance energy efficiency and reduce inference latency, requiring the quantization of CNNs before deployment. Combining Winograd-based convolution with quantization offers the potential for both performance acceleration and reduced energy consumption. However, prior research has identified significant challenges in this combination, particularly due to numerical instability and substantial accuracy degradation caused by the transformations required in the Winograd domain, making the two techniques incompatible on edge hardware. In this work, we describe our latest training scheme, which addresses these challenges, enabling the successful integration of Winograd-accelerated convolution with low-precision quantization while maintaining high task-related accuracy. Our approach mitigates the numerical instability typically introduced during the transformation, ensuring compatibility between the two techniques. Additionally, we extend our work by presenting a custom-optimized CUDA implementation of quantized Winograd convolution for NVIDIA edge GPUs. This implementation takes full advantage of the proposed training scheme, achieving both high computational efficiency and accuracy, making it a compelling solution for edge-based AI applications. Our training approach enables significant MAC reduction with minimal impact on prediction quality. Furthermore, our hardware results demonstrate up to a 3.4× latency reduction for specific layers, and a 1.44× overall reduction in latency for the entire DeepLabV3 model, compared to the standard implementation.

3D-MSFC: A 3D multi-scale features compression method for object detection

As machine vision tasks rapidly evolve, a new concept of compression, namely video coding for machines (VCM), has emerged. However, current VCM methods are only suitable for 2D machine vision tasks. With the popularization of autonomous driving, the demand for 3D machine vision tasks has significantly increased, leading to an explosive growth in LiDAR data that requires efficient transmission. To address this need, we propose a machine vision-based point cloud coding paradigm inspired by VCM. Specifically, we introduce a 3D multi-scale features compression (3D-MSFC) method, tailored for 3D object detection. Experimental results demonstrate that 3D-MSFC achieves less than a 3% degradation in object detection accuracy at a compression ratio of 2796×. Furthermore, its low-profile variant, 3D-MSFC-L, achieves less than a 2% degradation in accuracy at a compression ratio of 463×. The above results indicate that our proposed method can provide an ultra-high compression ratio while ensuring no significant drop in accuracy, greatly reducing the amount of data required for transmission during each detection. This can significantly lower bandwidth consumption and save substantial costs in application scenarios such as smart cities.

Research on Fault Diagnosis of Agricultural IoT Sensors Based on Improved Dung Beetle Optimization–Support Vector Machine

The accuracy of data perception in Internet of Things (IoT) systems is fundamental to achieving scientific decision-making and intelligent control. Given the frequent occurrence of sensor failures in complex environments, a rapid and accurate fault diagnosis and handling mechanism is crucial for ensuring the stable operation of the system. Addressing the challenges of insufficient feature extraction and sparse sample data that lead to low fault diagnosis accuracy, this study explores the construction of a fault diagnosis model tailored for agricultural sensors, with the aim of accurately identifying and analyzing various sensor fault modes, including but not limited to bias, drift, accuracy degradation, and complete failure. This study proposes an improved dung beetle optimization–support vector machine (IDBO-SVM) diagnostic model, leveraging the optimization capabilities of the former to finely tune the parameters of the Support Vector Machine (SVM) to enhance fault recognition under conditions of limited sample data. Case analyses were conducted using temperature and humidity sensors in air and soil, with comprehensive performance comparisons made against mainstream algorithms such as the Backpropagation (BP) neural network, Sparrow Search Algorithm–Support Vector Machine (SSA-SVM), and Elman neural network. The results demonstrate that the proposed model achieved an average diagnostic accuracy of 94.91%, significantly outperforming other comparative models. This finding fully validates the model’s potential in enhancing the stability and reliability of control systems. The research results not only provide new ideas and methods for fault diagnosis in IoT systems but also lay a foundation for achieving more precise, efficient intelligent control and scientific decision-making.

Adversarial Circuit Rewriting against Graph Neural Network-based Operator Detection

Recent work has shown that graph neural networks (GNNs) can be used to recover high-level word operators and their boundaries in gate-level netlists. Unlike formal methods, however, the GNN does not prove functional equivalence. As such, there is a question of whether structural transforms like circuit rewrites that preserve the functionality of the circuit can diminish the performance of GNN-based operator detection. In this work, we explore this problem by performing simple rewrites on benchmark circuits and showing that indeed there is a performance degradation. We propose new features and learning strategies that can alleviate this. Finally, we develop a rewriting framework that aims to optimize GNN accuracy degradation, which translates to a fine-grained adversarial circuit rewriting of sorts. This can have applications in circuit obfuscation or further improving reverse engineering by providing synthetic training data. We show how this methodology beats non-adversarial rewriting under the same area overhead.

Energy‐Efficient Hardware Implementation of Spiking‐Restricted Boltzmann Machines Using Pseudo‐Synaptic Sampling

Stochastic sampling is performed to reduce hardware energy consumption and prevent overfitting by reducing parameters, because not all data are required for learning. In this study, a new approach, pseudo‐synaptic sampling (PS2) method, which approximates the conventional synaptic sampling machine (S2M) method through a hardware‐friendly implementation while demonstrating superior efficiency, is introduced. By sampling in front of neurons rather than at each synapse, the PS2 method improves hardware energy efficiency and ensures scalability. Furthermore, it improves energy and area efficiency by eliminating the additional circuit required by other techniques, such as the random walk (RW) method previously used which requires an additional circuit to frequently charge/discharge the membrane potential. Herein, the average firing rate equation for the S2M method is modified to suit the experimental conditions of this study. Through this numerical simulations, it is confirmed that the activation function of the PS2 method aligns with that of the S2M method and verified that the PS2 method can implement stochasticity for restricted Boltzmann machine (RBM) neurons. Experimental validation of the PS2 method, compared to the RW method, for Modified National Institute of Standards and Technology database (MNIST) training and inference on field‐programmable‐gate‐array‐implemented spiking RBM chips reveals promising results. In an MNIST 100‐handwritten digit experiment, the PS2 method exhibits on‐chip training accuracy (92%) comparable to that of the RW method (93%). Furthermore, in the energy consumption analysis, it is shown that the PS2 method reduces power consumption by 94.94% compared to the RW method, highlighting its enhanced power efficiency due to a reduced number of circuit elements. In the investigation into the impact of increasing the frequency at which random bits are generated, it is shown that the RW method experiences accuracy degradation even with slight increases, whereas the PS2 method maintains accuracy over significantly longer periods. This enables further power reduction by allowing for a longer period during random bit generation. In this study, a foundation is laid for maximizing the energy efficiency of spiking neural network processors by optimizing internal noise generation mechanisms.

Evaluation of reduced-order models for the rapid aerodynamic analysis of supersonic and hypersonic bodies

Abstract This paper presents a comprehensive evaluation of reduced-order models (ROMs) for the determination of pressure coefficient distributions on supersonic and hypersonic bodies. The study investigates the limitations, aerodynamic precision and computational performance associated with various methodologies, ranging from simplistic Newtonian theory-based approaches to more advanced first and second-order shock-expansion theories. Validation is performed by comparing computed results with experimental and computational data for pressure distributions, drag and lift coefficients and centres of pressure for fundamental geometries and authentic vehicle design over a wide range of freestream conditions. The study also includes a comprehensive computational complexity analysis, demonstrating the superiority of finite-element ROM approaches over traditional finite-volume computational fluid dynamics (CFD) simulations. The primary objective of this paper is to scrutinise the extension of these methodological classes to the low supersonic regime. Hence, thermo-chemical reactions within the flow are disregarded, and the ideal gas law is adopted. A value of $\gamma = 1.4$ is chosen for consistency and comparability across the analyses. The proposed ROMs show remarkable potential for reducing high-speed simulation execution times by four orders of magnitude, maintaining accuracy within 20 per cent and as low as 1 per cent. The study unveils three key findings: first, the accuracy degradation of Newtonian-based theories for inclined elements, particularly around 45 degrees, and their reduced dependency on Mach number at large inclination. Secondly, the study presents novel insights into the impact of shock-wave-Mach-wave interactions on pressure distribution calculations, emphasising the Mach number as a crucial metric governing recompression effects. Lastly, the study demonstrates the exceptional accuracy of DeJarnette’s method, providing ${C_P}$ results within 2 per cent for a wide range of conditions, offering an attractive alternative to the Taylor-Maccoll equation.

Edge Computing for AI-Based Brain MRI Applications: A Critical Evaluation of Real-Time Classification and Segmentation.

Medical imaging plays a pivotal role in diagnostic medicine with technologies like Magnetic Resonance Imagining (MRI), Computed Tomography (CT), Positron Emission Tomography (PET), and ultrasound scans being widely used to assist radiologists and medical experts in reaching concrete diagnosis. Given the recent massive uplift in the storage and processing capabilities of computers, and the publicly available big data, Artificial Intelligence (AI) has also started contributing to improving diagnostic radiology. Edge computing devices and handheld gadgets can serve as useful tools to process medical data in remote areas with limited network and computational resources. In this research, the capabilities of multiple platforms are evaluated for the real-time deployment of diagnostic tools. MRI classification and segmentation applications developed in previous studies are used for testing the performance using different hardware and software configurations. Cost-benefit analysis is carried out using a workstation with a NVIDIA Graphics Processing Unit (GPU), Jetson Xavier NX, Raspberry Pi 4B, and Android phone, using MATLAB, Python, and Android Studio. The mean computational times for the classification app on the PC, Jetson Xavier NX, and Raspberry Pi are 1.2074, 3.7627, and 3.4747 s, respectively. On the low-cost Android phone, this time is observed to be 0.1068 s using the Dynamic Range Quantized TFLite version of the baseline model, with slight degradation in accuracy. For the segmentation app, the times are 1.8241, 5.2641, 6.2162, and 3.2023 s, respectively, when using JPEG inputs. The Jetson Xavier NX and Android phone stand out as the best platforms due to their compact size, fast inference times, and affordability.

Intelligent Fault Diagnosis Method Based on Neural Network Compression for Rolling Bearings

Rolling bearings are often exposed to high speeds and pressures, leading to the symmetry in their rotating structure being disrupted, which can lead to serious failures. Intelligent rolling bearing fault diagnosis is a critical part of ensuring operation of machinery, and it has been facilitated by the growing popularity of convolutional neural networks (CNNs). The outstanding performance of fault diagnosis CNNs results from complex and redundant network structures and parameters, resulting in huge storage and computational requirements, which makes it challenging to implement these models in resource-limited industrial devices. This study aims to address this problem by proposing a comprehensive compression method for CNNs that is applied to intelligent fault diagnosis. It involves several different compression methods, including tensor train decomposition, parameter quantization, and knowledge distillation for deep network compression. This results in a significant decrease in redundancy and speeding up the training of CNN models. Firstly, tensor train decomposition is applied to reduce redundant connections in both convolutional and fully connected layers. The next step is to perform parameter quantization to minimize the bits needed for parameter representation and storage. Finally, knowledge distillation is used to restore accuracy to the compressed model. The effectiveness of the proposed approach is confirmed by an experiment and ablation study with different models on several datasets. The results show that it can significantly reduce redundant information and floating-point operations with little degradation in accuracy. Notably, on the CWRU dataset, with about 60% parameter reduction, there is no degradation in our model’s accuracy. The proposed approach is a new attempt at the intelligent fault diagnosis of rolling bearings in industrial equipment.

Quantifying the Impacts of Fire‐Related Perturbations in WRF‐Hydro Terrestrial Water Budget Simulations in California's Feather River Basin

ABSTRACTWildfire activity in the western United States (WUS) is increasingly impacting water supply, and land surface models (LSMs) that do not explicitly account for fire disturbances can have critical uncertainties in burned areas. This study quantified responses from the Weather Research and Forecasting Hydrological modelling system (WRF‐Hydro) to a suite of fire‐related perturbations to hydrologic soil and runoff parameters, vegetation area, land cover classifications and associated vegetation properties, and snow albedo across the heavily burned Feather River Basin in California. These experiments were used to quantify the impacts of fire‐related perturbations in model simulations under the observed meteorological conditions during the 2000–2022 water years and determine whether applying these fire‐related perturbations enhanced post‐fire model accuracy across the 11–12 post‐fire months evaluated herein. The most comprehensive fire‐aware simulation consistently modelled enhanced annual catchment streamflow (by 8%–37%), subsurface flow (by 72%–116%), and soil moisture (by 4%–9%), relative to the baseline simulation which neglected fire impacts. Simulated fire‐enhanced streamflow was predominately attributable to fire‐induced vegetation area reductions that reduced transpiration. Simulated streamflow enhancements occurred throughout the water year, excluding early‐summer (e.g., May–June) when the baseline simulation modelled relatively more snowmelt and streamflow because fire perturbations caused earlier model snow depletion. Vegetation area reductions favoured increased model ground snow accumulation and enhanced snow ablation while imposed snow albedo darkening enhanced ablation, ultimately resulting in similar peak SWE and earlier snow disappearance (on average by 8‐days) from the most comprehensive fire‐aware simulation relative to the baseline simulation. The baseline simulation had large degradations in streamflow accuracy following major fire events that were likely partially attributable to neglecting fire disturbances. Applying fire‐related perturbations reduced post‐fire streamflow anomaly biases across the three study catchments. However, remaining large post‐fire streamflow uncertainties in the fire‐perturbed simulation underscores the importance of additional observationally constrained fire‐disturbance model developments.

LCGSC-YOLO: a lightweight apple leaf diseases detection method based on LCNet and GSConv module under YOLO framework.

In response to the current mainstream deep learning detection methods with a large number of learned parameters and the complexity of apple leaf disease scenarios, the paper proposes a lightweight method and names it LCGSC-YOLO. This method is based on the LCNet(A Lightweight CPU Convolutional Neural Network) and GSConv(Group Shuffle Convolution) module modified YOLO(You Only Look Once) framework. Firstly, the lightweight LCNet is utilized to reconstruct the backbone network, with the purpose of reducing the number of parameters and computations of the model. Secondly, the GSConv module and the VOVGSCSP (Slim-neck by GSConv) module are introduced in the neck network, which makes it possible to minimize the number of model parameters and computations while guaranteeing the fusion capability among the different feature layers. Finally, coordinate attention is embedded in the tail of the backbone and after each VOVGSCSP module to improve the problem of detection accuracy degradation issue caused by model lightweighting. The experimental results show the LCGSC-YOLO can achieve an excellent detection performance with mean average precision of 95.5% and detection speed of 53 frames per second (FPS) on the mixed datasets of Plant Pathology 2021 (FGVC8) and AppleLeaf9. The number of parameters and Floating Point Operations (FLOPs) of the LCGSC-YOLO are much less thanother related comparative experimental algorithms.

LDF-BNN: A Real-Time and High-Accuracy Binary Neural Network Accelerator Based on the Improved BNext.

Significant progress has been made in industrial defect detection due to the powerful feature extraction capabilities of deep neural networks (DNNs). However, the high computational cost and memory requirement of DNNs pose a great challenge to the deployment of industrial edge-side devices. Although traditional binary neural networks (BNNs) have the advantages of small storage space requirements, high parallel computing capability, and low power consumption, the problem of significant accuracy degradation cannot be ignored. To tackle these challenges, this paper constructs a BNN with layered data fusion mechanism (LDF-BNN) based on BNext. By introducing the above mechanism, it strives to minimize the bandwidth pressure while reducing the loss of accuracy. Furthermore, we have designed an efficient hardware accelerator architecture based on this mechanism, enhancing the performance of high-accuracy BNN models with complex network structures. Additionally, the introduction of multi-storage parallelism alleviates the limitations imposed by the internal transfer rate, thus improving the overall computational efficiency. The experimental results show that our proposed LDF-BNN outperforms other methods in the comprehensive comparison, achieving a high accuracy of 72.23%, an image processing rate of 72.6 frames per second (FPS), and 1826 giga operations per second (GOPs) on the ImageNet dataset. Meanwhile, LDF-BNN can also be well applied to defect detection dataset Mixed WM-38, achieving a high accuracy of 98.70%.

Analysis of the degradation in transmission accuracy of RV reducers considering the wear clearance between cycloidal wheels and pinwheels

To accurately determine the transmission accuracy degradation pattern of Rotate Vector (RV) reducers due to wear, a predictive method is proposed that simultaneously considers the initial clearance and wear clearance between the cycloid wheel and the pinwheel. The initial clearance is derived from a comprehensive gear system analysis, considering assembly errors, manufacturing errors and modifications. Building on this, the wear depth of the cycloid wheel tooth profile is numerically simulated using Archard's wear theory, revealing the wear distribution pattern under various operating conditions. A Gaussian regression model is used to predict the maximum wear amount for different operational durations. Based on the predicted wear amount, a multi-body dynamic simulation model incorporating wear clearance was established to analyze the transmission error under varying clearance conditions. Results indicate that the wear depth along the tooth profile initially increases and then decreases. The depth and region of wear along the tooth profile gradually increase with load and output speed, with the impact of load on the wear region being greater than that of output speed. Under the initial clearance condition, the maximum transmission error is 22.91″. As the wear clearance increases, transmission accuracy declines sharply. When the cumulative wear clearance reaches 0.2 mm, the transmission error exceeds the allowable backlash limit. This study provides a theoretical reference for predicting accuracy degradation and guides the manufacturing design of RV reducers.