Large Parameter Research Articles

High Throughput Characterization of Organic Thin Film Transistors.

Automation is vital to accelerating research. In recent years,the application of self-driving labs to materials discovery and device optimization has highlighted many benefits and challenges inherent to these new technologies. Successful automated workflows offer tangible benefits to fundamental science and industrial scale-up by significantly increasing productivity and reproducibility all while enabling entirely new types of experiments. However, it's implemtation is often time-consuming and cost-prohibitive and necessitates establishing multidisciplinary teams that bring together domain-specific knowledge with specific skillsets in computer science and engineering. This perspective article provides a comprehensive overview of how the research group has adopted "hybrid automation" over the last 8 years by using simple automatic electrical testers (autotesters) as a tool to increase productivity and enhance reproducibility in organic thin film transistor (OTFT) research. From wearable and stretchable electronics to next-generation sensors and displays, OTFTs have the potential to be a key technology that will enable new applications from health to aerospace. The combination of materials chemistry, device manufacturing, thin film characterization and electrical engineering makes OTFT research challenging due to the large parameter space created by both diverse material roles and device architectures. Consequently, this research stands to benefit enormously from automation. By leveraging the multidisciplinary team and taking a user-centered design approach in the design and continued improvement of the autotesters, the group has meaningfully increased productivity, explored research avenues impossible with traditional workflows, and developed as scientists and engineers capable of effectively designing and leveraging automation to build the future of their fields to encourage this approach, the files for replicating the infrastructure are included, and questions and potential collaborations are welcomed.

Lightweight pyramid attention residual network for intelligent fault diagnosis of machine under sharp speed variation

Deep learning based diagnostic methods have obtained great success in intelligent fault diagnosis of machines. However, most of the existing methods are developed for fault diagnosis task under stable working conditions, and cannot handle the data generated under sharp speed variation favorably. Besides, the large model parameters and high computing costs of these methods fail to meet the requirement of small memory and fast-precise prediction in practical applications. To address these challenges, a lightweight pyramid attention residual network (PARNet) is proposed in this paper. First, an efficient pyramid squeeze convolution module is proposed to capture multi-scale information from raw signals. Then, a soft squeeze-and-excitation attention is designed to establish long-range channel dependency, and fuse features of different scales. Based on the above improvements, a pyramid attention residual block is proposed as basic feature learning unit, and a concise CNN-based model PARNet is built using PAR block. Extensive experiments in two case studies are conducted to validate the effectiveness and adaptability of PARNet under different speed variation conditions. The test results demonstrate the superiority of the proposed method for machine fault diagnosis and reliability assessment in terms of recognition accuracy and model parameter complexity, compared with other advanced diagnostic methods.

AsymUNet: An Efficient Multi-Layer Perceptron Model Based on Asymmetric U-Net for Medical Image Noise Removal

With the continuous advancement of deep learning technology, U-Net–based algorithms for image denoising play a crucial role in medical image processing. However, most U-Net-based medical image denoising algorithms typically have large parameter sizes, which poses significant limitations in practical applications where computational resources are limited or large-scale patient data processing are required. In this paper, we propose a medical image denoising algorithm called AsymUNet, developed using an asymmetric U-Net framework and a spatially rearranged multilayer perceptron (MLP). AsymUNet utilizes an asymmetric U-Net to reduce the computational burden, while a multiscale feature fusion module enhances the feature interaction between the encoder and decoder. To better preserve the image details, spatially rearranged MLP blocks serve as the core building blocks of AsymUNet. These blocks effectively extract both the local and global features of the image, reducing the model’s reliance on prior knowledge of the image and further accelerating the training and inference processes. Experimental results demonstrate that AsymUNet achieves superior performance metrics and visual results compared with other state-of-the-art methods.

Electrodeposition of Non-Noble High-Entropy Alloys for Effective Hydrogen Evolution Electrocatalysts

As society works towards reducing its carbon footprint, various sectors such as energy, steelmaking and fertilizer production have begun to look increasingly towards decarbonization. Hydrogen production through electrolysis, when linked to renewables (solar, wind), provides a promising low-emission alternative as an energy carrier and chemical feedstock. With the decreasing cost of said renewables, greater focus is now on the capital cost of the electrolyser, and particularly the membrane/electrode structures. Conventional Proton Exchange Membrane (PEM) electrolysers rely on the use of electrocatalysts made with expensive noble metals such as platinum, ruthenium, and palladium [1]. Anion Exchange Membrane (AEM) electrolysers present an effective means of alkaline water splitting with low carbon emissions and can effectively utilize non-noble metal electrocatalysts for the Hydrogen Evolution Reaction (HER). A novel class of materials, High-Entropy Alloys (HEAs), made of non-noble metals or with reduced noble metal content have shown great promise as active electrocatalytic materials [2]. HEAs are typically defined as metal alloys made of five or more constituent elements each comprising at least 5% of the total material [3]. Having such a large number of principal alloying elements can lead to enhanced phase stability due to an increased configurational entropy, have unique chemical clustering, and can help prevent degradation of the material in harsh environments [2]. Additionally, the large lattice parameter mismatch between constituent elements can result in a high degree of strain which can help shift electronic states in a way that is favourable to hydrogen electrocatalysis [4]. These properties along with the so called “cocktail effect,” whereby combining multiple elements can result in unexpected synergistic interactions, makes HEAs uniquely suited to electrocatalytic applications [5].In this work, we investigate the electrocatalytic performance of electrodeposited FeNiCoMoW HEAs. The goal is to combine three hyper-d transition metals with two hypo-d transition metals to alter the electronic structure and improve electrocatalytic performance. Using chronoamperometric measurements, the Tafel slopes for FeNiCoMoW HEAs electrodeposited at pH 5, 6 and 7 were determined, with the pH 5 HEA demonstrating the smallest average Tafel slope of approximately 83 mV/dec. Compared to some electrodeposited binary alloys reported on in the literature, this marks an improvement and may be the result of unexpected interactions within the compositionally complex alloy. In the FeNiCoMoW HEA, the magnitude of the Tafel slope increased in tandem with the electrochemically active surface area (ECSA) as determined by Cyclic Voltammetry (CV) which suggests that the composition may be more important for improving performance. As the pH of the electrodeposition solution increased, the amount of cobalt in the as-deposited HEA decreased while the amounts of molybdenum and tungsten remained constant. The composition of nickel and iron appeared to vary inversely, reaching a maximum and minimum respectively at pH 6. From these results it seems that maximizing cobalt content and minimizing nickel content could potentially help to improve HER performance in the FeNiCoMoW system.

(Invited) Computational Framework and Insights Towards Electrolyte Design for Redox Flow Batteries

Achieving a fully renewable energy future requires the development of grid-scale energy storage technologies, and non-aqueous redox flow batteries (NRFBs) hold considerable potential.1 However, the redox-active materials (RAMs) currently in consideration as lead candidates for NRFBs exhibit shortcomings, including low solubility, impaired transport properties, and low reduction potentials. Utilizing theory and computation is invaluable in predicting the structure-property relationship of RAM as NRFB active materials. The numerous components within NRFB electrolytes, spanning solvents, active materials, and supporting salts, suggest that material discovery could experience a substantial acceleration through computational and machine-learning approaches.This presentation will provide a progress update on our efforts towards a computational framework involving quantum mechanical (QM), molecular dynamics (MD), and machine learning methods to predict fundamental chemical and physicochemical properties of redox-active materials for energy storage systems. We used an extremely stable, bio-inspired vanadium(4+) bis-hydroxyiminodiacetic acid (VBH) as the scaffold to explore the large parameter space of NRFB systems.2–4 First, we show our framework based on density functional theory (DFT) for calculating solubilities, incorporating both lattice and solvation energies, that can be calibrated and used with synthetic verifications. We found that lattice free energy, which has previously been neglected, has a significant role in tuning electrolyte solubility. Second, we describe our MD simulations on the concentration and composition dependence of transport properties, such as viscosity, using DFT-derived force field bonded parameters for the novel [VBH]2- anion to understand the interactions between the ions and the solvent. Third, we demonstrate that QM molecular design could finely adjust the reduction potential. These fundamental investigations would offer insights into electrolyte design and how molecular properties dictate the macroscopic properties that contribute to specific performance aspects of advanced battery electrolytes.1 Y. Huang, S. Gu, Y. Yan and S. F. Y. Li, Current Opinion in Chemical Engineering, 2015, 8, 105–113.2 H. Huang, R. Howland, E. Agar, M. Nourani, J. A. Golen and P. J. Cappillino, J. Mater. Chem. A, 2017, 5, 11586–11591.3 S. K. Pahari, T. C. Gokoglan, B. R. B. Visayas, J. Woehl, J. A. Golen, R. Howland, M. L. Mayes, E. Agar and P. J. Cappillino, RSC Adv., 2021, 11, 5432–5443.4 B. R. B. Visayas, S. K. Pahari, T. C. Gokoglan, J. A. Golen, E. Agar, P. J. Cappillino and M. L. Mayes, Chem. Sci., 2021, 12, 15892–15907.

Scanning Electron Microscopy Energy-Dispersive X-Ray Spectroscopy Analysis of Catalyst-Coated Porous Transport Layers

The transition of the United States energy infrastructure towards hydrogen energy involves the optimization of proton exchange membrane water electrolyzers (PEMWEs) for reliable hydrogen generation. PEMWE device typically consists of titanium porous transport layer (PTL) and an iridium-based anode catalyst layer. Several directions are being pursued to improve performance and to reduce the costs. To mitigate naturally occurring titanium passivation, protective coatings, typically platinum or other platinum-group metals (PGMs) are commonly applied to the PTLs. The drive to decrease the cost and the use of scarce materials has led to the development of catalyst layers with lower catalyst loadings. Reducing fabrication time and cost as well as increasing efficiency of the anode has led to the investigation of direct application of the iridium-based catalyst onto the Pt-coated Ti PTL. This motivated our investigation of porous-transport electrodes (PTEs) prepared with a variety of deposition methods, using different ink formulations, and targeting different catalyst loadings. There is need to develop a characterization approach that can be used to assess and compare quality of coatings across large parameter space or to monitor quality during production.Several sets of PTEs samples consisting of a titanium porous layer, platinum protective coating and iridium-based catalyst layer were prepared using a wide range of catalyst layer coating methods and parameters, including spray coating, gravure, airbrush and electrodeposition. All samples were characterized using scanning electron microscopy and energy-dispersive X-ray spectroscopy (SEM-EDS) to analyze the elemental composition and assess the distribution of elements of interest (Ti, Pt, and Ir). First, the atomic% of each element was used to calculate elemental ratios for quantitative assessment of the amount of deposited catalyst. However, it is crucial to recognize the limitations of relying solely on this information for assessing catalyst coating since it doesn’t account for heterogeneity of the coatings. Although elemental maps provide valuable insights into spatial distribution, they alone do not offer a complete picture. To better understand the catalyst coating distribution, each individual EDS elemental map was used to calculate the surface area, which served as a measure of PTL area covered by the catalyst. Further, surface area ratios (Pt/Ir) were calculated to determine variations in surface coverage. This poster will highlight differences between fabricated samples and will demonstrate that together, at% and surface area ratio values allow for a comprehensive quantitative assessment of the composition and distribution of catalyst surfaces that were produced through a wide range of fabrication variables. The implementation of proper characterization of these PTEs can help aid future decisions on which parameters should be used moving forward. This approach can also be used for quality control, enabling quick screening of commercially produced samples.

SwiftWater: A lightweight SwiftFormer-based underwater image enhancement network with auxiliary prompt module

Due to variations in light attenuation underwater and the complexities of light propagation in aquatic environments, underwater images commonly suffer from degradation, such as color distortion, blur, and low contrast. These challenges pose significant obstacles to researchers studying marine resources. In recent years, numerous solutions have been proposed for underwater image enhancement (UIE) tasks, with deep learning-based methods, including those employing convolution and transformer structures, demonstrating promising performance. However, deep learning approaches often encounter issues related to large parameter sizes and substantial computational demands. Addressing these concerns, this paper introduces SwiftWater, a lightweight transformer-based network specifically designed for UIE tasks, featuring a U-Net style architecture. Leveraging the efficient SwiftFormer encoder as the network backbone, we introduce a Hybrid Squeeze and Excitation Block (HSEB) to dynamically adjust representations within skip connections. Additionally, we introduce an Auxiliary Prompt Module (APM) with a Lightweight Prompt Block (LPB) that utilizes the HSV color space of underwater images to enhance training. LPB within APM comprises two modules: the Prompt Generation Module (PGM) for prompt generation and the Lightweight Prompt Interaction Module (LPIM) for prompt-feature interaction, generating rich auxiliary image representations. Compared to existing mainstream UIE methods employing transformer structures, our proposed approach offers reduced parameter counts while achieving state-of-the-art (SOTA) performance across both quantitative and qualitative metrics.

Artificial chemotaxis under electrodiffusiophoresis

HypothesisThrough a large parameter space, electric fields can tune colloidal interactions and forces leading to diverse static and dynamical structures. So far, however, field-driven interactions have been limited to dipole-dipole and hydrodynamic contributions. Nonetheless, in this work, we propose that under the right conditions, electric fields can also induce interactions based on local chemical fields and diffusiophoretic flows. ExperimentsHerein, we present a strategy to generate and measure 3D chemical gradients under electric fields. In this approach, faradaic reactions at electrodes induce global pH gradients that drive long-range transport through electrodiffusiophoresis. Simultaneously, the electric field induces local pH gradients by driving the particle's double layer far from equilibrium. FindingsAs a result, while global pH gradients lead to 2D focusing away from electrodes, local pH gradients induce aggregation in the third dimension. Evidence points to a mechanism of interaction based on diffusiophoresis. Interparticle interactions display a strong dependence on surface chemistry, zeta potential and diameter of particles. Furthermore, pH gradients can be readily tuned by adjusting the voltage and frequency of the electric field. For large Péclet numbers, we observed a collective chemotactic-like collapse of particles. Remarkably, such collapse occurs without reactions at a particle's surface. By mixing particles with different sizes, we also demonstrate, through experiments and Brownian dynamics simulations, the emergence of non-reciprocal interactions, where small particles are more drawn towards large ones.

Investigation of Machine Learning Regression Techniques to Predict Critical Heat Flux Over a Large Parameter Space

A unifying and accurate model to predict critical heat flux (CHF) over a wide range of conditions has been elusive since wall boiling research emerged. With the release of the experimental data utilized in the development of the 2006 Groeneveld CHF lookup table (LUT), by far the most extensive public CHF database available to date (nearly 25 000 data points), the development of data-driven prediction models over a large parameter space in simple geometry (vertical, uniformly heated round tubes) can be achieved. Furthermore, the popularization of machine learning techniques to solve regression problems has led to more advanced tools for analyzing large and complex databases. This work compares three machine learning algorithms to predict the CHF database. For each selected regression algorithm (ν-Support vector, Gaussian process, and neural network), a set of optimized hyperparameters is applied. Among the investigated algorithms, the neural network emerges as the most effective, achieving a CHF predicted/measured factor standard deviation of 12.3%, three times less than that of the LUT. In comparison, the Gaussian process regression and the ν-support vector regression achieve a standard deviation of 17.7%, about two times lower than the LUT. Hence, all considered algorithms significantly outperform the LUT prediction performance. The neural network model and training methodology are designed to prevent overfitting, which is confirmed by data analyses of the CHF predictions. Finally, future development directions (including data coverage, transfer learning, and uncertainty quantification) are discussed.

Determining the spin of light primordial black holes with Hawking radiation. Part II. High spin regime

We propose a method to determine the mass and spin of primordial black holes based on measuring the energy and emission rate at the dipolar and quadrupolar peaks in the primary photon Hawking spectrum, applicable for dimensionless spin parameters a~\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\overset{\\sim }{a} $$\\end{document} ≳ 0.6. In particular, we show that the ratio between the energies of the two peaks is only a function of the black hole spin, while the ratio between their emission rates depends also on the line-of-sight inclination. The black hole mass and distance from the Earth may then be inferred from the absolute values of the peak energies and emission rates. This method is relevant for primordial black holes born with large spin parameters that are presently still in the early stages of their evaporation process. While Hawking radiation from such black holes is somewhat below the reach of planned gamma-ray telescopes, we expect the required sensitivity to be attained in the not too distant future.

On a quenching problem of hyperbolic MEMS equation with a nonlocal term

We consider a nonlocal hyperbolic problem. It arises in the field of MEMS control. First, we review the stationary problem and state that there exists a unique and no solution for small and large parameter, respectively. Next, we introduce the results of global existence and dynamical properties of solution for the hyperbolic equation. Finally, we derive the result of the quenching behaviour. As proven in similar problems, we show that the quenching occurs for large parameter. This criterion is also expressed by means of the first eigenvalue of Laplace operator.

Asymptotic Representations of Solutions of $$n\times n$$ Systems of Ordinary Differential Equations with a Large Parameter

Asymptotic Representations of Solutions of $$n\times n$$ Systems of Ordinary Differential Equations with a Large Parameter

Analytical proxy to families of numerical solutions: the case study of spherical mini-boson stars

The Einstein field equations, or generalizations thereof, are difficult to solve analytically. On the other hand, numerical solutions of the same equations have become increasingly common, in particular concerning compact objects. Whereas analytic approximations to each individual solution within a numerical family have been proposed, proxies for whole families are missing, which can facilitate studying properties across the parameter space, data compression and a wider usage of such solutions. In this work we tackle this need, proposing a simple strategy based on two differentexpansions of the unknown functions in an appropriately chosen basis, to build such proxy. We use as an exploratory case-study spherical, fundamental mini-boson stars, to illustrate the feasibility of such an approach, emphasise its advantage in reducing the data size, and the challenges, say, in covering large parameter spaces.

A Lightweight CER-YOLOv5s Algorithm for Detection of Construction Vehicles at Power Transmission Lines

In the context of power-line scenarios characterized by complex backgrounds and diverse scales and shapes of targets, and addressing issues such as large model parameter sizes, insufficient feature extraction, and the susceptibility to missing small targets in engineering-vehicle detection tasks, a lightweight detection algorithm termed CER-YOLOv5s is firstly proposed. The C3 module was restructured by embedding a lightweight Ghost bottleneck structure and convolutional attention module, enhancing the model’s ability to extract key features while reducing computational costs. Secondly, an E-BiFPN feature pyramid network is proposed, utilizing channel attention mechanisms to effectively suppress background noise and enhance the model’s focus on important regions. Bidirectional connections were introduced to optimize the feature fusion paths, improving the efficiency of multi-scale feature fusion. At the same time, in the feature fusion part, an ERM (enhanced receptive module) was added to expand the receptive field of shallow feature maps through multiple convolution repetitions, enhancing the global information perception capability in relation to small targets. Lastly, a Soft-DIoU-NMS suppression algorithm is proposed to improve the candidate box selection mechanism, addressing the issue of suboptimal detection of occluded targets. The experimental results indicated that compared with the baseline YOLOv5s algorithm, the improved algorithm reduced parameters and computations by 27.8% and 31.9%, respectively. The mean average precision (mAP) increased by 2.9%, reaching 98.3%. This improvement surpasses recent mainstream algorithms and suggests stronger robustness across various scenarios. The algorithm meets the lightweight requirements for embedded devices in power-line scenarios.

Dynamic Analysis of a Uniform Microbeam Resting on a Nonlinear Foundation Considering Its Curvature Subjected to a Mechanical Impact and Electromagnetic Actuation.

This study proposes an investigation into the nonlinear vibration of a simply supported, flexible, uniform microbeam associated with its curvature considering the mechanical impact, the electromagnetic actuation, the nonlinear Winkler-Pasternak foundation, and the longitudinal magnetic field. The governing differential equations and the boundary conditions are modeled within the framework of a Euler-Bernoulli beam considering an element of the length of the beam at rest and using the second-order approximation of the deflected beam and the Galerkin-Bubnov procedure. In this work, we present a novel characterization of the microbeam and a novel method to solve the nonlinear vibration of the microactuator. The resulting equation of this complex problem is studied using the Optimal Homotopy Asymptotic Method, employing some auxiliary functions derived from the terms that appear in the equation of motion. An explicit closed-form analytical solution is proposed, proving that our procedure is a powerful tool for solving a nonlinear problem without the presence of small or large parameters. The presence of some convergence-control parameters assures the rapid convergence of the solutions. These parameters are evaluated using some rigorous mathematical procedures. The present approach is very accurate and easy to implement, even for complicated nonlinear problems. The local stability near the primary resonance is studied.

Self-Supervised Contrastive Learning for Automated Segmentation of Brain Tumor MRI Images in Schizophrenia

Schizophrenic patients’ brain tumor magnetic resonance imaging (MRI) images are important references for doctors to diagnose and treat schizophrenia. However, automatic segmentation of these images is a professional and tedious task. Existing methods suffer from problems such as large model parameters, long computation time, and inadequate image processing. To achieve more accurate segmentation of brain tumors, we propose brain tumor MRI images for automatic segmentation using self-supervised contrastive learning in schizophrenia patients (BTCSSSP). First, a denoising algorithm based on progressive principal component analysis approximation and adaptive clustering is designed to process the noisy MRI images. Second, a brightness-aware image enhancement algorithm is developed to address the problems of non-uniformity, unclear boundaries, and poor spatial resolution of the MRI images. Finally, a cross-scale U-Net network with selective feature fusion attention module is designed based on self-supervised contrastive learning to achieve automatic segmentation of brain tumor MRI images. The results show that the BTCSSSP method yields higher Recall and Precision than existing methods. The maximum recall is 95%, and the image segmentation precision is 95%, thus indicating good practical applicability.

Vertical line time domain green function and its applications in numerical simulation of ship seakeeping performance

In present study, a vertical line source time domain Green function is proposed and the subsequent numerical evaluation methods are described in detail. For which, Taylor expansion and Chebyshev approximation algorithms are originally extended from point source to line source for small parameters, as well as asymptotic expansions are extended for large parameters. Meanwhile, polynomials are originally proposed to replace the asymptotic expansions to improve the computational efficiency. The accuracy and stability of the proposed method are validated through several cases and good agreement can be obtained.Furthermore, a hybrid Green function method based on the vertical line source Green function and Rankine source is developed for time domain simulation of seakeeping performances of ships. With respect to the hybrid method, the flow domain is divided into inner part and outer part by an artificial control surface. The first-order Taylor Expansion Boundary Element Method (TEBEM) based on the Rankine panel distribution is applied in the inner part, while three-dimensional panel method based on the vertical line source Green function distribution is adopted in the outer part. Under this circumstance, motions of the single-ship (Slender Wigley and Blunt Wigley) and two-ship (two side-by-side ships) in head waves are numerically investigated. The numerical results agree well with the experimental data, which proved the feasibility and the accuracy of present hybrid Green function method for motion prediction of both single-ship and two-ship.

An unrolled neural network for accelerated dynamic MRI based on second-order half-quadratic splitting model

The reconstruction of dynamic magnetic resonance images from incomplete k-space data has sparked significant research interest due to its potential to reduce scan time. However, traditional iterative optimization algorithms fail to faithfully reconstruct images at higher acceleration factors and incur long reconstruction time. Furthermore, end-to-end deep learning-based reconstruction algorithms suffer from large model parameters and lack robustness in the reconstruction results. Recently, unrolled deep learning models, have shown immense potential in algorithm stability and applicability flexibility. In this paper, we propose an unrolled deep learning network based on a second-order Half-Quadratic Splitting(HQS) algorithm, where the forward propagation process of this framework strictly follows the computational flow of the HQS algorithm. In particular, we propose a degradation-sense module by associating random sampling patterns with intermediate variables to guide the iterative process. We introduce the Information Fusion Transformer(IFT) to extract both local and non-local prior information from image sequences, thereby removing aliasing artifacts resulting from random undersampling. Finally, we impose low-rank constraints within the HQS algorithm to further enhance the reconstruction results. The experiments demonstrate that each component module of our proposed model contributes to the improvement of the reconstruction task. Our proposed method achieves comparably satisfying performance to the state-of-the-art methods and it exhibits excellent generalization capabilities across different sampling masks. At the low acceleration factor, there is a 0.7% enhancement in the PSNR. Furthermore, when the acceleration factor reached 8 and 12, the PSNR achieves an improvement of 3.4% and 5.8% respectively.

YOLO-Granada: a lightweight attentioned Yolo for pomegranates fruit detection

Pomegranate is an important fruit crop that is usually managed manually through experience. Intelligent management systems for pomegranate orchards can improve yields and address labor shortages. Fast and accurate detection of pomegranates is one of the key technologies of this management system, crucial for yield and scientific management. Currently, most solutions use deep learning to achieve pomegranate detection, but deep learning is not effective in detecting small targets and large parameters, and the computation speed is slow; therefore, there is room for improving the pomegranate detection task. Based on the improved You Only Look Once version 5 (YOLOv5) algorithm, a lightweight pomegranate growth period detection algorithm YOLO-Granada is proposed. A lightweight ShuffleNetv2 network is used as the backbone to extract pomegranate features. Using grouped convolution reduces the computational effort of ordinary convolution, and using channel shuffle increases the interaction between different channels. In addition, the attention mechanism can help the neural network suppress less significant features in the channels or space, and the Convolutional Block Attention Module attention mechanism can improve the effect of attention and optimize the object detection accuracy by using the contribution factor of weights. The average accuracy of the improved network reaches 0.922. It is only less than 1% lower than the original YOLOv5s model (0.929) but brings a speed increase and a compression of the model size. and the detection speed is 17.3% faster than the original network. The parameters, floating-point operations, and model size of this network are compressed to 54.7%, 51.3%, and 56.3% of the original network, respectively. In addition, the algorithm detects 8.66 images per second, achieving real-time results. In this study, the Nihui convolutional neural network framework was further utilized to develop an Android-based application for real-time pomegranate detection. The method provides a more accurate and lightweight solution for intelligent management devices in pomegranate orchards, which can provide a reference for the design of neural networks in agricultural applications.

Spatiotemporal factor models for functional data with application to population map forecast

The proliferation of mobile devices has led to the collection of large amounts of population data. This situation has prompted the need to utilize this rich, multidimensional data in practical applications. In response to this trend, we have integrated functional data analysis (FDA) and factor analysis to address the challenge of predicting hourly population changes across various districts in Tokyo. Specifically, by assuming a Gaussian process, we avoided the large covariance matrix parameters of the multivariate normal distribution. In addition, the data were both time and spatially dependent between districts. To capture various characteristics, a Bayesian factor model was introduced, which modeled the time series of a small number of common factors and expressed the spatial structure through factor loading matrices. Furthermore, the factor loading matrices were made identifiable and sparse to ensure the interpretability of the model. We also proposed a Bayesian shrinkage method as a systematic approach for factor selection. Through numerical experiments and data analysis, we investigated the predictive accuracy and interpretability of our proposed method. We concluded that the flexibility of the method allows for the incorporation of additional time series features, thereby improving its accuracy.