Large Parameter Research Articles

Deep learning-based lightweight radar target detection method

For target detection tasks in complicated backgrounds, a deep learning-based radar target detection method is suggested to address the problems of a high false alarm rate and the difficulties of achieving high-performance detection by conventional methods. Considering the issues of large parameter count and memory occupation of the deep learning-based target detection models, a lightweight target detection method based on improved YOLOv4-tiny is proposed. The technique applies depthwise separable convolution (DSC) and bottleneck architecture (BA) to the YOLOv4-tiny network. Moreover, it introduces the convolutional block attention module (CBAM) in the improved feature fusion network. It allows the network to be lightweight while ensuring detection accuracy. We choose a certain number of pulses from the pulse-compressed radar data for clutter suppression and Doppler processing to obtain range–Doppler (R–D) images. Experiments are run on the R–D two-dimensional echo images, and the results demonstrate that the proposed method can quickly and accurately detect dim radar targets against complicated backgrounds. Compared with other algorithms, our approach is more balanced regarding detection accuracy, model size, and detection speed.

Discrete Carleman estimates and application to controllability for a fully-discrete parabolic operator with dynamic boundary conditions

We consider a fully-discrete approximation of a 1-D heat equation with dynamic boundary conditions for which we provide a controllability result. The proof of this result is based on a relaxed observability inequality for the corresponding adjoint system. This is conducted using a suitable Carleman estimate for such models where the discrete parameters h and Δt are connected to one of the large Carleman parameters.

Abstracts from The Aerosol Society Drug Delivery to the Lungs 33.

Abstracts from The Aerosol Society Drug Delivery to the Lungs 33.

Rapid estimation of cortical neuron activation thresholds by transcranial magnetic stimulation using convolutional neural networks

BackgroundTranscranial magnetic stimulation (TMS) can modulate neural activity by evoking action potentials in cortical neurons. TMS neural activation can be predicted by coupling subject-specific head models of the TMS-induced electric field (E-field) to populations of biophysically realistic neuron models; however, the significant computational cost associated with these models limits their utility and eventual translation to clinically relevant applications. ObjectiveTo develop computationally efficient estimators of the activation thresholds of multi-compartmental cortical neuron models in response to TMS-induced E-field distributions. MethodsMulti-scale models combining anatomically accurate finite element method (FEM) simulations of the TMS E-field with layer-specific representations of cortical neurons were used to generate a large dataset of activation thresholds. 3D convolutional neural networks (CNNs) were trained on these data to predict thresholds of model neurons given their local E-field distribution. The CNN estimator was compared to an approach using the uniform E-field approximation to estimate thresholds in the non-uniform TMS-induced E-field. ResultsThe 3D CNNs estimated thresholds with mean absolute percent error (MAPE) on the test dataset below 2.5% and strong correlation between the CNN predicted and actual thresholds for all cell types (R2 > 0.96). The CNNs estimated thresholds with a 2–4 orders of magnitude reduction in the computational cost of the multi-compartmental neuron models. The CNNs were also trained to predict the median threshold of populations of neurons, speeding up computation further. Conclusion3D CNNs can estimate rapidly and accurately the TMS activation thresholds of biophysically realistic neuron models using sparse samples of the local E-field, enabling simulating responses of large neuron populations or parameter space exploration on a personal computer.

Simulating the diversity of shapes of the Lyman-α line

ABSTRACT The Ly α line is a powerful probe of distant galaxies, which contains information about inflowing/outflowing gas through which Ly α photons scatter. To develop our understanding of this probe, we post-process a zoom-in radiation-hydrodynamics simulation of a low-mass (Mstar ∼ 109 M⊙) galaxy to construct 22 500 mock spectra in 300 directions from z = 3 to 4. Remarkably, we show that one galaxy can reproduce the variety of a large sample of spectroscopically observed Ly α line profiles. While most mock spectra exhibit double-peak profiles with a dominant red peak, their shapes cover a large parameter space in terms of peak velocities, peak separation, and flux ratio. This diversity originates from radiative transfer effects at interstellar medium and circum-galactic medium (CGM) scales, and depends on galaxy inclination and evolutionary phase. Red-dominated lines preferentially arise in face-on directions during post-starburst outflows and are bright. Conversely, accretion phases usually yield symmetric double peaks in the edge-on direction and are fainter. While resonant scattering effects at <0.2 × Rvir are responsible for the broadening and velocity shift of the red peak, the extended CGM acts as a screen and impacts the observed peak separation. The ability of simulations to reproduce observed Ly α profiles and link their properties with galaxy physical parameters offers new perspectives to use Ly α to constrain the mechanisms that regulate galaxy formation and evolution. Notably, our study implies that deeper Ly α surveys may unveil a new population of blue-dominated lines tracing inflowing gas.

Penalized Estimation of Sparse Markov Regime-Switching Vector Auto-Regressive Models

We consider sparse Markov regime-switching vector autoregressive (MSVAR) models in which the regimes are governed by a latent homogeneous Markov chain. In practice, even for moderate values of the number of Markovian regimes and data dimension, the associated MSVAR model has a large parameter dimension compared to a typical sample size. We provide a unified penalized conditional likelihood approach for estimating sparse MSVAR models. We show that our proposed estimators are consistent and recover the sparse structure of the model. We also show that, when the number of regimes is correctly or over-specified, our method provides consistent estimation of the predictive density. We develop an efficient implementation of the method based on a modified Expectation-Maximization (EM) algorithm. We discuss strategies for estimation of the number of regimes. We evaluate finite-sample performance of the method via simulations, and further demonstrate its utility by analyzing a real dataset.

Studies on a novel Ho3+:KBGM crystal toward the ∼2 μm laser applications

A Ho3+:KBGM crystal with good quality and large size was successfully grown via TSSG method. The crystal’s structure, ambient-temperature polarized absorption and emission spectra and fluorescence lifetime were studied. The J-O theory was adopted to analyze the absorption spectra and results show that crystal exhibits relatively large intensity parameters Ω2 and spectroscopic quality factor which is up to 3.12. At 453 nm, the absorption band is the strongest, and the corresponding cross-sections are 21.94 × 10–20 cm2, 26.66 × 10–20 cm2 and 18.56 × 10–20 cm2 for the -X, -Y and -Z polarizations, respectively. The emission spectra also demonstrate the broad and strong emission band at 2030 nm. The FWHMs and cross-sections are 74 nm, 0.37 × 10−20 cm2, 50 nm, 0.68 × 10−20 cm2, and 49 nm 0.52 × 10−20 cm2 for the three polarizations, respectively. In addition, the crystal has a superior quantum efficiency which is as high as 97.1%. Taking all the analyses into account, it is reasonable to assert that the Ho3+:KBGM crystal an ideal candidate for ∼2 µm laser applications in many areas.

The new shift perturbation method with applications to vibrational problems

Employing variable horizontal and vertical shifts for the unperturbed solution, a new perturbation method is proposed. The new method is named as the shift perturbation method. The existing perturbation methods are interpreted with respect to the horizontal and vertical shifts for the unperturbed solution. The new method is applied to some well-known vibrational problems. The method is capable of producing admissible approximate solutions for a wide range of problems such as the free and forced vibrations of the Duffing equation, equations with quadratic and cubic nonlinearities, variable amplitude nonlinear problems, and boundary layer problems. A variant of the method also produced valid solutions for large perturbation parameters.

Asymmetric dark matter with a spontaneously broken U(1)′ : Self-interaction and gravitational waves

Motivated by the collisionless cold dark matter small scale structure problem, we propose an asymmetric dark matter model where dark matter particle interact with each other via a massive dark gauge boson. This model easily avoid the strong limits from cosmic microwave background (CMB) observation, and have a large parameter space to be consistent with small scale structure data. We focus on a special scenario where portals between dark sector and visible sector are too weak to be detected by traditional methods. We find that this scenario can increase the effective number of neutrinos ($N_{\text{eff}}$).In addition, the spontaneous $U(1)'$ symmetry breaking process, which makes dark gauge boson massive, can generate stochastic gravitational waves with peak frequency around $10^{-6} - 10^{-7} \text{ Hz}$.

MSIA-Net: A Lightweight Infrared Target Detection Network with Efficient Information Fusion.

In order to solve the problems of infrared target detection (i.e., the large models and numerous parameters), a lightweight detection network, MSIA-Net, is proposed. Firstly, a feature extraction module named MSIA, which is based on asymmetric convolution, is proposed, and it can greatly reduce the number of parameters and improve the detection performance by reusing information. In addition, we propose a down-sampling module named DPP to reduce the information loss caused by pooling down-sampling. Finally, we propose a feature fusion structure named LIR-FPN that can shorten the information transmission path and effectively reduce the noise in the process of feature fusion. In order to improve the ability of the network to focus on the target, we introduce coordinate attention (CA) into the LIR-FPN; this integrates the location information of the target into the channel so as to obtain more expressive feature information. Finally, a comparative experiment with other SOTA methods was completed on the FLIR on-board infrared image dataset, which proved the powerful detection performance of MSIA-Net.

Power System Stabilizer Design using Genetic Algorithms and Particle Swarm Optimization

in this paper, Meta-heuristic techniques using genetic algorithms GA and particle swarm optimization PSO to tuning optimal design of power system stabilizer PSS proposed. This latter have been used for many years to add damping to electromechanical oscillations of power system, Based on this idea we have proposed multiobjective function composed with tow function, first maximize stability margin by increasing the damping factors while minimizing the real parts of the eigenvalues. Simulation results to comparative study between genetic algorithms and particle swarm optimization obtained by our realized graphical user interface (GUI) proved the efficiency of PSS optimized by genetic algorithms in comparison with Particle Swarm Optimization, showing stable system responses almost insensitive to large parameter variations and under different operating regime (under-excited, nominal and over excited regime).

A machine learning approach to drawing phase diagrams of topological lasing modes

Identifying phases and analyzing the stability of dynamic states are ubiquitous and important problems which appear in various physical systems. Nonetheless, drawing a phase diagram in high-dimensional and large parameter spaces has remained challenging. Here, we propose a data-driven method to derive the phase diagram of lasing modes in topological insulator lasers. The classification is based on the temporal behaviour of the topological modes obtained via numerical integration of the rate equation. A semi-supervised learning method is used and an adaptive library is constructed in order to distinguish the different topological modes present in the generated parameter space. The proposed method successfully distinguishes the different topological phases in the Su-Schrieffer-Heeger lattice with saturable gain. This demonstrates the possibility of classifying the topological phases without needing for expert knowledge of the system and may give valuable insight into the fundamental physics of topological insulator lasers via reverse engineering of the derived phase diagram.

Lightweight Multiscale CNN Model for Wheat Disease Detection

Wheat disease detection is crucial for disease diagnosis, pesticide application optimization, disease control, and wheat yield and quality improvement. However, the detection of wheat diseases is difficult due to their various types. Detecting wheat diseases in complex fields is also challenging. Traditional models are difficult to apply to mobile devices because they have large parameters, and high computation and resource requirements. To address these issues, this paper combines the residual module and the inception module to construct a lightweight multiscale CNN model, which introduces the CBAM and ECA modules into the residual block, enhances the model’s attention to diseases, and reduces the influence of complex backgrounds on disease recognition. The proposed method has an accuracy rate of 98.7% on the test dataset, which is higher than classic convolutional neural networks such as AlexNet, VGG16, and InceptionresnetV2 and lightweight models such as MobileNetV3 and EfficientNetb0. The proposed model has superior performance and can be applied to mobile terminals to quickly identify wheat diseases.

Gamma-rays buildup factor calculation using regression and Extra-Trees

The statistical fitting of existing data is an efficient method for calculating the gamma-ray buildup factors. However, the existing Geometric Progression (GP) fitting formula requires the interpolation of five parameters: a, b, c, d, and Xk. GP fitting formula also considers three factors: photon energy, shielding material and shielding in the interpolation process, leading to the complex calculations of buildup factor. On the other hand, the large and bloated parameter library also brings inconvenience to the application of GP formula in radiation calculation code. This paper proposes a regression method for calculating gamma-ray buildup factors based on Extra-Trees (ET). The training set is obtained from the exposure buildup factors of the ANS standard database to establish the ET with superior performance for buildup factor. The ET model can reproduce the data of exposure buildup factor in ANS without deviation. For materials not considered in ANS, the buildup factors calculated by ET model are consistent with the results calculated by GP fitting formula. The performance in time consumption is also satisfactory. The method proposed in this paper has the potential to replace GP fitting formula and be used in situations where the buildup factor needs to be calculated, such as design, shielding calculation and analysis of gamma-rays.

Optimization of a Twistable Hovering Flapping Wing Inspired by Giant Hummingbirds Using the Unsteady Blade Element Theory

Due to the complexity of tailoring the wing flexibility and selecting favorable kinematics, the design of flapping wings is a considerably challenging problem. Therefore, there is an urgent need to investigate methods that can be used to design wings with high energy efficiency. In this study, an optimization model was developed to improve energy efficiency by optimizing wing geometric and kinematic parameters. Then, surrogate optimization was used to solve the design optimization model. Finally, the optimal design parameters and the associated sensitivity were provided. The optimized flapping wing, inspired by hummingbirds, features large geometrical parameters, a moderate amplitude of the flapping angle, and low frequency. With the spanwise twisting deformation considered in the parameterization model, the optimization solver gave an optimized wing with a pitching amplitude of approximately 39 deg at the root and 76 deg at the tip. According to the sensitivity analysis, the length of the wing, flapping frequency, and flapping amplitude are the three critical parameters that determine both force generation and power consumption. The amplitude of the pitching motion at the wing root contributes to lowering power consumption. These results provide some guidance for the optimal design of flapping wings.

Secular Light Curves of main-belt comets 133P/Elst-Pizarro, 176P/LINEAR, 288P/(300163) and 433P/(248370)

Among the main-belt comets (MBCs) that have been discovered, 133P/Elst-Pizarro, 176P/LINEAR, 288P/(300163) and 433P/(248370) do not have obvious coma, and the amplitudes of their Secular Light Curves (SLCs) are less than 3 mag. Through the SLCs, we can better understand the activity status and decay process of these four low-level active MBCs, and then understand the activity evolution of the MBCs. To get the SLCs of these four MBCs, we collected 2,962 observations (span 5–6 apparitions) and used improved methods (subtract the opposition and aspect angle effects). According to the SLCs and related parameters, we get the following conclusions: (1) 133P/Elst-Pizarro and 176P/LINEAR had activity decay characteristics. For 133P/Elst-Pizarro, the amplitude of SLC was 2.4 mag in 1996+2001 perihelion passages, but changed to 1.2 mag in 2007+2013 perihelion passages. The time TOFF and distance ROFF at the turn off point of activity changed significantly, from TOFF=388±20 d and ROFF=3.04±0.02 au (1996+2001) to TOFF=230±13 d and ROFF=2.81±0.02 au (2007+2013). For 176P/LINEAR, the amplitude of SLC during 2005 perihelion passage was 0.37 mag, while during 2011, 2017 and 2022 perihelion passages, it did not show activity. (2) 133P/Elst-Pizarro and 176P/LINEAR had the aspect angle effect. That could mean the nuclei of 133P/Elst-Pizarro and 176P/LINEAR have large obliquity and axis ratio. (3) The equinox of 133P/Elst-Pizarro was close to perihelion (Δt=45d, anomaly=12°) and 176P reaches the solstice near perihelion (Δt=69d, anomaly=18°). (4) 288P/(300163) had a large asymmetry parameter ROFF/RON, which means that the sublimation takes place in depth, not just in the surface. (5) It can also be seen from SLC that 433P/(248370) had been active since 2010 perihelion passage, which was earlier than previously reported in 2016. But, SLC of 433P did not show activity near the perihelion in 2005. 433P had photometric anomaly (∼ 0.5 mag), from Δt = 94 to 138 d or R = 2.425 to 2.481 au in 2021 perihelion passage. (6) The PAGE and TAGE of four comets were determined.

On adjustable undular bore profiles based on the modified steady KdV–Burgers equation

In this study, a speed parameter is introduced into the steady Korteweg–de Vries (KdV)–Burgers equation which enables the theoretical undular bore profiles to be adjustable with a proper combination of the speed parameter and the viscous damping parameter. A new criterion for identifying the above two bores is then proposed with respect to these two parameters, whose influence on the undular bore profile is then discussed. For the theoretical solution with a small damping, error after introducing the variable speed parameter is limited. A large speed parameter corresponds to a wide range of acceptable dampings. From the energy perspective, it is confirmed that the speed parameter also denotes the nonlinearity effect. In addition, comparison between the theoretical and experimental results shows the superiority of the present model over the traditional model, which also reveals the physical meanings of the present model.

The cosmic web of X-ray active galactic nuclei seen through the eROSITA Final Equatorial Depth Survey (eFEDS)

Which galaxies in the general population turn into active galactic nuclei (AGNs) is a keystone of galaxy formation and evolution. Thanks to SRG/eROSITA’s contiguous 140 square degree pilot survey field, we constructed a large, complete, and unbiased soft X-ray flux-limited (FX > 6.5 × 10−15 erg s−1 cm−2) AGN sample at low redshift, 0.05 < z < 0.55. Two summary statistics, the clustering using spectra from SDSS-V and galaxy-galaxy lensing with imaging from HSC, are measured and interpreted with halo occupation distribution and abundance matching models. Both models successfully account for the observations. We obtain an exceptionally complete view of the AGN halo occupation distribution. The population of AGNs is broadly distributed among halos with a mean mass of 3.9−2.4+2.0 × 1012 M⊙. This corresponds to a large-scale halo bias of b(z = 0.34) = 0.99−0.10+0.08. The central occupation has a large transition parameter, σlog10(M) = 1.28 ± 0.2. The satellite occupation distribution is characterized by a shallow slope, αsat = 0.73 ± 0.38. We find that AGNs in satellites are rare, with fsat < 20%. Most soft X-ray-selected AGNs are hosted by central galaxies in their dark matter halo. A weak correlation between soft X-ray luminosity and large-scale halo bias is confirmed (3.3σ). We discuss the implications of environmental-dependent AGN triggering. This study paves the way toward fully charting, in the coming decade, the coevolution of X-ray AGNs, their host galaxies, and dark matter halos by combining eROSITA with SDSS-V, 4MOST, DESI, LSST, and Euclid data.

Transfer Deep Learning Network for Rolling Bearing Fault Diagnosis of Wind Turbines

Fault diagnosis of rolling bearings has become a critical measure to ensure the security, efficiency, and availability of wind turbine systems. In this work, an approach called a transfer deep learning network is reported to resolve the drawbacks of existing rolling bearing fault algorithms on the basis of deep learning, such as large training parameters, long training time, and insufficient training samples. The reported approach mainly consists of two steps: feature transfer using transfer component analysis and network transfer using the pre-trained convolutional neural network. One dataset of rolling bearings with four fault types is gathered to estimate the classification performance, and the peer model is employed for a comparison study. The derived research results show that the reported algorithm based on the feature transfer and the network transfer can yield outstanding results in the rolling bearing fault diagnosis.

Estimating ECG intervals from a deep learning morphology model

The statistical fitting of existing data is an efficient method for calculating the gamma-ray buildup factors. However, the existing Geometric Progression (GP) fitting formula requires the interpolation of five parameters: a, b, c, d, and Xk. GP fitting formula also considers three factors: photon energy, shielding material and shielding in the interpolation process, leading to the complex calculations of buildup factor. On the other hand, the large and bloated parameter library also brings inconvenience to the application of GP formula in radiation calculation code. This paper proposes a regression method for calculating gamma-ray buildup factors based on Extra-Trees (ET). The training set is obtained from the exposure buildup factors of the ANS standard database to establish the ET with superior performance for buildup factor. The ET model can reproduce the data of exposure buildup factor in ANS without deviation. For materials not considered in ANS, the buildup factors calculated by ET model are consistent with the results calculated by GP fitting formula. The performance in time consumption is also satisfactory. The method proposed in this paper has the potential to replace GP fitting formula and be used in situations where the buildup factor needs to be calculated, such as design, shielding calculation and analysis of gamma-rays.