Gaussian Parameters Research Articles

Modeling and prediction of set‑up errors in breast cancer image‑guided radiotherapy using the Gaussian mixture model.

The aim of the present study was to develop a prediction model for set-up error distribution in breast cancer image-guided radiotherapy (IGRT) using a Gaussian mixture model (GMM). To achieve this, the image-guided set-up errors data of 80 patients with breast cancer were selected, and the GMM was used to develop the set-up errors distribution prediction model. The predicted error center points, covariance and probability were calculated and compared with the planning target volume (PTV) margin formula. A total of 1,200 sets of set-up errors in IGRT for breast cancer were collected. The results of the Gaussian model parameters showed that the set-up errors were mainly in the direction of µ1-µ4 center points. All the raw errors in the lateral, longitudinal and vertical directions were -6.30-4.60, -5.40-1.47 and -2.70-1.70 mm, respectively. According to the probability of each center, the set-up error was most likely to shift in the µ1 direction, reaching 0.53. The set-up errors of the other three centers, µ2, µ3 and µ4, were 0.11, 0.34 and 0.12, respectively. According to the covariance parameters of the GMM, the maximum statistical standard deviation of the set-up errors reached 29.06. In conclusion, the results of the present study demonstrated that the GMM can be used to quantitatively describe and predict the distribution of set-up errors in IGRT for breast cancer, and these findings could be useful as a reference for set-up error control and tumor PTV expansion in breast cancer radiotherapy without routine, daily IGRT.

Realization and optimization of the gold-mesoporous silica based bimodal surface plasmon resonance sensor with digital Gaussian filter

Abstract This work presented a novel biomdal surface plasmon resonance (BSPR) sensor with mesoporous silica film (MSF) and digital Gaussian filter. A modified Stöber solution growth approach was applied to prepare MSF film on the gold film. The conventional SPR sensor was coupled with a digital Gaussian filter to realize the BSPR sensor. After that, the Gaussian parameters were modified and the BSPR angular spectrum was obtained for further sensing experiments. Porosity and thickness of the MSF film were measured and calculated by means of a combination of simulation and measurement. The simulation results indicate that the refractive index (RI) sensitivity of the BSPR sensor could reach 75.11 deg/RIU which is 66.91% higher than the 45 deg/RIU of the conventional gold-MSF SPR sensor. Based on the experiments, the RI resolution of BSPR sensor was improved by 37.1% to 6.61 × 10−6 RIU, the limit of detection (LOD) for glucose was raised from 320 mg l−1 to 131 mg l−1, and the LOD for CTAB molecule was raised from 124.98 nM to 63.78 nM when compared to the gold-MSF SPR sensor.

Unraveling gluon TMDs in J/ψ and pion production at the EIC

We investigate the azimuthal asymmetries such as cos2ϕT and Sivers symmetry for J/ψ and π± production in electron-proton scattering, focusing on scenarios where the J/ψ and the pion are produced in an almost back-to-back configuration. The electron is unpolarized, while the proton can be unpolarized or transversely polarized. For the J/ψ formation, we use nonrelativistic QCD (NRQCD), while π± is formed due to parton fragmentation. In this kinematics, we utilize the transverse momentum-dependent factorization framework to calculate the cross sections and asymmetries. We consider both quark and gluon-initiated processes and show that the gluon contribution dominates. In this work, we used the generalized parton model (GPM) for the transverse momentum distribution (TMD) parametrizations and did not consider the effect of TMD evolution. We provide numerical estimates of the upper bounds on the azimuthal asymmetries, as well as employ a Gaussian parametrization for the gluon TMDs, within the kinematical region accessible by the upcoming Electron-Ion Collider (EIC). Published by the American Physical Society 2024

Impact of the Interruption Duration on Photoluminescence Properties of MOCVD-Grown GaAsP/InAlGaAs Quantum Well Structures.

The growth interruption technology is introduced to the growth of GaAsP/InAlGaAs quantum well (QW) structure using metal-organic chemical vapor deposition (MOCVD). The effect of growth interruption time (GIT) on the crystalline quality and optical properties are investigated. The two distinctive emission peaks are the transition recombination between the electron level of conduction band and the light and heavy hole level of valence band in the photoluminescence (PL) at room temperature. The PL peaks present a redshift and merge together with the increasing GIT, which is attributed to the QW energy level shift caused by the increase in arsenic concentrations in GaAsP QW, the diversified thickness of QW and the variations of indium components in the InAlGaAs barrier layer. The Gaussian deconvolution parameters in temperature-dependent PL (TDPL) show that the GaAsP/InAlGaAs QW with a GIT of 6 s has a 565.74 meV activation energy, enhancing the carrier confinement in QW and the PL intensity, while the 6 s-GIT GaAsP QW has the increasing interface roughness and the non-radiative centers at the InGaAsP intermediate layer, leading to a spectral broadening. The QW with 10 s-GIT exhibits a small full width at half maximum (FWHM) with the various temperature, indicating reduced interface roughness and excellent crystal quality. An increase in GIT may be suitable for optimizing the optical properties of GaAsP/InAlGaAs QW.

Bayesian modelling of multiple plasma diagnostics at Wendelstein 7-X

Abstract Inference of electron density and temperature has been performed using multiple, diverse sets of plasma diagnostic data at Wendelstein 7-X. Predictive models for the interferometer, Thomson scattering and helium beam emission spectroscopy (He-BES) systems have been developed within the Minerva framework and integrated into a unified model. Electron density and temperature profiles are modelled using Gaussian processes. Calibration factors for the Thomson scattering system and predictive uncertainties are considered as additional unknown parameters. The joint posterior probability distribution for the electron density and temperature profiles as well as Gaussian process hyperparameters and model parameters is explored through a Markov chain Monte Carlo algorithm. Samples from this distribution are numerically marginalised over the hyperparameters and model parameters to yield marginal posterior distributions for the electron density and temperature profiles. The profile inferences incorporate various data combinations from the interferometer and Thomson scattering as well as constraints at the limiter/divertor positions through virtual observations or edge data from He-BES. Additionally, the integration of x-ray imaging crystal spectrometer data into the model for ion temperature profiles is presented. All profiles presented in this study are inferred with optimally selected hyperparameters and model parameters by exploring the joint posterior distribution, inherently applying Bayesian Occam’s razor.

Monte Carlo dosimetric analyses on the use of 90Y-IsoPet intratumoral therapy in canine subjects

Objective. To investigate different dosimetric aspects of 90Y-IsoPet™ intratumoral therapy in canine soft tissue sarcomas, model the spatial spread of the gel post-injection, evaluate absorbed dose to clinical target volumes, and assess dose distributions and treatment efficacy. Approach. Six canine cases treated with 90Y-IsoPet™ for soft tissue sarcoma at the Veterinary Health Center, University of Missouri are analyzed in this retrospective study. The dogs received intratumoral IsoPet™ injections, following a grid pattern to achieve a near-uniform dose distribution in the clinical target volume. Two dosimetry methods were performed retrospectively using the Monte Carlo toolkit OpenTOPAS: imaging-based dosimetry obtained from post-injection PET/CT scans, and stylized phantom-based dosimetry modeled from the planned injection points to the gross tumor volume. For the latter, a Gaussian parameter with variable sigma was introduced to reflect the spatial spread of IsoPet™. The two methods were compared using dose-volume histograms (DVHs) and dose homogeneity, allowing an approximation of the closest sigma for the spatial spread of the gel post-injection. In addition, we compared Monte Carlo-based dosimetry with voxel S-value (VSV)-based dosimetry to investigate the dosimetric differences. Main results. Imaging-based dosimetry showed differences between Monte Carlo and VSV calculations in tumor high-density areas with higher self-absorption. Stylized phantom-based dosimetry indicated a more homogeneous target dose with increasing sigma. The sigma approximation of the 90Y-IsoPet™ post-injection gel spread resulted in a median sigma of approximately 0.44 mm across all cases to reproduce the dose heterogeneity observed in Monte Carlo calculations. Significance. The results indicate that dose modeling based on planned injection points can serve as a first-order approximation for the delivered dose in 90Y-IsoPet™ therapy for canine soft tissue sarcomas. The dosimetry evaluation highlights the non-uniformity of absorbed doses despite the gel spread, emphasizing the importance of considering tumor dose heterogeneity in treatment evaluation. Our findings suggest that using Monte Carlo for dose calculation seems more suitable for this type of tumor where high-density areas might play an important role in dosimetry.

Executive functioning in posttraumatic stress disorder: Understanding how inhibition, switching, and test modality affect reaction times.

Posttraumatic stress disorder (PTSD) has been linked to deficits in executive functioning, but the literature suggests these associations are inconsistent. Results vary depending on the task used, test modality, and the specific subdomain being measured, such as inhibitory control (interference resolution, response inhibition) or set shifting (task switching, rule switching). Notably, deficits are more consistently observed in computerized tasks that measure precise reaction times (RTs) than in classic paper-and-pencil measures, but few studies have parsed specific executive functioning deficits in PTSD using detailed analyses of RT data. The present study used a cued-switching Stroop Task to examine both interference resolution and task switching in 28 veterans with PTSD and 28 age-matched controls. Each trial required attending to a randomly presented cue and responding to the specified target while ignoring irrelevant or opposing information. Analyses of RT distributions estimated both Gaussian (normal) and ex-Gaussian (exponential) parameters. Veterans with PTSD had slower and more variable RTs than the controls on trials that required ignoring conflicting information (interference resolution, d' = .68). These effects were confined to the normal distribution, not to excessively slow responses (as estimated by ex-Gaussian parameters). Veterans with PTSD also showed modestly slower RTs on trials that required switching between cues, but Bayesian evidence for this was weak, and measures by ex-Gaussian parameters were not significant. These results highlight the importance of examining executive functioning in PTSD with a more nuanced approach, as clarity around these deficits may have important implications for future intervention and rehabilitation strategies. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Stacked collaborative transformer network with contrastive learning for video moment localization

Video grounding intends to perform temporal localization in multimedia information retrieval. The temporal bounds of the target video span are determined for the given input query. A novel interactive multi-head self-attention (IMSA) transformer is proposed to localize an unseen moment in the untrimmed video for the given image. A new semantic-trained self-supervised approach is considered in this paper to perform cross-domain learning to match the image query – video segment. It normalizes the convolution function enabling efficient correlation and collecting of semantically related video segments across time based on the image query. A double hostile Contrastive learning with Gaussian distribution parameters method is advanced to learn the representations of video. The proposed approach performs dynamically on various video components to achieve exact semantic synchronization and localization among queries and video. In the proposed approach, the IMSA model localizes frames greatly compared to other approaches. Experiments on benchmark datasets show that the proposed model can significantly increase temporal grounding accuracy. The moment occurrence is identified in the video with a start and end boundary ascertains an average recall of 86.45% and a mAP of 59.3%.

Development and application of a three-dimensional pseudo-Voigt function for muon spin relaxation analysis of weakly magnetic materials.

We present a three-dimensional pseudo-Voigt function to analyze muon spin relaxation (μSR) in weakly magnetic materials. Our approach approximates the Voigt function by superimposing Gaussian and Lorentzian functions using a one-dimensional method proposed by Di Rocco and Cruzado [Acta Phys. Pol., A 122, 666 (2012)]. We derive the peak of the Voigt function analytically and express the Half Width at Half Maximum (HWHM) of the Voigt function as a function of the HWHMs of the Gaussian and Lorentzian functions. We compare the pseudo-Voigt function to the exact Voigt function and find a maximum normalized discrepancy of ∼20% at the tail of the distribution function, depending on the ratio of Lorentzian to Gaussian HWHMs and internal magnetic field. We apply the derived three-dimensional pseudo-Voigt function to calculate μSR functions for zero- and longitudinal-field experiments and use them to fit μSR time spectra of La2-xSrxCuO4 with 2.4% Sr, employing a strong collision model with the static-based pseudo-Voigt muon spin relaxation function as the initial condition. Our results show that the Gaussian- and Lorentzian-fitted parameters and fluctuation rate are in good agreement with results from the exact Voigt function for a temperature range of 30-200K, with the deviation of Gaussian and Lorentzian width parameters reaching ∼0.15G below 30K.

Axially Non-Uniform Properties Effected Electro-Elastic Fields in Bended Piezoelectric Semiconductor Beams

This paper studied the effects of axially non-uniform material and non-uniform doping on the electro-elastic fields in bent piezoelectric semiconductor beams. The mathematical model is established based on the phenomenological theory of piezoelectric semiconductor and Euler’s beam theory. To solve the governing equations with variable coefficients, the differential quadrature method is adopted. As two representative structures, the pure bending beam and cantilever beam are studied, respectively. From the calculated results, it can be found the varied material properties changed the effective stiffness, as a result, all the field quantities are altered. Meanwhile, the symmetry of all field quantities in a pure bending beam is broken. Considering different Gaussian doping parameters [Formula: see text] and [Formula: see text], it can be observed that the distribution rule of perturbation carrier density becomes complex. The increment of [Formula: see text] broadens the width of optimum range-producing carriers, however, the increment of [Formula: see text] describes the moving of the optimum range from left to right. Importantly, the introduction of non-uniform doping breaks the limitation of perturbation carrier density in a cantilever beam with uniform material properties, the maximum value of perturbation carrier density does not appear at the fixed end only. The obtained results could be the guidance in designing high-performance electric devices.

LED-based temporal variant noise model for Fourier ptychographic microscopy

Fourier ptychographic microscopy (FPM) is a technique to reconstruct a high-resolution image from a set of low-resolution images captured with different illumination angles, which is susceptible to ambient noise, system noise, and weak currents when acquiring large-angle images, especially dark field images. To effectively address the noise problem, we propose an adaptive denoising algorithm based on a LED-based temporal variant noise model. Taking the results of blank slide samples as the reference value of noise, and analyzing the distribution of noise, we establish a statistical model for temporal variant noise, describing the relationship between temporal noise and LED spatial location. Based on this model, Gaussian denoising parameters are selected to adaptively denoise the images with different locations, with which high-resolution images can be reconstructed. Compared with other methods, the experimental results show that the proposed method effectively suppresses the noise, recovers more image details, increases the image contrast, and obtains better visual effects. Meanwhile, better objective evaluation also mirrors the advantages of the proposed algorithms.

Towards Sustainable Road Pavements: Sound Absorption in Rubber-Modified Asphalt Mixtures

In the last decade, various asphalt paving materials have undergone investigation for sound attenuation purposes. This research aims to delve into the innovative design of sustainable road pavements by examining sound absorption in rubber-modified asphalt mixtures. More specifically, the impact of alternative sustainable materials on the sound absorption of asphalt mixtures across different temperatures, precisely crumb rubber (CR) derived from recycling of end-of-life tires, was investigated. The acoustic coefficient and its Gaussian fit parameters (Peak, BandWidth, and Area Under the Curve) were evaluated. Five different types of asphalt mixtures were studied, encompassing dense, discontinuous, and open mixtures with 0%, 0.75%, and 1.50% CR incorporated through the dry process (DP). The results of sound absorption indicated a slight influence of crumb rubber at temperatures ranging from 10 °C to 60 °C, particularly in mixtures with high void content. On the other hand, as expected, the void content proved to be highly correlated with sound absorption. These findings facilitated the establishment of predictive models that correlate acoustic absorption spectra with the characteristics of asphalt mixtures. As a result, these models will be valuable in the design of the next generation of sound-absorbing pavements.

Potential energy landscape of a coarse grained model for water: ML-BOP.

We quantify the statistical properties of the potential energy landscape for a recently proposed machine learning coarse grained model for water, machine learning-bond-order potential [Chan et al., Nat. Commun. 10, 379 (2019)]. We find that the landscape can be accurately modeled as a Gaussian landscape at all densities. The resulting landscape-based free-energy expression accurately describes the model properties in a very wide range of temperatures and densities. The density dependence of the Gaussian landscape parameters [total number of inherent structures (ISs), characteristic IS energy scale, and variance of the IS energy distribution] predicts the presence of a liquid-liquid transition located close to P = 1750 ± 100 bars and T = 181.5 ± 1 K.

Multi-distribution bi-deniable inner product encryption based on Decision-LWE

Bi-deniable Encryption scheme means that when the sender and the receiver are coerced, the coercer can obtain fake plaintext, random numbers, and secret keys. It’s a solution strategy in the case of information leakage. Compared with traditional encryption, deniable encryption can provide secret communications in situations of coercion in the Post-Quantum era. Compared with sender-deniable encryption and receiver-deniable encryption, bi-deniable encryption can achieve secret communications in the situation that both sender and receiver are coerced by the coercer. So, we propose to design a bi-deniable encryption scheme under the multi-distribution model. In our bi-deniable encryption scheme, we construct a bi-deniable encryption scheme based on the assumption of Decision-Learning With Errors (DLWE) under the multi-distribution model. Firstly, the principle of Inner Product Predicate Encryption (IPPE) is applied in our scheme. Secondly, we apply the framework of Bi-Translucent Set (BTS), combined with inner product predicate encryption. Thirdly, we construct a series of probabilistic polynomial time algorithms, which apply linear transformation between different lattice structures, and Regev dual encryption. Fourthly, the statistical indistinguishability between the sampling algorithm with discrete Gaussian sampling algorithm, and the computational indistinguishability between LWE’s ciphertext samples with uniform samples, the property of indistinguishability is applied in theorem proving, which obtain Indistinguishability under Chosen Plaintext Attacks (IND-CPA) security and the property of bi-deniability. Given the value range of the Gaussian parameter and security parameter in our scheme, the correctness of the bi-deniable encryption scheme is guaranteed. We also give the security proof of IND-CPA and bi-deniability’s property by a series of games. The ‘Inner-Product Bi-Translucent Set’ Bi-Deniable Encryption scheme under the multi-distribution model is based on Decision-LWE assumption, and can avoid quantum-resistant attacks. The bi-deniable encryption scheme is firstly constructed by the inner product with the Decision-LWE assumption, which can provide better properties of security and deniability.

A new approach for atmospheric turbulence removal using low-rank matrix factorization.

In this article, a novel method for removing atmospheric turbulence from a sequence of turbulent images and restoring a high-quality image is presented. Turbulence is modeled using two factors: the geometric transformation of pixel locations represents the distortion, and the varying pixel brightness represents spatiotemporal varying blur. The main framework of the proposed method involves the utilization of low-rank matrix factorization, which achieves the modeling of both the geometric transformation of pixels and the spatiotemporal varying blur through an iterative process. In the proposed method, the initial step involves the selection of a subset of images using the random sample consensus method. Subsequently, estimation of the mixture of Gaussian noise parameters takes place. Following this, a window is chosen around each pixel based on the entropy of the surrounding region. Within this window, the transformation matrix is locally estimated. Lastly, by considering both the noise and the estimated geometric transformations of the selected images, an estimation of a low-rank matrix is conducted. This estimation process leads to the production of a turbulence-free image. The experimental results were obtained from both real and simulated datasets. These results demonstrated the efficacy of the proposed method in mitigating substantial geometrical distortions. Furthermore, the method showcased the ability to improve spatiotemporal varying blur and effectively restore the details present in the original image.

Computational extreme value statistics of strength for WC-Co composites at small specimens

The mechanical behavior of ceramic–metal (WC-Co) composites at small specimen level is subject to significant uncertainty due to the random assembly of (1) its constituent phases, (2) the crystal orientation of the WC grains, (3) the intrinsic residual stresses, and (4) the variable level of confinement of the metallic binder, within the contiguous network of much stiffer WC grains. In the existing literature, a Weibull distribution of strength is typically assumed at large specimen level. However, it is unknown whether this assumption can be carried over to small samples. Thus, the aim of this study is to shed light into the randomness in the mechanical strength of WC-Co composites, by using Monte Carlo simulations employing Latin Hypercube Sampling. The marginal distribution of the model parameters is assumed to be a grafted zero-threshold Weibull-Gaussian probability distribution, applied individually to the ceramic particles and the metallic binder using autocorrelation fields. The deterministic models are the recently calibrated microplane models M7WC and MPJ2 for the ceramic phase (WC) and the metallic matrix (Co) respectively. Test data in the form of strength histograms, gathered from literature on tensile tests conducted on nanowires of WC-Co composites, are used to calibrate the Weibull and Gaussian parameters, the grafting point locations and the autocorrelation lengths. The deterministic models M7WC and MPJ2 are assumed to be predictive in this study, as they were already calibrated in a previously published work by the authors. The finite element meshes used in the Monte Carlo simulations are generated using real tomography images, which easily captures the Gaussian part of the probability density function (pdf) without any random variables. This approach is shown to account successfully for the randomness observed in the strength of these ceramic–metal composites. Predictions are made for the strength distribution of differently sized nanowire specimens in tension.

SBHSR: single-band super-resolution method for hyperspectral images based on blind degradation and fusion of auxiliary band

Despite natural image super-resolution (SR) methods have achieved great success, super-resolution methods for hyperspectral image (HSI) with rich spectral features are still a very challenging task. Furthermore, due to the diversity of HSI captured by different cameras, their degradation conditions vary. Currently available HSI SR methods are mainly based on fixed degradation models, and their performance is severely affected when the actual degradation does not match the assumed degradation model. To address this issue, this paper proposes a single-band hyperspectral image super-resolution (SBHSR) method for HSI. This method assumes that the degradation Gaussian blur kernel parameters of the HSI are unknown, and adapts to various degradation conditions by conducting blind super-resolution on the image. It also combines another band with better spatial structure as an auxiliary band for feature fusion. To address the spectral differences between the super-resolution band and the auxiliary band, we use the Spatial Adaptation Module (SAM) to map the feature distribution, ensuring consistent spectral brightness between the auxiliary band and the super-resolution band. Due to the use of blind degradation super-resolution method, our method is more robust compared to previous HSI SR methods. Additionally, as it is specifically designed for single-band hyperspectral image super-resolution, it has the advantage of being faster and more efficient. Experimental validation on two hyperspectral datasets demonstrates the superiority of our method, as it improves the spatial resolution of the image while preserving the spectral features, and performs better than existing blind super-resolution methods.

A Sparse Model for Electricity Price Forecasting with LASSO-GRBFN and Brain Storm Optimization

This paper proposes a new method for predicting electricity prices. The method is based on the Generalized Radial Basis Function Network (GRBFN), an extension of the RBFN of Artificial Neural Network (ANN) used for one-step-ahead forecasting. The GRBFN is superior to the RBFN because it evaluates the Gaussian function parameters during learning. Conventional ANN methods consider overfitting with the weight decay method, based on the L2 norm of weights between neurons, but there is still room for improvement. Therefore, the paper proposes using the Least Absolute Shrinkage and Selection Operator (LASSO) to improve the model’s performance. Furthermore, the paper presents the Brain Storm Optimization (BSO) of evolutionary computation to evaluate the cost function with the term of the L1 norm. Conventional methods with the gradient are not suitable for the L1 norm. The proposed method is successfully applied to actual data of ISO New England in the USA.

Z-Splat: Z-Axis Gaussian Splatting for Camera-Sonar Fusion.

Differentiable 3D-Gaussian splatting (GS) is emerging as a prominent technique in computer vision and graphics for reconstructing 3D scenes. GS represents a scene as a set of 3D Gaussians with varying opacities and employs a computationally efficient splatting operation along with analytical derivatives to compute the 3D Gaussian parameters given scene images captured from various viewpoints. Unfortunately, capturing surround view (360° viewpoint) images is impossible or impractical in many real-world imaging scenarios, including underwater imaging, rooms inside a building, and autonomous navigation. In these restricted baseline imaging scenarios, the GS algorithm suffers from a well-known 'missing cone' problem, which results in poor reconstruction along the depth axis. In this paper, we demonstrate that using transient data (from sonars) allows us to address the missing cone problem by sampling high-frequency data along the depth axis. We extend the Gaussian splatting algorithms for two commonly used sonars and propose fusion algorithms that simultaneously utilize RGB camera data and sonar data. Through simulations, emulations, and hardware experiments across various imaging scenarios, we show that the proposed fusion algorithms lead to significantly better novel view synthesis (5dB improvement in PSNR) and 3D geometry reconstruction (60% lower Chamfer distance).

Vision-based strawberry classification using generalized and robust deep networks

Grading of agricultural products such as fruits and vegetables based on ripeness level and visual defects for the purpose of export, storage and waste control is a process of special importance. Various methods have been used to detect levels of ripeness and the quality of agricultural products, some of which are destructive and some non-destructive. The machine vision system is one of the non-destructive and accurate systems in the field of detecting the quality of agricultural products. In this study, we propose a robust and generalized model via fine-tuning the pre-trained networks for the classification of strawberry fruit. A dataset containing 800 confirmed strawberry images in four classes (unripe, half-ripe, ripe, and damaged) was used. Instead of using fundamental data augmentation (FDA) techniques to prevent overfitting problem and increase the robustness of the model, we employed a novel learning-to-augment strategy (LAS) using noisy images that creates new noisy variant of data via original images. By using the Bayesian optimization algorithm, controllers were used to select the optimal noise parameters of Gaussian and speckle noise to generate new noise images. The best policies of data augmentation based on LAS was used to fine-tune pre-trained cutting-edge models (GoogleNet, ResNet18, and ShuffleNet). The results show that in all the proposed scenarios (i.e. using original data without data augmentation, employing FDA, and applying LAS) the GoogleNet model was able to achieve 96.88 %, 97.50 %, and 98.85 % accuracy, respectively.