Gaussian Parameters Research Articles

Modeling Annotation Uncertainty with Gaussian Heatmaps in Landmark Localization

In landmark localization, due to ambiguities in defining their exact position, landmark annotations may suffer from large observer variabilities, which result in uncertain annotations. To model the annotation ambiguities of the training dataset, we propose to learn anisotropic Gaussian parameters modeling the shape of the target heatmap during optimization. Furthermore, our method models the prediction uncertainty of individual samples by fitting anisotropic Gaussian functions to the predicted heatmaps during inference. Besides state-of-the-art results, our experiments on datasets of hand radiographs and lateral cephalograms also show that Gaussian functions are correlated with both localization accuracy and observer variability. As a final experiment, we show the importance of integrating the uncertainty into decision making by measuring the influence of the predicted location uncertainty on the classification of anatomical abnormalities in lateral cephalograms.

Radial Basis Function Network with Differential Privacy

Differential privacy (DP) remains a potent solution to what is arguably the defining issue in machine learning: balancing user privacy with an ever-increasing need for data. Practitioners must respect privacy, especially in sensitive healthcare domains. DP strives towards this aim by adding noise to training data to occlude its origin and nature, and is ideal for multiple Neural Network (NN) types. This includes deep varieties that utilise multiple hidden layers, and shallow ones with single hidden layers such as the Radial Base Function Network (RBFN). The work herein explores DP within this context by devising a model that leverages Gaussian RBF parameters to add privacy during training. Our model’s efficacy is examined against two real and three synthetic datasets, with results showing reasonable trade-offs between accuracy and privacy. With high intra-class variation, we retained 100% accuracy for two synthetic datasets and a drop of only 1.72% for another. If privacy is prioritised with low intra-class variation, we achieved accuracy drops of 8%–23% with an inherited epsilon that never exceeds one, indicating a good privacy guarantee. We also show that timely training is achievable on a high-dimensional dataset consisting of 2M records and 170 features.

Copula-Based Analysis of Power Allocation for Two-Sensor Composite Fading MAC With Dependent Channel Coefficients

In this article, the problem of power allocation in a wireless multiple access channel with two sensors is addressed, where squared channel coefficients, and, hence, signal-to-noise ratios are assumed to have Gaussian mixture distributions and the dependence between channel coefficients is described by using copula theory. For this model, outage probability is analyzed as the objective function for formulating the power allocation, and for the product copula function or in the independent channels case, a closed form of outage probability is obtained. Considering that the corresponding optimizations do not lead to convex optimization problems, MATLAB simulation is employed to solve the power allocation problem. Various Gaussian mixture distribution parameters are exploited for the links between the sensors and the fusion center with different dependencies and the results are compared.

Lattice‐based hash‐and‐sign signatures using approximate trapdoor, revisited

For the purpose of improving the efficiency of the cryptosystems built upon lattice trapdoors, Chen, Genise and Mukherjee at ASIACRYPT 2019 modified the gadget trapdoor (G-trapdoor) to an approximate trapdoor, which enables one to sample short preimages approximately from a discrete Gaussian distribution. The implementation shows that the sizes of the hash-and-sign signature scheme can be reduced to 3.67 kB for an estimation of 81.67-bit security, and 9.97 kB for an estimation of 168.81-bit security. In this study, the spherical sampling method is adapted to the non-spherical setting, without leaking any information about the trapdoor. Due to the fact that the signature size and the concrete security are closely related to the Gaussian parameter of the sampling algorithm, this technique provides a tradeoff between them. Specifically, two modes of parameters were set up for different goals. (a) Mode 1 admits to achieve the ‘win–win’ scenario, that is, gain concrete security and simultaneously reduce the signature size. Our proof-of-concept implementation shows that for an estimation of 94.5- and 185.88-bit security, the signature sizes can be reduced to 3.3 and 6.98 kB. (b) Mode 2 aims mainly to further reduce the signature sizes, without a decrease in the security level. The implementation shows that the signature size can be reduced to 2.35 kB for an estimation of 81.67-bit security, and 5.75 kB for an estimation of 168.82-bit security.

Atom cloud detection and segmentation using a deep neural network

We use a deep neural network (NN) to detect and place region-of-interest (ROI) boxes around ultracold atom clouds in absorption and fluorescence images—with the ability to identify and bound multiple clouds within a single image. The NN also outputs segmentation masks that identify the size, shape and orientation of each cloud from which we extract the clouds’ Gaussian parameters. This allows 2D Gaussian fits to be reliably seeded thereby enabling fully automatic image processing. The method developed performs significantly better than a more conventional method based on a standardized image analysis library (Scikit-image) both for identifying ROI and extracting Gaussian parameters.

Fabrication of an Anti-Reflective Microstructure on ZnS by Femtosecond Laser Bessel Beams.

As an important mid-infrared to far-infrared optical window, ZnS is extremely important to improve spectral transmission performance, especially in the military field. However, on account of the Fresnel reflection at the interface between the air and the high-strength substrate, surface optical loss occurs in the ZnS optical window. In this study, the concave antireflective sub-wavelength structures (ASS) on ZnS have been experimentally investigated to obtain high transmittance in the far-infrared spectral range from 6 μm to 10 μm. We proposed a simple method to fabricate microhole array ASS by femtosecond Bessel beam, which further increased the depth of the microholes and suppressed the thermal effects effectively, including the crack and recast layer of the microhole. The influence of different Gaussian and Bessel beam parameters on the microhole morphology were explored, and three ASS structures with different periods were prepared by the optimized Bessel parameters. Ultimately, the average transmittance of the sample with the ASS microhole array period of 2.6 μm increased by 4.1% in the 6 μm to 10 μm waveband, and the transmittance was increased by 5.7% at wavelength of 7.2 μm.

Reduction of pixel deviation effect in three-dimensional light-field display based on the fitting algorithm with dense-viewpoints

In three-dimensional (3D) light-field displays, the quality of a reconstructed image degrades with increasing depth of field due to the pixel deviation effect. A fitting algorithm of dense-viewpoints is proposed to reduce the pixel deviation with high image quality. To analyze the pixel deviation effect, the human vision-based perspectives perceived from 3D light-field display system with the holographic functional screen are mathematically modeled. The optimized synthetic image is coded according to the mapping relationship from the display model and the optimal Gaussian distribution parameters achieved with the maximum likelihood estimation method. The effectivenesses of the proposed model and the optimization method are verified through simulations as well as optical experiments. A human vision-based high resolution 3D light-field display with 45-degree viewing angle and the large depth of field is achieved.

Gaussian Parameters Correlate with the Spread of COVID-19 Pandemic: The Italian Case

Until today, numerous models have been formulated to predict the spreading of Covid-19. Among them, the actively discussed susceptible-infected-removed (SIR) model is one of the most reliable. Unfortunately, many factors (i.e., social behaviors) can influence the outcomes as well as the occurrence of multiple contributions corresponding to multiple waves. Therefore, for a reliable evaluation of the conversion rates, data need to be continuously updated and analyzed. In this work, we propose a model using Gaussian functions, coming from the solution of an ordinary differential equation representing a logistic model, able to describe the growth rate of infected, deceased and recovered people in Italy. We correlate the Gaussian parameters with the number of people affected by COVID-19 as a function of the large-scale anti-contagion control measures strength, and also of vaccines effects adopted to reach herd immunity. The superposition of gaussian curves allow modeling the growth rate of the total cases, deceased and recovered people and reproducing the corresponding cumulative distribution and probability density functions. Moreover, we try to predict a time interval in which all people will be infected or vaccinated (with at least one dose) and/or the time end of pandemic in Italy when all people have been infected or vaccinated with two doses.

Optimization of broadband Λ -type quantum memory using Gaussian pulses

Optical quantum memory, the ability to store photonic quantum states and retrieve them on demand, is an essential resource for emerging quantum technologies and photonic quantum information protocols. Simultaneously achieving high efficiency and high-speed, broadband operation is an important task necessary for enabling these applications. We investigate the optimization of a large class of optical quantum memories based on resonant and near-resonant interaction with ensembles of $\mathrm{\ensuremath{\Lambda}}$-type level systems with the restriction that the temporal envelope of all optical fields must be Gaussian, which reduces experimental complexity. Through this optimization we demonstrate an experimentally simple path to saturation of the protocol-independent storage efficiency bound that is valid for a wide range of memory bandwidths, including those that are broadband and high speed. Examining the resulting optimal Gaussian control field parameters, we find a continuous transformation between three physically distinct resonant quantum memory protocols. We compare this Gaussian optimization scheme with standard shape-based optimization.

Gaussian Mixture Model Parameter Optimization in Range Points Migration-Based Three-Dimensional Radar Imaging

Millimeter-wave (MMW) imaging radar is one of the major environmental sensors in recent self-driving or driver assistance systems and provides environmentally robust sensing even in optically challenging conditions. As a promising three-dimensional (3-D) radar imaging technique, this study focuses on the range points migration (RPM) method, which has advantages in terms of accuracy and low complexity. In the original RPM algorithm, the parameters are manually or empirically determined by considering the sensor configuration and target shape or distance. To address this limitation, we introduced a novel parameter optimization scheme based on the Gaussian mixture model (GMM) and the expectation maximization (EM) algorithm. In addition, we used the k-space decomposition-based parameter determination scheme to determine target shape-dependent parameter selection. The results, assuming the human body imaging scenario, showed that our proposed method retains a highly accurate target image without requiring empirical parameter selection.

Robustness meets algorithms

In every corner of machine learning and statistics, there is a need for estimators that work not just in an idealized model, but even when their assumptions are violated. Unfortunately, in high dimensions, being provably robust and being efficiently computable are often at odds with each other. We give the first efficient algorithm for estimating the parameters of a high-dimensional Gaussian that is able to tolerate a constant fraction of corruptions that is independent of the dimension. Prior to our work, all known estimators either needed time exponential in the dimension to compute or could tolerate only an inverse-polynomial fraction of corruptions. Not only does our algorithm bridge the gap between robustness and algorithms, but also it turns out to be highly practical in a variety of settings.

Adapting Gaussian YOLOv3 with transfer learning for overhead view human detection in smart cities and societies

Nowadays, deep neural networks are widely applied in sustainable smart cities and societies, including smart manufacturing, healthcare, industries, agriculture, surveillance, and various artificial intelligence-based real-life applications. In this regard, the human detection system has gained notable attention since it is recognized as a crucial task in intelligent surveillance applications. Researchers practiced a variety of computer vision and deep neural networks-based techniques for human detection-based applications; however, they often focused on the frontal view camera perspective. Thus, in this work, we have introduced a human detection system for intelligent surveillance in smart cities and societies with a completely distinct perspective, i.e., an overhead perspective that can provide sufficient visibility and coverage of a scene in congested and obstructed environments. However, human appearance can be difficult from such an extreme point of view, as there are significant variations in humans’ poses and appearances. Therefore, in this work, leveraging the deep neural network-based object detection technique, the Gaussian YOLOv3 algorithm is used for human detection. The algorithm determines the bounding box uncertainty by modeling its coordinates as a Gaussian parameter, improving accuracy and reducing false positives. A Gaussian YOLOv3 is combined with channel attention and feature intertwine modules to improve specific feature maps. The channel attention module is combined with the feature map to learn each channel's weight autonomously, improve the key features, and enhance the network's ability to discriminate between humans and background. At the same time, different channels of the feature map are intertwined to obtain more representative features. Finally, the features obtained from the attention and feature intertwine modules are fused to form an improved feature map. In addition, to further increase the detection accuracy of the algorithm for human detection, transfer learning is adopted. The experimental outcomes reveal that training improves the Gaussian YOLOv3 algorithm's potential for human detection with an overall detection accuracy of 94%.

A Novel Anti-Jamming Technique for INS/GNSS Integration Based on Black Box Variational Inference

In this paper, a novel anti-jamming technique based on black box variational inference for INS/GNSS integration with time-varying measurement noise covariance matrices is presented. We proved that the time-varying measurement noise is more similar to the Gaussian distribution with time-varying mean value than to the Inv-Gamma or Inv-Wishart distribution found by Kullback–Leibler divergence. Therefore, we assumed the prior distribution of measurement noise covariance matrices as Gaussian, and calculated the Gaussian parameters by the black box variational inference method. Finally, we obtained the measurement noise covariance matrices by using the Gaussian parameters. The experimental results illustrate that the proposed algorithm performs better in resisting time-varying measurement noise than the existing Variational Bayesian adaptive filter.

Hot spots localization in proteins by optimized short time Ramanujan Fourier transform.

Specific functions in biological processes are dependent on protein-protein interactions. Hot spot residues play a key role in the determination of these interactions and have wide applications in engineering proteins and drug discovery. Experimental techniques to identify hotspots are often labor intensive and expensive. Also, most of the computational methods which have been developed are structure based and need some training. In this work, hotspots have been identified by sequence information alone using the Resonant Recognition Model (RRM). The proposed method uses characteristic period in place of traditionally used characteristic frequency by RRM-based methods. The characteristic period has been extracted from the consensus spectrum of protein families using the Ramanujan Fourier Transform (RFT). Position-period plots for proteins have been generated using Short Time RFT (ST-RFT) with a Gaussian window. Hot spots have been identified by thresholding of the signal corresponding to the protein's characteristic period in the ST-RFT. To enhance the performance of the ST-RFT, Gaussian window shape parameter has been optimized using concentration measure as a metric. Better sensitivity of this method has been observed compared to other reported RRM-based methods. Since the method is model independent it does not requires any training and can be readily used for any protein sequence provided its interface residues and protein family are known.

Confidence-Aware Object Detection Based on MobileNetv2 for Autonomous Driving.

Object detection is an indispensable part of autonomous driving. It is the basis of other high-level applications. For example, autonomous vehicles need to use the object detection results to navigate and avoid obstacles. In this paper, we propose a multi-scale MobileNeck module and an algorithm to improve the performance of an object detection model by outputting a series of Gaussian parameters. These Gaussian parameters can be used to predict both the locations of detected objects and the localization confidences. Based on the above two methods, a new confidence-aware Mobile Detection (MobileDet) model is proposed. The MobileNeck module and loss function are easy to conduct and integrate with Generalized-IoU (GIoU) metrics with slight changes in the code. We test the proposed model on the KITTI and VOC datasets. The mean Average Precision (mAP) is improved by 3.8 on the KITTI dataset and 2.9 on the VOC dataset with less resource consumption.

A 2D Gaussian beam launcher applied in parallel plate waveguide field mapping systems

Abstract3D Gaussian beam generation techniques have been studied by many researchers, while the problem of 2D Gaussian beam generations have not been addressed fully. However, 2D Gaussian beams are still useful in research and applications. This article presents the mechanism, design, simulation, and measurement of a 2D Gaussian beam launcher, which allows the easy control of the width and the location of Gaussian beam waist. The proposed Gaussian beam launcher consists of a radiator array fed with the designable excitation phase and magnitude to each radiator, so as to generate the necessary boundary condition that the Gaussian beam can be excited and propagate. Both simulation and measurement results are nearly identical with the profile of the lowest order 2D Gaussian beam mode. The proposed 2D Gaussian beam launcher is suitable as the excitation source of parallel plate waveguide measurement systems. By integrating tunable components, it can be easily extended in the future to control Gaussian beam parameters dynamically.

Deepwater gas concentration feature extraction based on fluid mechanics

In order to solve the complex production process, high technical content, bad working environment, frequent blowout accidents, and other problems. A method based on the concentration distribution characteristics of explosion flue gas emission in deepwater area is proposed. Based on the computational fluid dynamics theory, the basic governing equations of blowout and explosion of natural gas in deepwater are obtained. Based on the basic governing equations of deepwater natural gas blowout and explosion, the diffusion of natural gas in deepwater natural gas blowout and explosion is studied by using the K-turbulence model. The diffusion process of gas explosion in deepwater is obtained. Two Gaussian smoke model parameters, atmospheric diffusion parameter and average wind speed, were realized by combining the segmenting smoke flow with Gaussian smoke mass concentration distribution. The blowout and explosion process of natural gas in deepwater is simulated by MATLAB software, and the distribution characteristics of flue gas concentration are analyzed. The results showed that the smoke concentration and the change of smoke concentration were not affected by the change.

Aperture Averaged Scintillation of Gaussian Beam in Strong Oceanic Turbulence

Oceanic turbulence caused by salinity and temperature fluctuations underwater affects the characteristics of a transmit laser beam resulting in turbulence-induced intensity fluctuations (i.e., scintillation) at the receiver. Oceanic optical wireless communication (OOWC) systems employ a receiver lens of aperture to focus the collected light onto a photodetector. This way, aperture averaging takes place and the scintillation on the detector is reduced. Using the modified Rytov theory, aperture averaged scintillation of the Gaussian beam passing through strong oceanic turbulence is examined in this study. Effects of Gaussian beam parameters and the turbulence parameters on the aperture averaged scintillation and the aperture averaging factor are illustrated. The scintillation behaviors of the limiting cases of a spherical wave and a plane wave are also reported. Results show that aperture averaged scintillation decreases with increasing the size of the receiver aperture for any turbulence level. The effect of Gaussian size on the aperture averaged scintillation varies depends on the turbulence level. It is also shown that there is a close match between the point scintillation index values obtained from the modified Rytov theory and the conventional Rytov theory in low levels of turbulence.

Chirped super–Gaussian and super–sech pulse perturbation of nonlinear Schrödinger's equation with quadratic–cubic nonlinearity by variational principle

We apply variational method to the perturbed nonlinear Schrödinger equation having quadratic-cubic form of nonlinearity, to study localized optical pulses. Super-Gaussian and super-sech solitons are used as envelopes for the trial function. Numerical simulations are presented for specific values of the Gaussian and super-sech pulse parameters. The impact of the quadratic-cubic terms on the evolution for different parameters is assessed. In general, when the nonlinear quadratic and cubic coefficients increase, the frequency of the oscillations of the collective variables also increases.

Robot grasping method optimization using improved deep deterministic policy gradient algorithm of deep reinforcement learning

Robot grasping has become a very hot research field so that the requirements for robot operation are getting higher and higher. In previous research studies, the use of traditional target detection algorithms for grasping is often very inefficient, and this article is dedicated to improving the deep reinforcement learning algorithm to improve the grasping efficiency and solve the problem of robots dealing with the impact of unknown disturbances on grasping. Using the characteristic that deep reinforcement learning actively explores the unknown environment, a Gaussian parameter Deep Deterministic Policy Gradient (Gaussian-DDPG) algorithm based on the Importance-Weighted Autoencoder (IWAE) is proposed to realize the robot's autonomous learning of the grasping task. Traditional coordinate positioning methods and deep learning methods have poor grasping effects for disturbed situations (such as the movement of the target object). The IWAE algorithm is used to compress the high-dimensional information of the original visual input to the hidden space and pass it to the deep reinforcement learning network as part of the state value. Based on the classic DDPG algorithm, it smoothly adds Gaussian parameters to improve the exploratory nature of the algorithm, dynamically sets the robot grasping space parameters to adapt to the workspace of multiple scales, and finally, realizes the accurate grasping of the robot. Relying on the possible position information deviation of the visual information, the control of the grasping position by the manipulator torque information is further optimized to improve the grasping efficiency of disturbed objects.