摘要 :
The goal of this work is to parallelize the multistep method for the numerical approximation of the Backward Stochastic Differential Equations (BSDEs) in order to achieve both, a high accuracy and a reduction of the computation ti...
展开
The goal of this work is to parallelize the multistep method for the numerical approximation of the Backward Stochastic Differential Equations (BSDEs) in order to achieve both, a high accuracy and a reduction of the computation time as well. In the multistep scheme the computations at each grid point are independent and this fact motivates us to select massively parallel GPU computing using CUDA. In our investigations we identify performance bottlenecks and apply appropriate optimization techniques for reducing the computation time, using a uniform domain. Finally, a Black-Scholes BSDE example is provided to demonstrate the achieved acceleration on GPUs.
收起
摘要 :
The goal of this work is to parallelize the multistep method for the numerical approximation of the Backward Stochastic Differential Equations (BSDEs) in order to achieve both, a high accuracy and a reduction of the computation ti...
展开
The goal of this work is to parallelize the multistep method for the numerical approximation of the Backward Stochastic Differential Equations (BSDEs) in order to achieve both, a high accuracy and a reduction of the computation time as well. In the multistep scheme the computations at each grid point are independent and this fact motivates us to select massively parallel GPU computing using CUDA. In our investigations we identify performance bottlenecks and apply appropriate optimization techniques for reducing the computation time, using a uniform domain. Finally, a Black-Scholes BSDE example is provided to demonstrate the achieved acceleration on GPUs.
收起
摘要 :
The goal of this work is to parallelize the multistep method for the numerical approximation of the Backward Stochastic Differential Equations (BSDEs) in order to achieve both, a high accuracy and a reduction of the computation ti...
展开
The goal of this work is to parallelize the multistep method for the numerical approximation of the Backward Stochastic Differential Equations (BSDEs) in order to achieve both, a high accuracy and a reduction of the computation time as well. In the multistep scheme the computations at each grid point are independent and this fact motivates us to select massively parallel GPU computing using CUDA. In our investigations we identify performance bottlenecks and apply appropriate optimization techniques for reducing the computation time, using a uniform domain. Finally, a Black-Scholes BSDE example is provided to demonstrate the achieved acceleration on GPUs.
收起
摘要 :
Generally the pulse repetition frequency in airborne SAR system is much higher than signal bandwidth to guarantee the maximum slant range requested. In order to reduce the large processing amount caused by high PRF, we always prep...
展开
Generally the pulse repetition frequency in airborne SAR system is much higher than signal bandwidth to guarantee the maximum slant range requested. In order to reduce the large processing amount caused by high PRF, we always preprocess the echo signal in azimuth before the whole processing procedure. The conventional method is pre-filtering, which has some approximation and relatively large computation amount. This paper mainly studies pre-accumulation as another preprocessing method with relatively small computation amount. On the basis of SAR signal processing theory, we analyze and simulate the result of direct accumulation of incoherent signal in azimuth, as well as the phase loss caused by incoherence and the factors affecting the phase loss. Furthermore, present a resolution to eliminate the phase loss, which adjusting signal phase by multiplying phase difference before accumulating, and compare the computation amount with that of pre-filtering. After that, point out the condition suitable to use this method. Finally, we apply this method to raw data processing of airborne SAR and the result shows that the proposed method can achieve the image result with the similar quality to that of original processing, while reduce the computation amount considerably.
收起
摘要 :
Generally the pulse repetition frequency in airborne SAR system is much higher than signal bandwidth to guarantee the maximum slant range requested. In order to reduce the large processing amount caused by high PRF, we always prep...
展开
Generally the pulse repetition frequency in airborne SAR system is much higher than signal bandwidth to guarantee the maximum slant range requested. In order to reduce the large processing amount caused by high PRF, we always preprocess the echo signal in azimuth before the whole processing procedure. The conventional method is pre-filtering, which has some approximation and relatively large computation amount. This paper mainly studies pre-accumulation as another preprocessing method with relatively small computation amount. On the basis of SAR signal processing theory, we analyze and simulate the result of direct accumulation of incoherent signal in azimuth, as well as the phase loss caused by incoherence and the factors affecting the phase loss. Furthermore, present a resolution to eliminate the phase loss, which adjusting signal phase by multiplying phase difference before accumulating, and compare the computation amount with that of pre-filtering. After that, point out the condition suitable to use this method. Finally, we apply this method to raw data processing of airborne SAR and the result shows that the proposed method can achieve the image result with the similar quality to that of original processing, while reduce the computation amount considerably.
收起
摘要 :
Buoyancy-driven flow in high temperature difference region has attracted more and more attention due to its application in nuclear industry, such as fatigue analysis. A high-precision Non-Boussinesq model is developed to simulate ...
展开
Buoyancy-driven flow in high temperature difference region has attracted more and more attention due to its application in nuclear industry, such as fatigue analysis. A high-precision Non-Boussinesq model is developed to simulate turbulent natural convection by adding source terms through User Defined Functions (UDF) in Fluent software. The analyses of velocity and temperature profile in an air-filled-rectangular cavity are conducted for the validation of this developed model. A correlation between average Nusselt number and Rayleigh number for the large aspect ratio cavity is obtained to predict heat transfer characteristic. Finally, application of Non-Boussinesq model to a high temperature difference region of heat exchanger is investigated. The results reveal that scale effect is important for the heat transfer of natural convection, and the decrease of scale will lead to an increase of heat transfer coefficient.
收起
摘要 :
Buoyancy-driven flow in high temperature difference region has attracted more and more attention due to its application in nuclear industry, such as fatigue analysis. A high-precision Non-Boussinesq model is developed to simulate ...
展开
Buoyancy-driven flow in high temperature difference region has attracted more and more attention due to its application in nuclear industry, such as fatigue analysis. A high-precision Non-Boussinesq model is developed to simulate turbulent natural convection by adding source terms through User Defined Functions (UDF) in Fluent software. The analyses of velocity and temperature profile in an air-filled-rectangular cavity are conducted for the validation of this developed model. A correlation between average Nusselt number and Rayleigh number for the large aspect ratio cavity is obtained to predict heat transfer characteristic. Finally, application of Non-Boussinesq model to a high temperature difference region of heat exchanger is investigated. The results reveal that scale effect is important for the heat transfer of natural convection, and the decrease of scale will lead to an increase of heat transfer coefficient.
收起
摘要 :
Buoyancy-driven flow in high temperature difference region has attracted more and more attention due to its application in nuclear industry, such as fatigue analysis. A high-precision Non-Boussinesq model is developed to simulate ...
展开
Buoyancy-driven flow in high temperature difference region has attracted more and more attention due to its application in nuclear industry, such as fatigue analysis. A high-precision Non-Boussinesq model is developed to simulate turbulent natural convection by adding source terms through User Defined Functions (UDF) in Fluent software. The analyses of velocity and temperature profile in an air-filled-rectangular cavity are conducted for the validation of this developed model. A correlation between average Nusselt number and Rayleigh number for the large aspect ratio cavity is obtained to predict heat transfer characteristic. Finally, application of Non-Boussinesq model to a high temperature difference region of heat exchanger is investigated. The results reveal that scale effect is important for the heat transfer of natural convection, and the decrease of scale will lead to an increase of heat transfer coefficient.
收起
摘要 :
This study proposes a novel approach for image super-resolution (SR) that combines deep learning algorithms, adaptive multi-path structures, and classification models. The goal is to enhance industrial product details, improve vis...
展开
This study proposes a novel approach for image super-resolution (SR) that combines deep learning algorithms, adaptive multi-path structures, and classification models. The goal is to enhance industrial product details, improve visual quality, and achieve better quantitative metrics such as PSNR and SSIM. The proposed multi-path super-resolution algorithm utilizes a big-size convolution kernel and residual network to extract features at different scales, enabling the capture and enhancement of fine details in low-resolution images. Two different loss functions are incorporated to improve the visual quality and fidelity of the SR images. Furthermore, the integration of a super-resolution model with a ResNet-18 classification model enhances image clarity, detail retention, and overall performance,. The experimental results demonstrate that our proposed super-resolution (SR) algorithm outperforms several typical methods. Additionally, incorporating the ResNet-18 classification model improves the performance of the model on the NEU-CLS dataset, achieving higher accuracy, recall, precision, and F1-score compared to the original dataset.
收起
摘要 :
In recent times, shared bikes have become a new trend for improving mobility in many cities. More and more people choose shared bikes as their "final 1-mile" solution for urban transportation. However, modeling to estimate the opt...
展开
In recent times, shared bikes have become a new trend for improving mobility in many cities. More and more people choose shared bikes as their "final 1-mile" solution for urban transportation. However, modeling to estimate the optimal number of shared bikes deployed has not been well addressed. To support bike-sharing companies in better deploying shared bikes, in this research, we propose a new deep residual network model to determine the optimal number of shared bikes. The novelty of this model is that residual networks are adopted to create a deep learning model, which is the first to be used in the shared bike deployment domain. Moreover, in the proposed model, three strategies have been considered to balance the profit of the service providers and the welfare of the public. Simulation results show that our model has achieved a coefficient of determination (R2 score) of 0.8998, showing that the model performs satisfactorily in determining the optimal number of shared bikes when compared to several typical prediction approaches, such as (a) gradient boosters, (b) support vector machines, (c) boosting trees, and (d) extreme gradient boosting trees.
收起