We developed a highly parallel simulator of Optical Coherence Tomography (OCT) of objects with arbitrary spatial distributions. This Monte Carlo method based simulator models the object as a tetrahedron-based mesh, an...
详细信息
ISBN:
(纸本)9781628413632
We developed a highly parallel simulator of Optical Coherence Tomography (OCT) of objects with arbitrary spatial distributions. This Monte Carlo method based simulator models the object as a tetrahedron-based mesh, and implements an advanced importance sampling scheme. This new method makes OCT simulations more practical, since the corresponding serial Central Processing Unit (CPU) based implementation requires approximately 360 hours to simulate OCT imaging of a single B-scan. We implemented this new simulator on Graphics Processing Units (gpus) using the Compute Unified Device Architecture (CUDA) platform and programming model by NVIDIA. We demonstrated that our new simulator requires one order of magnitude less time, compared to its serial implementation, to simulate the same OCT images. Our new parallel OCT simulator could be an important and practical tool to study different OCT phenomena and to design novel OCT systems with superior imaging performance.
In this study, we combined the Maximum Likelihood Estimator from our previous works with a Sequential Importance Resampling (SIR) particle filter to estimate the states of the stochastic Gompertz tumor growth model. W...
详细信息
ISBN:
(纸本)9781728157429
In this study, we combined the Maximum Likelihood Estimator from our previous works with a Sequential Importance Resampling (SIR) particle filter to estimate the states of the stochastic Gompertz tumor growth model. We also implemented a parallel version in CUDA for the SIR filter in order to reduce its execution time. Extensive simulations with synthetic data were run to examine whether the SIR filter can provide more accurate state estimates in respect to the Normalized Mean Squared Deviation criterion compared to those provided by the deterministic Gompertz model. Moreover, we monitored and compared the execution time of the SIR's parallel and sequential implementations for different numbers of particles. The results showed that the SIR filter can estimate the system's states very accurately, even at the early tumor growth stages. Additionally, the parallel implementation that ran on the gpu was way more efficient than the implementation that ran on the CPU. By combining the Maximum Likelihood Estimator (MLE) with an SIR filter, we were able to obtain very accurate estimates of the tumors' volume. Furthermore, the execution time for the SIR filter was significantly decreased by taking advantage of the gpus ability to perform a very large number of computations in parallel.
The spectral transforms of logic functions have numerous applications in logic synthesis, signal processing, pattern recognition, and in many related areas. It is very important to be able to efficiently compute these...
详细信息
ISBN:
(纸本)9781665485005
The spectral transforms of logic functions have numerous applications in logic synthesis, signal processing, pattern recognition, and in many related areas. It is very important to be able to efficiently compute these transforms. Python frameworks show extremely high computation performance for (Fast Fourier Transform) FFT-based algorithms executed on graphics processing units (gpus). Thus, this paper presents a comparison of computation time for different implementations of spectral transform of logic functions, performed on gpus using three different Python frameworks. It is used TensorFlow, PyTorch, and CuPy, frameworks. The experiments were performed on Nvidia GeForce RTX 2060 gpu, which belongs to the middle gpu price range. Computational times are compared using randomly-generated truth vectors of size up to 4096. The aim of this paper is to identify the computing platform and Python programming framework which produces the fastest spectral transform of logic functions.
A null-space free method with the FFT-based matrix-vector multiplications was proposed to solve the Maxwell equations that model the three-dimensional photonic crystals. The most time-consuming parts of this method we...
详细信息
A null-space free method with the FFT-based matrix-vector multiplications was proposed to solve the Maxwell equations that model the three-dimensional photonic crystals. The most time-consuming parts of this method were the FFT-based matrix-vector multiplications. In this article, we propose new mathematical formulas to compute the FFT-based matrix-vector multiplications and derive highly efficient algorithms on top of the NVIDIA gpu architecture. The resulting algorithms are approximately two- to threefold faster than the previous algorithms. We have successfully used a single NVIDIA Tesla P100 gpu to solve a set of generalized eigenvalue problems of 5, 184, 000 dimensions in 17 to 22 seconds for each problem. Furthermore, we ported the codes to a gpu cluster and achieved near linear scalability. To our knowledge, these gpu implementations of the proposed algorithms are the fastest implementations. The schemes can be applied to simulate a three-dimensional photonic crystal with all 14 Bravais lattices. These highly efficient schemes and codes raise possibilities for large-scale and near real-time numerical simulations for novel physical discoveries and engineering applications of photonic crystals. (C) 2019 Elsevier B.V. All rights reserved.
暂无评论