We put forward an approach to the semantics of probabilistic programs centered on an action-based language equipped with a small-step operational semantics. This approach provides benefits in terms of both clarity and...
详细信息
ISBN:
(纸本)9783031505201;9783031505218
We put forward an approach to the semantics of probabilistic programs centered on an action-based language equipped with a small-step operational semantics. This approach provides benefits in terms of both clarity and effective implementation. Discrete and continuous distributions can be freely mixed, unbounded loops are allowed. In measure-theoretic terms, a product of Markov kernels is used to formalize the small-step operational semantics. This approach directly leads to an exact sampling algorithm that can be efficiently simd-parallelized. An observational semantics is also introduced based on a probability space of infinite sequences, along with a finite approximation theorem. Preliminary experiments with a proof-of-concept implementation based on TensorFlow show that our approach compares favourably to state-ofthe-art tools for probabilistic programming and inference.
The aim of the work reported here is to build a useful toolset for 3D model-based vision on an simd parallel machine, the AMT DAP. Included in the toolset are facilities for model specification, manipulation and rende...
详细信息
Digital holography (DH) systems have the potential to perform single-shot imaging through deep turbulence by incorporating emerging algorithms, such as model-based iterative reconstruction (MBIR), that jointly estimat...
详细信息
ISBN:
(纸本)9781510639003
Digital holography (DH) systems have the potential to perform single-shot imaging through deep turbulence by incorporating emerging algorithms, such as model-based iterative reconstruction (MBIR), that jointly estimate both the phase-errors and speckle-free image. However, the high computational cost of MBIR poses a challenge for use in practical applications. In this paper, we propose a method that makes MBIR feasible for real-time DH systems. Our method uses surrogate optimization techniques to simplify and speed up the reflectance and phase-error updates in MBIR. Further, our method accelerates computation of the surrogate-updates by leveraging cache-prefetching and simd vector processing units on each CPU core. We analyze the convergence and real CPU time of our method using simulated data sets, and demonstrate its dramatic speedup over the original MBIR approach.
Monte Carlo based methods such as path tracing are the only to do the physically correct simulations of global illumination. Due to its high capability to exploit acceleration structures and simd parallelism of modern...
详细信息
ISBN:
(纸本)9781424463916
Monte Carlo based methods such as path tracing are the only to do the physically correct simulations of global illumination. Due to its high capability to exploit acceleration structures and simd parallelism of modern processors, path tracing is becoming potential in the sense of real time. Basically path tracing generates images with heavy noise. Adaptive sampling is an interesting way to produce less noisy images. In this paper, we make use of both the intra-pixel information and the inter-pixel information to propose a new refinement criterion for adaptive sampling in path tracing to lower the Monte Carlo noise appeared in the generated image. For the intra-pixel information, we take advantage of the non-extensive Tsallis entropy as a homogeneity measurement of sample values within a pixel. For the inter-pixel information, we develop an innovative inter-pixel coherence measure based on the magnitude of spatial gradient and the impulsiveness of the distribution. Implementation results demonstrate that our novel method can perform significantly better than the previously typical ones.
General purpose graphical processing units were proven to be useful for accelerating computationally intensive algorithms. Their capability to perform massive parallel computing significantly improve performance of ma...
详细信息
General purpose graphical processing units were proven to be useful for accelerating computationally intensive algorithms. Their capability to perform massive parallel computing significantly improve performance of many algorithms. This thesis focuses on using graphical processors (GPUs) to accelerate algorithms based on adversarial search. We investigate whether or not the adversarial algorithms are suitable for single instruction multiple data (simd) type of parallelism, which GPU provides. Therefore, parallel versions of selected algorithms accelerated by GPU were implemented and compared with the algorithms running on CPU. Obtained results show significant speed improvement and proof the applicability of GPU technology in the domain of adversarial search algorithms.
暂无评论