Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userFully Parallel Particle Learning for GPGPUs and Other Parallel Devices
114
0
0.0
(
0
)
Added by
Kenichiro McAlinn
Publication date
2012
fields
Mathematical Statistics
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
We develop a novel parallel resampling algorithm for fully parallelized particle filters, which is designed with GPUs (graphics processing units) or similar parallel computing devices in mind. With our new algorithm, a full cycle of particle filtering (computing the value of the likelihood for each particle, constructing the cumulative distribution function (CDF) for resampling, resampling the particles with the CDF, and propagating new particles for the next cycle) can be executed in a massively and completely parallel manner. One of the advantages of our algorithm is that every single numerical computation or memory access related to the particle filtering is executed solely inside the GPU in parallel, and no data transfer between the GPUs device memory and the CPUs host memory occurs unless for further processing, so that it can circumvent the limited memory bandwidth between the GPU and the CPU. To demonstrate the advantage of our parallel algorithm, we conducted a Monte Carlo experiment in which we apply the parallel algorithm as well as conventional sequential algorithms for estimation of a simple state space model via particle learning, and compare them in terms of execution time. The results show that the parallel algorithm is far superior to the sequential algorithm.
rate research
Read More
Parallel tempering (PT) methods are a popular class of Markov chain Monte Carlo schemes used to sample complex high-dimensional probability distributions. They rely on a collection of $N$ interacting auxiliary chains targeting temper
Monte Carlo (MC) methods are widely used for Bayesian inference and optimization in statistics, signal processing and machine learning. A well-known class of MC methods are Markov Chain Monte Carlo (MCMC) algorithms. In order to foster better exploration of the state space, specially in high-dimensional applications, several schemes employing multiple parallel MCMC chains have been recently introduced. In this work, we describe a novel parallel interacting MCMC scheme, called {it orthogonal MCMC} (O-MCMC), where a set of vertical parallel MCMC chains share information using some horizontal MCMC techniques working on the entire population of current states. More specifically, the vertical chains are led by random-walk proposals, whereas the horizontal MCMC techniques employ independent proposals, thus allowing an efficient combination of global exploration and local approximation. The interaction is contained in these horizontal iterations. Within the analysis of different implementations of O-MCMC, novel schemes in order to reduce the overall computational cost of parallel multiple try Metropolis (MTM) chains are also presented. Furthermore, a modified version of O-MCMC for optimization is provided by considering parallel simulated annealing (SA) algorithms. Numerical results show the advantages of the proposed sampling scheme in terms of efficiency in the estimation, as well as robustness in terms of independence with respect to initial values and the choice of the parameters.
We explore an online learning reinforcement learning (RL) paradigm for optimizing parallel particle tracing performance in distributed-memory systems. Our method combines three novel components: (1) a workload donation model, (2) a high-order workload estimation model, and (3) a communication cost model, to optimize the performance of data-parallel particle tracing dynamically. First, we design an RL-based workload donation model. Our workload donation model monitors the workload of processes and creates RL agents to donate particles and data blocks from high-workload processes to low-workload processes to minimize the execution time. The agents learn the donation strategy on-the-fly based on reward and cost functions. The reward and cost functions are designed to consider the processes workload change and the data transfer cost for every donation action. Second, we propose an online workload estimation model, in order to help our RL model estimate the workload distribution of processes in future computations. Third, we design the communication cost model that considers both block and particle data exchange costs, helping the agents make effective decisions with minimized communication cost. We demonstrate that our algorithm adapts to different flow behaviors in large-scale fluid dynamics, ocean, and weather simulation data. Our algorithm improves parallel particle tracing performance in terms of parallel efficiency, load balance, and costs of I/O and communication for evaluations up to 16,384 processors.
In recent years, mesh subdivision---the process of forging smooth free-form surfaces from coarse polygonal meshes---has become an indispensable production instrument. Although subdivision performance is crucial during simulation, animation and rendering, state-of-the-art approaches still rely on serial implementations for complex parts of the subdivision process. Therefore, they often fail to harness the power of modern parallel devices, like the graphics processing unit (GPU), for large parts of the algorithm and must resort to time-consuming serial preprocessing. In this paper, we show that a complete parallelization of the subdivision process for modern architectures is possible. Building on sparse matrix linear algebra, we show how to structure the complete subdivision process into a sequence of algebra operations. By restructuring and grouping these operations, we adapt the process for different use cases, such as regular subdivision of dynamic meshes, uniform subdivision for immutable topology, and feature-adaptive subdivision for efficient rendering of animated models. As the same machinery is used for all use cases, identical subdivision results are achieved in all parts of the production pipeline. As a second contribution, we show how these linear algebra formulations can effectively be translated into efficient GPU kernels. Applying our strategies to $sqrt{3}$, Loop and Catmull-Clark subdivision shows significant speedups of our approach compared to state-of-the-art solutions, while we completely avoid serial preprocessing.
Bayesian Additive Regression Trees (BART) is a Bayesian approach to flexible non-linear regression which has been shown to be competitive with the best modern predictive methods such as those based on bagging and boosting. BART offers some advantages. For example, the stochastic search Markov Chain Monte Carlo (MCMC) algorithm can provide a more complete search of the model space and variation across MCMC draws can capture the level of uncertainty in the usual Bayesian way. The BART prior is robust in that reasonable results are typically obtained with a default prior specification. However, the publicly available implementation of the BART algorithm in the R package BayesTree is not fast enough to be considered interactive with over a thousand observations, and is unlikely to even run with 50,000 to 100,000 observations. In this paper we show how the BART algorithm may be modified and then computed using single program, multiple data (SPMD) parallel computation implemented using the Message Passing Interface (MPI) library. The approach scales nearly linearly in the number of processor cores, enabling the practitioner to perform statistical inference on massive datasets. Our approach can also handle datasets too massive to fit on any single data repository.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
