The Megopolis Resampler: Memory Coalesced Resampling on GPUs

TL;DR

The paper introduces Megopolis, a GPU-optimized resampling algorithm that enhances performance and stability in importance sampling tasks without tuning parameters, enabling real-time applications.

Contribution

It presents Megopolis, a novel memory-coalesced Metropolis-based resampling algorithm optimized for GPUs, improving speed and quality without parameter tuning.

Findings

Megopolis outperforms existing Metropolis variants in speed and quality.

The algorithm is numerically stable across experiments.

Open-source implementation facilitates future research and fair comparisons.

Abstract

The resampling process employed in widely used methods such as Importance Sampling (IS), with its adaptive extension (AIS), are used to solve challenging problems requiring approximate inference; for example, non-linear, non-Gaussian state estimation problems. However, the re-sampling process can be computationally prohibitive for practical problems with real-time requirements. We consider the problem of developing highly parallelisable resampling algorithms for massively parallel hardware architectures of modern graphics processing units (GPUs) to accomplish real-time performance. We develop a new variant of the Metropolis algorithm -- Megopolis -- that improves performance without requiring a tuning parameter or reducing resampling quality. The Megopolis algorithm is built upon exploiting the memory access patterns of modern GPU units to reduce the number of memory transactions without the need for tuning parameters. Extensive numerical experiments on GPU hardware demonstrate that the proposed Megopolis algorithm is numerically stable and outperforms the original Metropolis algorithm and its variants -- Metropolis-C1 and Metropolis-C2 -- in speed and quality metrics. Further, given the absence of open tools in this domain and facilitating fair comparisons in the future and supporting the signal processing community, we also open source the complete project, including a repository of source code with Megopolis and all other comparison methods.