Организация нечеткого логического вывода на основе многоуровневого параллелизма

In this paper, a method for constructing hierarchical systems of fuzzy inference based on multilevel parallelism, in particular, second-level parallelism, is developed, theoretically substantiated and implemented. This approach is designed to accelerate the computation of hierarchical fuzzy systems having complex dependency graphs between blocks of fuzzy rules. The concept of multilevel parallelism is formulated and presented. The notion of the level of parallelism is introduced. The theorem is formulated and proved, and a method for theoretical estimation of the maximum possible acceleration for systems constructed on the basis of parallelism of the level n is developed. An approach to designing hierarchical fuzzy systems based on multilevel parallelism for NVIDIA graphics accelerators is developed. Using NVIDIA CUDA technology, an experimental software system was designed for hierarchical systems of fuzzy inference based on multilevel parallelism for systems having complex graphs of dependencies between blocks of fuzzy rules. Experimental estimates of the acceleration are obtained. Also, based on the developed method, theoretical estimates of the maximum possible acceleration are found. A comparative characteristic of the theoretical and experimental estimates of the acceleration of hierarchical fuzzy systems is given.

Abstract licence: CC BY-ND

2 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.