GPU technology as a platform for accelerating physiological systems modeling based on Laguerre-Volterra networks.

Area of Science:

• Biomedical Engineering
• Computational Biology
• Scientific Computing

Background:

• Mathematical modeling of biological systems is crucial but computationally intensive, often requiring nonlinear approaches with many parameters.
• General-Purpose Graphics Unit Processing (GPGPU) programming, using CUDA-enabled algorithms on GPU cards, offers significant computational speedups.
• Existing models may not fully leverage the parallel processing capabilities of GPUs for complex biological simulations.

Purpose of the Study:

• To develop a CUDA-enabled version of a nonlinear identification model for biological applications.
• To implement Laguerre-Volterra networks, combining basis expansions and polynomial-type networks, on a GPGPU platform.
• To evaluate the performance improvements of the GPU-based model compared to traditional implementations.

Main Methods:

• Developed a software implementation utilizing the GPGPU programming paradigm with CUDA.
• Integrated a nonlinear identification approach combining basis expansions and polynomial-type networks (Laguerre-Volterra networks).
• Executed model calculations on the GPU card of a host computer system to exploit inherent parallelism.

Main Results:

• The CUDA-enabled Laguerre-Volterra network model demonstrated performance improvements over the original MATLAB implementation.
• GPU acceleration significantly reduced computation time for nonlinear modeling in biological applications.
• The parallel nature of the modeling approach is well-suited for GPGPU acceleration.

Conclusions:

• The GPGPU implementation of Laguerre-Volterra networks provides a computationally efficient solution for complex biological modeling.
• This approach enhances the feasibility of detailed mathematical modeling in biomedical engineering and biology.
• Further applications of GPU acceleration in computational biology are warranted.