A hardware-based computational platform for Generalized Laguerre-Volterra MIMO model for neural activities

Area of Science:

• Computational Neuroscience
• Hardware Acceleration
• System Modeling

Background:

• Identifying time-varying neural dynamics is crucial for understanding brain function.
• Generalized Laguerre-Volterra MIMO systems are essential for modeling neural spike activities.
• Existing computational models can be slow and inefficient for real-time analysis.

Purpose of the Study:

• To propose a parallelized and pipelined FPGA architecture for modeling Generalized Laguerre-Volterra MIMO systems.
• To accelerate the identification of time-varying neural dynamics.
• To present ongoing work on a Self Reconfiguration Platform for advanced system modeling.

Main Methods:

• Design and implementation of a parallelized and pipelined architecture using Xilinx Virtex-6 FPGA.
• Development of a higher-level Self Reconfiguration Platform.
• Performance comparison with a C model on an Intel i7-860 Quad Core Processor.

Main Results:

• The proposed FPGA design achieves a data sample processing speed of 1.33 × 10^6/s.
• This represents a 3.1 × 10^3 times speedup compared to the C model.
• Initial test results demonstrate the feasibility and efficiency of the proposed architecture.

Conclusions:

• The developed FPGA architecture offers a significant speed improvement for modeling neural dynamics.
• The Self Reconfiguration Platform shows promise for advanced computational neuroscience applications.
• This approach enables faster and more efficient analysis of neural spike activities.