C11214: Event Related Neural Simulation ToolNovelty:
A software program that provides an environment for the simulation of large, integrate-and-fire neural networks. The tool allows for optimized processing of an event-driven algorithm to efficiently and accurately calculate simulated neural network activity.
Value Proposition:
Existing neural network simulation tools have been demonstrated to have an inadequate balance of performance and realism. The event-driven algorithm in this tool, developed by JHU Researchers, allows for accurate calculation of neuron spike times up to machine precision; solutions for systems neuroscience experiments modeled with this tool are exact solutions as opposed to approximations subject to discretization errors, as in many existing simulation tools. The efficiency of this tool allows for the simulation of large networks of millions of neurons, providing capabilities to realistically model neurophysiological experiments. Other advantages of this software program include:
• Programmed in JAVA: Can be run on a large number of computer architectures without the need for compilation or knowledge of special scripting language
• Parameter definition and simulated neuronal network documented in XML: Allows for easy user parameter input and solution access
Technical Details:
Johns Hopkins researchers have developed a user-ready neural network simulation software application that combines a highly-efficient, single-neuron model with a state-of-the-art, event-based neural communication protocol. Single-neuron models are linked by the simulator to create a network model based on user specifications of neural activity and connectivity. The simulator is designed for repeated simulation trials of neural networks that are much larger than can be implemented on one commodity computer; therefore, the network is mapped onto host machines in such a way as to optimize the number of trials that can be running simultaneously. This novel protocol optimizes simulations through fully parallel computation and minimization of communication events between machines.
Looking for Partners:
To develop and commercialize the technology as a neuroscience research tool for efficient computer simulation of biologically realistic neural networks.
Stage of Development:
Proof of Concept
Data Availability:
Prototype
Publications/Associated Cases:
Information Sciences and Systems (CISS), 2011 45th Annual Conference on, Issue Date: 23-25 March 2011.
