Jolley Hall, Room 309
"The Development of Scalable Simulator for Spiking Neural Network"
Jae Sang Ha
Thesis Adviser: Shantanu Chakrabartty
A neural network simulator for Spiking Neural Network (SNN) is a useful research tool to model brain functions with a computer. With this tool, different parameters can be explored easily compared to using real brain. For several decades, researchers have developed many software packages and simulators to accelerate research in computational neuroscience. However, despite their advantages, different neural simulators possess different limitations, such as flexibility of choosing a neuron. In this paper, I demonstrate an efficient and scalable spiking neural simulator that is based on growth transform neurons and runs on a single machine. The growth transform neuron model’s update is based on matrix-vector multiplication, which is optimized using external libraries—BLAS and sparseBLAS. Using sparseBLAS, the scability of the simulator was optimized with sparse representation of matix. This simulator can simulate up to 1 million neurons and is flexible with neuron model changes behind the simulator. With different parameters, a researcher can design a variety of network topology, visualize a coupling matrix and simulate a designed network. I intend to make this simulator widely available to the public so that researchers can have an easy access to large-scale simulations.