Kevin Bender, Professor

Open (1) NeuroGPU – Neuronal simulation environment on graphical processing units.

Open. Apprentices needed for the fall semester. Please do NOT contact faculty before September 11th (the start of the 4th week of classes)! Enter your application on the web beginning August 16th. The deadline to apply is Tuesday, August 29th at 8 AM.

Neuronal modeling, from individual channels to neurons and circuits, is a major method that neuroscientists use to understand the brain. Currently we use electrophysiology recording with neuronal models in order to study the biophysical properties of the neuron we recorded from. This is done by using optimization algorithms to constrain the model’s parameters until its output mimics the recorded neuron. Running these simulations are computationally taxing. We therefore developed a platform using Graphical Processing Units (GPUs) to accelerate neuronal simulations. Our platform, NeuroGPU, is capable of running simulations 150 times faster than conventional methods. We are seeking a computer science apprentice student interested in helping us further develop NeuroGPU to its implementation phase. This will include adding a novel numerical algorithm, extending the environment to simulate neuronal circuits, and helping to make NeuroGPU more available to other research groups.

Each of the projects described below may result in a peer-reviewed paper. Furthermore each of these projects will advance the accessibility of the neuroscience community to low-cost, highly-detailed neuronal simulations. You would be joining a dynamic lab that more broadly studies mechanisms and function of neuromodulation. There would be opportunities to collaborate and learn to translate computer science skills to neuroscience.

We are seeking CS undergraduate students who would like to take part in this project. Possible tasks include:
1. Applying the SPIKE algorithm to a special case tri-diagonal solver.
2. Simulating different neuronal morphologies simultaneously.
3. Incorporating neuronal connections (synapses) for network simulations.
4. Profile and optimize the application’s performance.

Day-to-day supervisor for this project: Roy Ben-Shalom, Post-Doc

Qualifications: Required proficient skills in: C\C++ Matlab Desirable but not essential: Python CUDA

Weekly Hours: 9-12 hrs

Off-Campus Research Site: Can work remotely or at:
UCSF Mission Bay
675 Nelson Rising Ln, San Francisco, CA 94158

Related website: http://keck.ucsf.edu/neurograd/faculty/bender.html