General Interests
I am interested in the computational properties of the brain. Specifically, how sensory information is processed and used to guide behavior. I am currently examining this phenomenon in the fish midbrain, specifically the optic tectum and the nucleus isthmi.
Computational Neuroscience
I model the fish (bluegill sunfish and goldfish) optic tectum and satilite nuclei. This intriguing circuit processes sensory input from a variety of modalities and participates in commanding various behaviors.
The brain computes using spiking neurons. Therefore, to realistically model neural computation, I use simplified spiking neuron models (leaky integrate-and-fire, adaptive exponential integrate-and-fire, spike response, compartmental...) to examine tectal processing. Although there are a variety of computational modeling packages available, I find that developing my own framework has not only helped me to better understand how neuronal networks function but also provided me with the flexibility necessary to model this unique system.
In my most recent efforts I have been experimenting with using a graphics card to process image data and convert grayscale images into spikes using a fragment shader implementation of a leaky integrate-and-fire neuron. To the right is an example of a movie frame (of my hand) before and after GPU process. The GPU convolved the image with a laplacian of gaussian kernel and high-pass filtered the image.
I hope to understand how visual input is processed and utilized to control avoidance behavior.
Electrophysiology
I am primarily interested in the physiological response of the midbrain nucleus isthmus. The activity of this nucleus signals the proximity of an approaching object. I am interested in discovering the limits of this computation by presenting stimuli of various sensory modalities (lateral line water currents, and visual) and of various sizes and speeds of approach. By understanding the limits of the system, I hope to expose how the nucleus isthmus and the reciprocally connected optic tectum analyzes near, salient objects.
Robotics
After developing neural models in simulation, the next step is to test the network performance in a real-world environment. To accomplish this, I use a iRobot Roomba robotics kit custom fit with a Jetway mini-itx motherboard as a test platform. The robot runs Ubuntu linux and a version of the Player robotics server (from the player-stage project).