Meet a RC Researcher- Randy O'Reilly
Randy O’Reilly is a professor of psychology and neuroscience at CU Boulder. In his research, he uses computer models to simulate various parts of the brain. He is currently focusing on the posterior cortex, the site of visual recognition; dopamine models that demonstrate the way the brain learns from reward and punishment; and the prefrontal cortex, which enables the brain to stay on task and focus on the goals of the present moment.
O’Reilly develops these models using studies of animal behavior, usually rats. Once he can determine the connection between animal behavior and brain activity, he can build a computer model that simulates the way animals respond to different stimuli. Much of O’Reilly’s research involves testing the model against actual experimental results to verify its accuracy. Once developed, the model can be used to make novel predictions that have not been tested experimentally. O’Reilly’s model predicted that a certain drug would make Parkinson’s patients more prone to gambling, and experimental testing with real Parkinson’s patients verified his prediction. The models have the potential to be used in place of actual experimentation on the human brain, which could lead to exciting advancements in psychology and neuroscience.
Running simulations on models of the brain requires computing speed and power beyond that of a single computer. In order to run his computer simulations, O’Reilly received a grant from the Office of Naval Research to buy 36 nodes in Research Computing's "Condo" cluster. The cluster's many nodes, all working together, enable O’Reilly to run simulations on his model far more quickly than he could on a single computer. Research Computing configured the system, and now maintains and runs it for O’Reilly and his team. In his previous experience with clusters, O’Reilly maintained them himself, and he appreciates the time he saves by delegating that task to Research Computing. Of Research Computing’s role in his research, O’Reilly says, “It’s been great. The Condo model’s been fantastic for us, but it’s a huge amount of work, figuring out what system to get, configuring that system, and maintaining the system, and Research Computing does all that. It’s been a huge time saver for us.”