Mark Cohen: A Simple Question – How Does the Brain Create the Mind?
For UCLA Professor-in-Residence Mark Cohen, the Hoffman2 Shared Research Cluster is more than just a useful resource. It’s an absolute necessity.
Cohen’s research focuses on answering a deceptively simple question: how does the brain create the mind? A pioneer in the development of ultra-fast MRI applications and functional MRI, Cohen uses modern methods of neuroimaging to explore the relationships between structure and function in the human brain. The images are integral to his research on topics such as cognition, perception, attention and drug abuse.
As you might imagine, a single image of the brain contains an immense amount of data. Now imagine thousands of such images, which are studied using modern data analysis approaches such as statistical pattern analysis and machine learning. That’s just a glimpse into Cohen’s research. As you can see, there’s nothing simplistic about it.
“We need the [Hoffman2] cluster just to go through data of this size,” he said.
UCLA’s Hoffman2 is the largest and most powerful cluster in the UC system. Faculty members (and their students) across all disciplines are welcome to join the cluster, which is hosted by the Institute for Digital Research and Education. For Cohen, who holds appointments in the departments of Psychiatry and Behavioral Science, Neurology, Psychology, Radiological Science and Bioengineering, the decision to opt in was a no-brainer.
“Buying these kinds of toys for our lab is impracticably expensive,” he said. “Apple had given me a small cluster that I housed in the [Academic Technology Services] area. I segued into the big cluster.”
For a study about seizures and epilepsy, Cohen’s trainees are analyzing post-seizure EEGs from 1,800 individuals. The goal is to detect statistical patterns in the data that might one day lead to the development of a predictive tool regarding epilepsy. Once again, he has turned to the Hoffman2 for his computational needs.
“We couldn’t do this at all without the cluster,” he said. “The EEG data sets are huge. The Hoffman2 takes a couple of days to go through all the data, using about 1,000 processors at a time.”
By comparison, Cohen estimated that a solitary high-end computer would take 40 times longer to get through the data, turning a two-day project into a nearly three-month project. It’s no wonder he considers the cluster a necessary resource.
“We’re not looking back,” he said. “It’s the right way to go.”