for Computer Science Inaugurated
No one knows how the brain encodes and transfers information internally. Dr. Larry Manevitz is trying to find out. Manevitz, who is not a medical researcher, biologist, or chemist, is doing this computationally. A computer scientist, he heads the University's Haifa Interdisciplinary Research Center for Advanced Computer Science (HIACS). The Center is affiliated with the new Caesarea Edmond Benjamin de Rothschild Institute for Interdisciplinary Applications of Computer Science, which the University inaugurated at its 29th Board of Governors Meeting. Members of the Caesarea Foundation participated in the ceremony.
Prof. Martin Golumbic, who directs the new Institute, and Manevitz have already brought some of the world's foremost researchers in the field to the University to brainstorm aspects of computer science applications in the future. In Manevitz's case, it was an international workshop on "Computational/Mathematical Problems (and Solutions?) Arising from Neurophysiology."
According to Manevitz, Israel is the world center for the study of neurophysiology. The objective of the workshop was to see whether mathematical models can help the experimentalists—the neurophysiologists. Put more broadly, the question is, "How can one discipline's tools be made available to another discipline?"
The computer scientist illustrated the point. Magnetic resonance imaging (MRI) involves a mathematical reconstruction of the body without the use of radioactive material, unlike CAT scans. MRIs enable a surgeon to be more exact in planning an operation. They also allow researchers to see, for example, what part of the brain is active when a person thinks of certain things. In the latter case, the researcher asks: "How is the brain computing this? What algorithm is at work?" It is, he said, "a computational way of looking at things.
" "We do not understand," he continued, "how information is passed from one neuron to another." A neuron, which Manevitz describes as more complex than a computer, is a cell whose specialized processes make it the fundamental functional unit of the nervous system. The goal of the meeting, accordingly, was to look at problems and, more important, to try to look for right explanations, the right structural level. "We are "trying to decode essential mechanisms…trying to define what code the brain uses to transfer information."
Manevitz himself has been working with a senior Technion Medical School researcher on the question of how a system recognizes time; or in his words, "how time evolves out of a neurostructure."
The rate of concentration of ions in a cell was described back in the late 1950s. At the time, and indeed until very recently, researchers thought that a cell's timing was fixed in advance and the neuron would always act accordingly. Timing refers to the time scale between firings, or biochemical reactions. It is now seen that the behavior of neurons does change, but scientists are so far at a loss to explain why.
The UH computer scientist and his partner constructed a model of reactions that affect one another. They abstracted the model into a mathematical graph, the nodes of which represented the reactions. The rate of the reaction determines the waiting time. Using computer simulation, they found this timing to be non-monotonic, even if the reaction was simple. The cell reacts, but then stabilizes, and sets itself into a certain pattern until affected by other parameters. He and his colleague had found that by taking away details, the explanation could be simple. "The changes, which very, very small," Manevitz remarked, "looked mysterious, but in the end, when looked at in the right way, they were obvious.
" Asked about practical applications of this work, he immediately cited machine learning. "The strongest method of machine learning today," Manevitz said, "uses artificial neural methods. These can be adjusted to make machines learn. An example is the way robots can be made to learn to use their hands."
The scientist noted, however, that this method was not good for temporal learning, to learn a song or how to play tennis. There needs to be deeper knowledge of neurons in the temporal realm for that, he said.
As to the two-day workshop itself, Manevetz said that participants were "quite enthusiastic about coming, despite the Intifada and travel advisories." He pointed to the interdisciplinary interest that the event generated, with the Philosophy Department considering it important. The formal sessions themselves lasted from 9:30 in the morning to 7 in the evening, but speakers and audience stayed even later to discuss issues. "Very unusual for a conference," he concluded in evident satisfaction.