Dharmendra S. Modha

My Work and Thoughts.

Archives for 2006

The Swartz Foundation for Computational Neuroscience

By

From my perspective, cognitive computing is "neuroscientifically-inspired computing". The flip side of which is "computationallly-enabled (inspired) neuroscience".

Today, a major force shaping the field of computational neuroscience is Dr. Jerome (Jerry) Swartz.  Dr. Swartz is an incredible human being: a scientist (an inventor of over 200 patents), a technological innovator (winner of National Medal of Technology), a successful entrepreneur (founder of Symbol Technologies — recently acquired by Motorola), and a philanthropist.

In 1994, Dr. Swartz established the Swartz Foundation. The Foundation has established 3 centers at Cold Spring Harbor Laboratory, Columbia University and UC San Diego, and has partnered with the Sloan Foundation to establish five centers at Salk Institute, Cal Tech, NYU/Courant, Brandeis, and UC San Francisco. In effect, the Foundation has created a "Virtual Neuroscience Institute" that brings together the very best minds in the field together.

Recently, I had an opportunity to visit the Swartz Center for Computational Neuroscience at UC San Diego, and to spend time with its director, Dr. Scott Makeig. The center is equipped with state-of-the-art EEG labs and a high-performance compute cluster. Dr. Makeig’s personal interest is in applying Independent Component Analysis (a la Bell and Sejnowski) to EEG data. EEG data analysis allows us to understand how multiple brain areas interact dynamically in exhibiting a number of cognitive phenomena. He has led the development of the widely used EEGLAB software which "is an interactive Matlab toolbox for processing continuous and event-related EEG, MEG and other electrophysiological data using independent component analysis (ICA), time/frequency analysis, artifact rejection, and several modes of data visualization". Dr. Makeig has formed an impressive array of partnerships and projects, and has attracted top-notch collaborators. Very recently, EEGLAB was used by Professor Robert Knight and colleagues at UC Berkeley and UC San Francisco in their paper "High Gamma Power Is Phase-Locked to Theta Oscillations in Human Neocortex" that appeared in Science, 15 September 2006, 313: 1626-1628. Due to its noninvasive nature, EEG is likely to have a number of mainstream applications in brain-machine interfaces.

“A new look at the visual system” by Robert Galambos

By

Professor Robert Galambos is an iconic figure in Neurosciences. Along with six other faculty, he was instrumental in establishing the first academic neurosciences department in the world at UCSD. He was elected to the National Academy of Sciences in 1960. He is most widely known for his co-discovery of the phenomenon of echolocation in bats.

I had the incredible good fortune of listening to his talk at UCSD. At an age that is nearing 91, he is still amazingly sharp and full of energy. He described joint work with G. Juhasz’s group. A key idea is that eye fixates on different aspects of the visual scene at rate of 3 fixations/second. Each fixation causes retina to transmit an optic nerve signal that lasts 300ms. The amazing fact is that 300ms seems to be a constant. Different stimuli, different mental states of animal, different duration of stimuli, or different magnitude of stimuli do not seem to affect this number! This was established via experiments in rats.

In humans, he described the following experiment. A red LED is flashed followed by inter stimulus interval (ISI) followed by a green LED. Four cases are observed:

  1. When ISI is less than 60 ms, the subject sees a single flash of orange color.
  2. When ISI is in range 61-90ms, the subject sees a smaller flash followed by another larger flash — both are orange in color.
  3. When ISI is in range 91-290ms, the subject sees a smaller red flash followed by a larger green flash.
  4. When ISI is 300ms or more, the subject sees a red flash followed by a green flash of equal durations.

He argued how these observations can be used to demystify a whole range of psychological phenomena. For example, each saccade (the blur between two fixations) is 40 ms. Above data may explain why we never see it (so called saccadic suppression).

He proposed that the principle of retina is "create the briefest possible optic nerve volley that includes all information in a scene, and send it off to cortex at the highest possible rate."

At the very end, he raised several intriguing questions: (a) If optic nerve volley is 300ms, how do we perceive motion when pictures are flashed at the rate of 24 Hz? (b) Observe that when ISI is less than 90 ms, the subject perceives orange. Why? 

Nathaniel Rochester III (1919-2001)

By

Nathaniel Rochester

On August 31, 1955, J. McCarthy, Dartmouth College,  M. L. Minsky, Harvard University,  N. Rochester, I.B.M. Corporation, and C.E. Shannon, Bell Telephone Laboratories submitted “A PROPOSAL FOR THE DARTMOUTH SUMMER RESEARCH PROJECT ON ARTIFICIAL INTELLIGENCE” to the Rockfeller Foundation for support. This proposal is widely credited for coining the term “artificial intelligence”.

I was quite pleasantly surprised to find that an IBMer played a significant role in drafting a historically important scientific document. I was curious about him and his achievements.

In 1956, he published a classic article in which he and collegues simulated a network of neurons on IBM 701 and 704 calculators (yes, that is what computers were known as then!) to test Hebb’s theories. When he co-wrote the proposal, he was Manager of Information Research at IBM in Poughkeepsie, New York. His photograph was published in Time Magazine,  May 11, 1981. Here is his biography that was published by IBM Journal of Research and Development: 25th Anniversary Issue, 1981. (see pg. 842).

Mr. Rochester, an IBM Fellow, is currently working on the development of a portable personal computer. After working on radar in the Massachusetts Institute of Technology Radiation Laboratory and at GTLE, he joined IBM in 1948 in Poughkeepsie, New York. He was the architect of the test assembly and headed the architecture efforts for the tape processing machine and the IBM 701. He wrote the first symbolic assembly program, a predecessor of SAP. He managed the IBM 700 series engineering during the design of the IBM 703,704, 705, and the start of the 709. He joined IBM Research when it began in 1955 and directed work in computer theory and experimental computer design. In 1961, he joined the Data Systems Division to start a group that, among other things, designed IBM’s first two timesharing systems, QWIKTRAN and CPS, and accomplished the initial design of the PL/I language. His patents on the arithmetic unit of the 701 and on the variable-word-size architecture of the tape processing machine earned him an Outstanding Invention Award from IBM. He has served on government panels on air defense, antisubmarine warfare, cryptanalysis, and air traffic control. At M.I.T., Mr. Rochester received a B.S. in electrical engineering in 1941 and was elected a member of Tau Beta Pi and Sigma XI; in 1958 he was a Visiting Professor; and currently he is a Visiting Scientist there. Mr. Rochester is a Fellow of the Institute of Electrical and Electronics Engineers.

At the end, the 1955 proposal listed several people who would be interested in attending the event. Amongst these were six IBMers: John Backus, Alex Bernstein, W. L. Duda, Herbert Gelernter, Nathaniel Rochester, and David Sayre. WOW!

References:

  1. http://www-formal.stanford.edu/jmc/history/dartmouth.html
  2. Rochester, Nathaniel and H. L. Gelernter, “Intelligent Behavior in Problem-Solving Machines” IBM Journal of Research and Development 2:4 (October 1958): 336-345.
  3. Rochester, N., Holland, J.H., Haibt L.H. and Duda, W.L. (1956). Tests on a Cell Assembly Theory of the Action of the Brain Using a Large Digital Computer, IRE Transaction of Information Theory IT-2:80-93
  4. http://www.library.rochester.edu/index.cfm?page=3228

Machina Speculatrix

By

W. Grey Walter, a British physiologist, in 1950 and 1951, published two Scientific American articles describing his wonderful Machina Speculatrix (fondly known as Grey Walter’s "turtle").

Walter’s turtles represent a milestone in the conception and design of intelligent machines, and a second generation turtle is preserved at the Smithsonian Institution. Walter Grey’s turtles have inspired Rodney Brooks amongst many others in their quest for building intelligent machines. Walter Freeman has referred to Grey Walter as "Godfather of truly intelligent machines".

A turtle’s brain was built with two valves, two mechanical relays, and two capacitors. On the movement side, a turtle was equipped with three wheels and two motors. One motor powered the front wheel whereas other helped steer the front wheel. On the sense side, a turtle had two receptors. One internal "touch" receptor provided the ability to detect if it had hit an obstacle or an incline — in which case it would stop, back up, and try to navigate around the obstacle. Other external receptor was a photocell. A turtle searched for "moderate level illumination" of where the precise threshold changed with the amount of charge left in its batteries. It was housed in a brightly lit hutch which it found aversive when its batteries were charged and attractive when the batteries were running out. A turtle had an internal status light that went on when it was turning. If it saw its own light in a mirror or a light of another turtle, it would engage in a remarkable turning and backing behavior.

References:

  1. Walter J Freeman, W.Grey Walter: Biographical Essay. Encyclopedia of Cognitive Science(2003) 4: 537-539.
  2. Michael Gasperi’s Machina Speculatrix page.
  3. W. Grey Walter, The Living Brain, W. W. Norton, New York, 1963.
  4. W. Grey Walter, An Imitation of Life, Scientific American, pp. 42-45. May 1950.
  5. W. Grey Walter, A Machine that Learns, Scientific American, pp. 60-63, Aug 1951.

Fastest supercomputer to be built by IBM

By

"Computer giant IBM will build the world’s most powerful supercomputer at a US government laboratory."

"The machine, codenamed Roadrunner, could be four times more potent than the current fastest machine, BlueGene/L, also built by IBM."

"The new computer is a "hybrid" design, using both conventional supercomputer processors and the new "cell" chip designed for Sony’s PlayStation 3."

"Roadrunner will be installed at the Los Alamos National Laboratory, New Mexico."  

"The new machine will be able to achieve "petaflop speeds," said IBM."

See original article –>

I wonder what new avenues in computational neuroscience can be opened with such computational power.  On a personal note, I would love to play on this machine!