Previous Fora / 2003

• Permanent update
• Invitation
• Programme
• Participants
• Credits
• News & Views
• Budapest & Hungary
• Contact
• FAQ
• Satellite events
• Video archive

• Plenary Session
• Plenary Session II.
• Knowledge, environment, development
• Knowledge-based economy
• Closing Plenary Session

Speakers

		Professor Tamás Roska
		Professor of Information Technology Hungary

 

Dr. Tamás Roska received his M.A. in electrical engineering at the Budapest University of Technology and Economics in 1964. In 1967 he received his PhD, in 1973 became a candidate of technical sciences, and in 1982, a doctor for technical sciences. The Hungarian Academy of Sciences elected Dr. Roska as corresponding member in 1993 and as an ordinary member in 1998. Currently, he is research professor at the Computer and Automation Institute of the Hungarian Academy of Sciences, and dean of the Faculty of Information Technology at Pázmány Péter Catholic University. Since 1990, Dr. Roska has intensely cooperated with the research team of the academician and neurobiologist József Hámori. In the beginning of 1993, they collaborated to establish the post-graduate Centre for Neuromorph Information Technology, which is an interdisciplinary doctoral school incorporating the participation of four universities.

Following the emergence of chips and multimedia application, the third wave of information technology progress came as the development of mechanics, microelectronics and artificial intelligence. These research paths will best evolve in an interdisciplinary environment, and so the Faculty of IT's at Pázmány Péter Catholic University encourages interdisciplinary action.

In the beginning Tamás Roska's career, his work was related to the theory of non-linear electrical circuits and the computer aided design of electrical circuits. He was member of the scientific school established by the academician çrp?d Csurgay, where he investigated the issues of neural networks through the theoretical questions of non-linear electronics. Later, Dr. Roska deepened his knowledge concerning non-linear causality at the University of California at Berkeley.

When at UC Berkeley, in the mid-1970's, Dr. Roska met and collaborated with Professor L. O. Chua on the principles of the first analogical super computer, the Cellular Neural Network Computer (CNN). The principal of the CNN is based on the usage of traditional small computers with analogue and logical memory, which are interconnected to a regular geometrical grid. The small computers serve as cells and are connected locally over small distances. This construction is similar to the cellular automata of János Neumann, although the processing units of the CNN are not logical, but analogue and dynamic. This is a procedure that controls all cells of the analogical (analogue and logical) cellular computers in a way so they work non-linearly, dynamically and completely parallel. The CNN uses basic methods of proceeding waves and spatio-temporal elementary instructions.

When Frank Werblin, professor of neurobiology at UC Berkeley, joined the research activities of Roska and Chua, they began simulating and researching the visual recognition processes in the retina with the CNN. Interesting discoveries resulted from this work, namely that the analogical CNN computer can realize the combination of the retinas functioning (CNN bionic eye) and other sensory modalities (audition, touch).

The structure of the retina and that of a chip is similar. The chip is a multilayered plate in which one pixel of a picture is captured by one small computer. In the human and animal eye, neuron groups separately process and then simplify the picture on the retina. The brain receives dozens of information about different abstract features of the picture in the form of spatio-temporal activations. Werblins research points out that neurons of the retina form different layers, which are sensitive to different representations of the world. Different ganglion cell populations capture different layers and forward the information to higher visual centers. This process can be likened to simultaneously watching several blurred movies. For example, one cell population forwards the edges of a picture, another the bigger, more coherent parts, and the third cell population captures movements, or if a picture element disappears.

This high speed visual processing can be simulated on the CNN computer's artificial neural circuits. According to Dr. Roska, the significance of this system is to provide the ability to reveal pictures in this complex and noisy world.

 

 Resources on the Web

Analogical and Neural Computing Systems Laboratory:
http://lab.analogic.sztaki.hu/