On results of David Marr Memorial Lab on Brain Reverse Engineering in years 2011-2013
Course duration: 6 h
Studies of the brain mechanisms are among the central problems of modern science. Indeed, it is widely believed that the brain hides very important and yet undiscovered “mysteries of life”. Due to the importance of the problem, it is permanently surrounded with intensive and hot discussions. Their range extends from demands to ban brain studies to statements that this problem can be never solved scientifically, or, at least, it cannot be solved on grounds of contemporary natural sciences. Our project takes an extreme stand among others (say, European Human Brain Project, or the recent USA initiative BRAIN). First, we suppose that the problem can be solved in a limited time (before/on January 1, 2016). Second, we ask for the work about 200 times less money than the mentioned super-projects. The methods, which we believe might bring success, are described at the websites rebrain.2045.ru and rebrain.2045.com.
In the lecture, we will give concrete initial results of our works. They were exposed earlier (Karandashev, 2013; Solovyeva, 2013; Dunin-Barkowski, 2013) at the Neuroscience session of the Barrow Neurological Institute (December 11, 2012, Phoenix, Arizona, USA), at the All-Russia Conference Nejroinformatika 2013 (January 25, 2013, Moscow, Russia), and at the conference Global Future 2045 (June 16, 2013, Lincoln Center, New York, USA).
Our works are connected with the data formats in neural systems. On grounds of the ideas on discrete and continuous attractors of activity dynamics in neural networks (Marr, 1971; Amari, 1974; Hopfield, 1982; Dunin-Barkowski, 1984; and later publications) and experimental data on inborn connections in cortex (Markram, 2011; and others) we analyzed several hypothetical types of networks with inborn neural attractors and explored methods of their application for vector quantization (pattern recognition) and Kohonen’s type self-organizing maps (Kohonen, 1982). We conclude that implementation of the above mentioned functions (they are definitely performed in real brains) with the help of neural attractors has significant advantages comparing to implementation of them using uncoupled neurons. Thus, we have tentatively revealed a candidate for a new basic principle of the work of real neuronal systems.
The work was supported by the Establishment “Russia 2045” and grant no. 13-07-01004 of the Russian Foundation for Basic Research.
Tutor
Prof. Witali Dunin-Barkowski
Country: Russia
Place of employment: Professor, Department of Neuroinfomatics, Scientific Research Institute for System Analysis
Spheres of researches: experimental and theoretical biophysics of different parts of nervous system- little brain, brainstem, Ammon's horn and biophysics of cells, the theory and analysis of natural and artificial neural systems, problems of general artificial intelligence.
Phone: +7 449 135 78 02
Email: This email address is being protected from spambots. You need JavaScript enabled to view it.