![]() |
ИСТИНА |
Войти в систему Регистрация |
ФНКЦ РР |
||
It is necessary to develop the specific theoretical background for supercomputing instead of using the idea of parallel processes emulation by sequential calculations. This idea has theoretical foundation presented by synchronous instructions of aggregating Turing machines. Hence, the parallel processes reduce to the sequential ones. Metaphorically speaking, it looks like the journey from Moscow and St. Petersburg to Guangzhou through Rio de Janeiro. This way of computation is not appropriate for Big Data supercomputing because an each unique specification or a singular rule requires a particular calculator. The non-electronic prototype of a serial computer was Analytical Engine invented by British scientists C. Babbage and A. Lovelace (1842), so any contemporary supercomputer is formally based on the principles of its functioning. The other principle of parallel computing that is not implemented in modern technology was offered earlier: in 1832 Russian engineer S.N. Korsakov published a description of the engine working as "homeoscope", i.e. as device processing common aspects in various logical ideas. "Calculations" were carried out through single-moment connections of particular features of a simple idea and simple components of a complex idea. The complex parallel-sequential Korsakov-Turing machine is proposed as a formal definition of supercomputing. The "connections" of Korsakov's machine and the "instructions" of the Turing machine are complementary to each other. The Korsakov-Turing machine is the basis of such methodological AI problems solutions as connective-symbolic representation, non-text context, deep learning, integrity of a frame, the description of a dynamic system. There are several suggestions of ways of technical implementation of this anti-reductionist supercomputer project.