Active solid-state devices (e.g. – transistors – solid-state diode – Field effect device – Having insulated electrode
Patent
1996-07-01
1998-10-06
Crane, Sara W.
Active solid-state devices (e.g., transistors, solid-state diode
Field effect device
Having insulated electrode
257316, 327427, 395 24, H01L 2976
Patent
active
058180818
DESCRIPTION:
BRIEF SUMMARY
SPECIFICATION
1. Technological Field
The present invention relates to a semiconductor device, and in particular, provides a highly functional semiconductor integrated circuit device for the realization of neural network computers (neuron computers). Furthermore, the present invention provides a highly functional semiconductor integrated circuit device for the realization of multivalent analog memory.
2. Background of the Invention
The pace of development in semiconductor integrated circuit technology has been startling, so that, for example, using dynamic memory as an example, memories from 4 megabits to 6 megabits are already in production, while ultra-high density memories having capacities of 64 megabits or more are being realized at the level of research. In 64 megabit memories, approximately 120 million MOS transistors are integrated on a silicon chip measuring at most 1 cm.sup.2. Such ultra LSI technology is applied not merely to memory circuits, but also to logical circuits; a variety of highly functional integrated circuits have been developed, beginning with CPUs of from 32 bits to 64 bits.
However, when such logical circuits are used to construct, for example, a computer employing a method in which digital signals, that is to say, signals having two values, "1" and "0", are used to conduct calculations, it is necessary to employ the Neumann method, in which instructions are executed one by one in accordance with a previously determined program. When such a method is employed, extremely high speed calculations are possible with respect to simple arithmetical calculations; however, an enormous amount of time is required for, for example, the calculations of pattern recognition or image processing or the like. Furthermore, such a method is highly disadvantageous for the type of data processing at which human beings excel, such as, for example, association, learning, and the like, and research is currently being conducted into various software technologies; however, the current state of affairs is such that rapid results have not been obtained. Here, a separate branch of research has been conducted in order to solve all these problems; in this research, the functioning of animal brains is being studied, and attempts are being made to develop computers which are capable of conducting calculations and processing which imitates such functions, that is to say, to develop neural circuit computers (neuron computers).
This research began in the 1940's; however, the field has become extremely active in recent years. This is the case because, in concert with the development of LSI technologies, it has become possible to realize such neuron computers in hardware.
However, there remain a number of problems with the realization of such neuron computers on LSI chips using present semiconductor LSI technology, and the goal of practical application has not yet been reached.
A discussion of the technological problems involved in the realization of such computers in terms of LSI follows.
The human brain has an extremely complex structure, and the functions thereof are of an extremely high level; however, the basic composition of the brain is extremely simple. That is to say, the brain comprises nerve cells having a calculating function, termed neurons, and nerve fibers, which transmit the results of such calculations to other neurons, thus serving, so to speak, as wiring.
In FIG. 14, a simplified model of the composition of the fundamental units of the brain is shown. References 1401a, 1401b, and 1401c indicate neurons, while references 1402a, 1402b, and 1402c indicate nerve fibers. 1403a, 1403b, and 1403c are termed synapse junctions; these apply, for example, a weighting w.sub.a to a signal transmitted from, for example, nerve fiber 1402a, and input this into neuron 1401a. Neuron 1401a calculates the linear sum of the signal strengths with are inputted thereinto, and when the total value thereof is in excess of a threshold value, the nerve cell becomes active, and outputs a signal to nerve fiber 1402b. If the tota
REFERENCES:
patent: 4258378 (1981-03-01), Wall
patent: 5258657 (1993-11-01), Shibata et al.
patent: 5280446 (1994-01-01), Ma et al.
patent: 5625591 (1997-04-01), Kato et al.
Kosaka Hideo
Ohmi Tadahiro
Shibata Tadashi
Yamashita Takeo
Crane Sara W.
Ohmi Tadahiro
Shibata Tadashi
LandOfFree
Semiconductor device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Semiconductor device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Semiconductor device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-81223