JP7087252B2 - Arithmetic processing system and auxiliary equipment - Google Patents

Description

The present invention relates to an arithmetic processing system and an auxiliary device.

Attempts have been made to reproduce reactions such as double loops and triple loops that occur in the human brain judgment process by executing neural networks, deep learning, etc. on a computer device that is a Von Neumann computer.

On the other hand, a technique for estimating the state of human emotions and physiology without using a calculation method (that is, a binary number) using a huge number of switches like a Neumann computer such as a computer device has been proposed. (See, for example, Patent Document 1). That is, in Patent Document 1, the amount of homeostatic deviation in the subject is obtained by using the information indicating the physiological state of the subject and the information indicating the emotion of the subject and the activity of organs such as the brain, and the subject is obtained from the obtained deviation amount. Calculate the energy that acts on the emotions and activities of organs. Then, in Patent Document 1, the calculated energy is used to rotate a plurality of gears indicating each activity of the subject's emotions and organs in a virtual space, and the pathological condition of the subject is estimated from the state of rotation of each gear. do.

Japanese Unexamined Patent Publication No. 2015-128579

However, in a Von Neumann computer such as a computer device, only the probability and the derivative in the calculation means in the binary system can correspond, and the calculation is repeated infinitely. For this reason, the Von Neumann computer keeps the amount of calculation within the storage capacity of memory or the like without diverging the calculation (reaction such as double loop) that is repeatedly executed as in the judgment process of the human brain. It's difficult. That is, in the Von Neumann computer, the timing of reproducing the reaction such as double loop in the judgment process of the human brain and determining the intention and judgment (the timing of converging the reaction such as double loop (calculation executed repeatedly)). Is difficult to determine. In addition, it is difficult for a Von Neumann computer to determine various intentions and judgments according to one's mood, situation, situation, and the like. These are known as artificial intelligence frame problems.

An object of the present invention is to provide an arithmetic processing system and an auxiliary device capable of accurately reproducing information transmission in a neural circuit as compared with the conventional one.

An arithmetic processing system from one viewpoint includes a computer device, a plurality of gyros that rotate according to an auxiliary process among the arithmetic processes executed by the computer device, and a frame in which a plurality of gyros are arranged in a balanced manner in the initial state. , An auxiliary device having an adjusting unit arranged on the frame and adjusting the angle of each of the plurality of gyros and a detecting unit for detecting the movement of the frame, and the detecting unit assists the movement of the detected frame. Is output to the computer device as a result of.

The auxiliary device from another viewpoint is an auxiliary device that assists the arithmetic processing executed by the computer device, and a plurality of gyros that rotate according to the assisted processing and a plurality of gyros are arranged in a balanced manner in the initial state. It includes a frame, an adjustment unit that adjusts the angle of each of a plurality of gyros arranged in the frame, and a detection unit that detects the movement of the frame, and the detection unit is a result of processing that assists the movement of the detected frame. Output to a computer device.

It is a device that enables simultaneous calculation of digital and analog from continuous quantity signal input such as current, and function output by applying it to a balance type calculation method using these concepts and transistors and quantum computer chips. You may.

INDUSTRIAL APPLICABILITY The present invention can reproduce the transmission of information in a neural circuit with higher accuracy than in the past.

It is a figure which shows an example of the arithmetic processing system of 1st Embodiment. It is a figure which shows an example of the auxiliary apparatus shown in FIG. It is a figure which shows an example of a neural network. It is a figure which shows an example of the change of the membrane potential in a nerve cell. It is a figure which shows an example of the processing of the signal received from the auxiliary device in the computer device shown in FIG. 1. It is a figure which shows an example of the arithmetic processing in the arithmetic processing system shown in FIG. It is a figure which shows an example of the arithmetic processing system of 2nd Embodiment. It is a figure which shows an example of the computer apparatus shown in FIG. 7. It is a figure which shows an example of the slider device corresponding to the auxiliary device shown in FIG. (A) An example of the case where the auxiliary device is virtually realized by a program stored in the storage unit of the computer device, (b) an example of the case where the auxiliary device is virtually realized by a program stored in another computer device, ( c) It is a figure which shows an example of the case where the auxiliary device is virtually realized by the program stored in the server device.

Hereinafter, embodiments will be described with reference to the drawings.

<First Embodiment>
FIG. 1 shows an example of an arithmetic processing system according to the first embodiment.

The arithmetic processing system SYS shown in FIG. 1 includes a computer device 100 and an auxiliary device 200.

The computer device 100 is a Von Neumann computer having an arithmetic processing unit such as a CPU (Central Processing Unit) and a storage device such as a hard disk device. The arithmetic processing apparatus of the computer apparatus 100 executes a program stored in the storage apparatus by receiving an instruction of arithmetic processing by a user via, for example, an input device such as a keyboard included in the computer apparatus 100. The operation of the computer device 100 will be described with reference to FIGS. 5 and 6.

The auxiliary device 200 is a device that assists the arithmetic processing executed by the computer device 100.

FIG. 2 shows an example of the auxiliary device 200 shown in FIG. FIG. 2A shows an auxiliary device 200 viewed from the Y-axis direction (lateral direction). FIG. 2B shows the auxiliary device 200 as viewed from the Z-axis direction (upward direction).

The auxiliary device 200 includes a frame FR, a support column ST (ST (1), ST (2)), a bearing AX, a motor M1 (M1 (1), M1 (2)), a disk E (E (1), E (2)). )) And a hinge HG (HG (1), HG (2)). Further, the auxiliary device 200 includes a motor M2 (M2 (1), M2 (2)), a clamp CL (CL (1), CL (2)), a laser pointer PT, and an optical sensor LS.

The frame FR is rotatably arranged in the XZ plane on the bearing AX in the Y-axis direction arranged on the support column ST. Motors M1 are arranged at both ends of the frame FR via hinges HG such as hinges. Further, a motor M2 for changing the direction of the motor M1 and a clamp CL are arranged on the frame FR. Further, a laser pointer PT that emits laser light is arranged on the frame FR. As shown in FIG. 2, as the initial state of the auxiliary device 200, the motors M1, M2, the disk E, the hinge HG, the clamp CL, and the laser pointer PT are on the frame FR so that the balance of the frame FR is maintained. Is placed in.

The motor M1 is arranged with a disk E made of metal or the like, and operates as a gyro based on a control instruction from the computer device 100. Further, the motor M1 is connected to the motor M2 via a clamp CL made of metal or the like.

The motor M2 operates in response to a control instruction from the computer device 100, and changes the direction of each of the motors M1 (gyro) via the clamp CL. By changing the direction of each motor M1, the moment of each motor M1 changes, the equilibrium state of the frame FR is broken, and the frame FR moves in the XZ plane. The motor M2 and the clamp CL are examples of the adjusting unit.

The laser pointer PT is a laser light source and emits laser light toward the optical sensor LS. As a result, the laser pointer PT outputs the movement of the frame FR according to the operation of the motors M1 and M2 to the optical sensor LS.

The optical sensor LS is a CCD (Charge-Coupled Device), CMOS (Complementary Metal Oxide Semiconductor) image sensor, PSD (Position Sensitive Device), or the like, and detects the position of the laser beam emitted by the laser pointer PT. The optical sensor LS outputs the position of the detected laser beam, that is, the movement of the frame FR, to the computer device 100. The optical sensor LS and the laser pointer PT are examples of detection units.

The auxiliary device 200 shown in FIG. 2 is a seesaw-type or balance-type auxiliary device in which the frame FR operates as a seesaw (or balance) having two arms in which a gyro of a motor M1 and a disk E is arranged. There is. However, the auxiliary device 200 may be a seesaw-type or balance-type auxiliary device having a frame having two or more arms, and a gyro of a motor M1 and a disk E is arranged on each of the plurality of arms. In this case, the bearing AX is preferably a spherical bearing. In this case, the frame FR operates in three dimensions in the XYZ space, and the auxiliary device 200 detects the position of the laser beam of the laser pointer PT in the two-dimensional plane of the XY plane. Further, the optical sensor LS is preferably a CCD or CMOS image sensor in order to detect the position of the laser beam of the laser pointer PT on a two-dimensional plane.

Further, instead of the laser pointer PT, an acceleration sensor, an electronic gyro, or the like may be arranged on the frame FR to directly measure the tilt angle of the frame FR. As a result, the optical sensor LS can be omitted, and the auxiliary device 200 can be miniaturized.

FIG. 3 shows an example of a neural network. The neural network shown in FIG. 3 is a multi-layer perceptron having an input layer, an intermediate layer and an output layer. The input layer has three nodes, the middle layer has four nodes, and the output layer has one node. It is preferable that the number of nodes in the input layer, the intermediate layer, and the output layer is appropriately determined according to the field to which the neural network is applied, the required accuracy, and the like. Further, the neural network may have a plurality of intermediate layers.

In the neural network shown in FIG. 3, the information input to the three nodes of the input layer is output via the four nodes of the intermediate layer and one node of the output layer. In this case, the information input to each of the three nodes of the input layer is weighted according to the relationship with each of the four nodes of the intermediate layer and added to each of the four nodes of the intermediate layer. .. Then, the information added by each of the four nodes in the intermediate layer is weighted according to the relationship with one node in the output layer, and is added by one node in the output layer. The node of the output layer outputs the added information. Conventionally, the weighting between each node is determined to a predetermined value by machine learning or the like using teacher data or the like in which the relationship between the information input to the neural network and the information output from the neural network is clear in advance. Will be done. Alternatively, the weighting between each node is determined using a differentiable function such as a sigmoid function. However, the connections between synapses in nerve cells are not in a constant state like a predetermined value, nor are they in a state showing a fixed predetermined fluctuation such as a sigmoid function.

FIG. 4 shows an example of a change in membrane potential in a nerve cell. The vertical axis of FIG. 4 indicates the membrane potential, and the horizontal axis of FIG. 4 indicates time. As shown in FIG. 4, in the case of a membrane potential of about −70 mV, the nerve cell is at a resting potential in a state where it does not receive information from other nerve cells. Then, when an excitatory stimulus current is input, the nerve cell raises the membrane potential and starts depolarization. When the membrane potential exceeds the threshold (for example, a potential 15 mV or more higher than the resting potential) due to the progress of depolarization, the nerve cell becomes an action potential state, and the action potential rapidly changes from a negative potential to a positive potential. It changes and reaches a peak of about 30 mV. After that, the action potential changes from a positive potential to a negative potential due to repolarization, and becomes a hyperpolarized state in which the potential is swung back to a potential lower than the resting potential. Then, the nerve cell returns to the resting potential. The membrane potential shown in FIG. 4 shows the case of nerve cells, but the same waveform is also shown for the current flowing through synapses, muscles, and the like. From this, the waveform of the membrane potential shown in FIG. 4 is a basic shape of information transmission in the human body, and is different from the sigmoid function and the like.

That is, the essence of the brain is to process information input from the outside world and to plastically adjust nerve function accordingly. The efficiency of synaptic transmission is not constant in many parts of the brain and changes in response to stimuli. This phenomenon is called synaptic plasticity. One of the synaptic plasticities is spike-timing dependence plasticity (STDP), which changes depending on the firing time difference between presynaptic cells and postsynaptic cells. The STPD shows long-term potentiation (LTP), which shows a plastic change in which the neurotransmission efficiency increases over a long period of time, and a plastic change, in which the neurotransmission efficiency decreases over a long period of time. There is long-term depression (LTD). Synaptic connections change in response to the relationship between LTP and LTD, affecting memory and learning in the brain.

The inventor noticed that the swing back in the waveform of the membrane potential shown in FIG. 4 is similar to the movement of the top when the rotation stops. Therefore, the auxiliary device 200 has a gyro of the motor M1 (1) and the disk E (1) and a gyro of the motor M1 (2) and the disk E (2) at both ends of the frame FR, so that FIG. 4 shows. Reproduce the swingback of the waveform of the shown membrane potential. For example, in the case of the neural network shown in FIG. 3, the gyro of the motor M1 (1) and the disk E (1) corresponds to the node of the input layer or the node of the intermediate layer, and the motor M1 (2) and the disk E (2) correspond to each other. ) Corresponds to the node in the middle layer or the node in the output layer. In the case of a neural circuit, the gyro of the motor M1 (1) and the disk E (1) corresponds to the presynaptic cell, and the gyro of the motor M1 (2) and the disk E (2) corresponds to the postsynaptic cell. ..

The auxiliary device 200 deviates from the equilibrium state of the frame FR by changing the rotation speed of each motor M1 based on, for example, a control instruction from the computer device 100. However, due to the action of maintaining the equilibrium state on the frame FR, the auxiliary device 200 reproduces the same swing back as the waveform of the membrane potential shown in FIG. The auxiliary device 200 causes the optical sensor LS to detect the movement of the frame FR as the position of the laser beam of the laser pointer PT by the laser pointer PT emitting the laser beam to the optical sensor LS. Then, the auxiliary device 200 outputs a signal including the detected position information to the computer device 100. The computer device 100 uses the signal received from the auxiliary device 200 to determine the weighting between each node of the neural network shown in FIG. The operation of the computer device 100 will be described with reference to FIGS. 5 and 6.

By providing the motor M1 to the frame FR, the operation of the frame FR accompanying the operation of the motor M1 is limited to a predetermined range according to the length of the frame FR. Therefore, the action of maintaining an equilibrium state corresponds to human homeostasis.

Further, since the coma shows the precession movement, the auxiliary device 200 has a “fluctuation component” (or “chaos”) in the waveform shown in FIG. 4 due to the precession movement of the two gyros arranged in the frame FR. Outputs the superimposed waveform. Since the precession movement can be controlled according to the rotation speed of the gyro, the fluctuation component is controllable chaos. This makes it difficult for the auxiliary device 200 to reproduce the same waveform, unlike the case of the sigmoid function. Therefore, the arithmetic processing system SYS can accurately reproduce the transmission of information in a neural circuit such as a neural network as compared with the conventional one.

In addition, since homeostasis changes according to the health condition of a person, it is considered that the waveform of the membrane potential shown in FIG. 4 also changes. Therefore, in the auxiliary device 200, each motor M1 is provided at both ends of the frame FR by using a hinge HG and a clamp CL, and the angle is changed together with the rotation speed of each motor M1 based on a control instruction from the computer device 100. This makes it possible to reproduce waveforms in various human states.

Further, the arithmetic processing system SYS may determine the weighting among all the nodes of the neural network shown in FIG. 3 by using the waveform output by one auxiliary device 200. Alternatively, the arithmetic processing system SYS has a plurality of auxiliary devices 200 corresponding to each of the nodes of the neural network shown in FIG. 3, and the waveform output by each auxiliary device 200 is used between the corresponding nodes. The weighting may be determined.

FIG. 5 shows an example of processing of a signal received from the auxiliary device 200 in the computer device 100 shown in FIG. FIG. 5 shows some elements of the auxiliary device 200 shown in FIG. 2, and omits other elements. Further, FIG. 5 shows the frame FR in the equilibrium state by a broken line. Note that FIG. 5 describes the case of the neural network shown in FIG. 3, but the same applies to the case of a neural circuit or the like.

The computer device 100 converts the information on the position of the laser beam of the laser pointer PT included in the signal received from the auxiliary device 200 into an analog value of the waveform shown in FIG. 4 and a digital value of an N-ary number. For example, the computer device 100 acquires the position of the frame FR in the equilibrium state indicated by the broken line in advance, and obtains the frame from the difference between the position of the laser beam of the laser pointer PT of the signal received from the auxiliary device 200 and the position in the equilibrium state. The angle θ of the slope of FR is calculated. The angle θ is a positive angle in the positive Z-axis direction and a negative angle in the negative Z-axis direction when the positive X-axis direction is used as a reference (0 degree).

For example, as shown in FIG. 5, when the frame FR is tilted toward the positive X-axis side and the angle θ is a negative angle and the digital value is a binary number (N = 2), the computer device 100 is set to ". Obtain a digital value of "0". On the other hand, when the frame FR is tilted toward the negative X-axis side and the angle θ is a positive angle, the computer device 100 obtains a digital value of “1”. When the digital value is "0", the computer device 100 determines that the nodes of the neural network to which the auxiliary device 200 corresponds are at resting potential and information is not transmitted (OFF state). On the other hand, when the digital value is "1", the computer device 100 determines that the state (ON state) in which the information is transmitted by changing from the resting potential to the action potential between the nodes. That is, the auxiliary device 200 operates as a switch between the nodes, similarly to the conventional Von Neumann computer.

Further, as shown in FIG. 5, the computer device 100 is based on the movable range of the frame FR, that is, the irradiation range RA on the optical sensor LS by the laser beam of the laser pointer PT and the position of the laser beam of the laser pointer PT. , Obtain a continuous quantity of analog values indicating a ratio x: y. The obtained analog values x and y change with time indicate the trajectory of the movement of the frame FR, and are the waveforms shown in FIG. The computer device 100 stores the acquired digital values and analog values x and y in the storage device of the computer device 100. The ratio x: y is a state between the gyro between the motor M1 (1) and the disk E (1) and the gyro between the motor M1 (2) and the disk E (2), for example, a node of a neural network. It shows the state of information transmission between presynaptic cells and between presynaptic cells and postsynaptic cells. That is, when x> y, the information from the presynaptic cell is strongly output to the postsynaptic cell, and when x <y, the information from the presynaptic cell is weakly output to the postsynaptic cell.

The computer device 100 uses the obtained digital values and analog values x and y to determine the state and weighting of information transmission between the nodes corresponding to the auxiliary device 200. For example, when the digital value is "0", the computer device 100 is in an OFF state in which information is not transmitted between the nodes, so that the weighting between the nodes corresponding to the auxiliary device 200 is not performed or the value is determined to be arbitrary. ..

On the other hand, when the digital value is "1", the computer device 100 is in the ON state in which information is transmitted between the nodes, so that the analog values x and y (that is, the waveform of the membrane potential shown in FIG. 4) indicate. Determine the weighting according to the state between the nodes. For example, the computer device 100 increases the weighting between the nodes as the action potential swings from a negative potential to a positive potential and approaches the peak when the nodes are in a depolarized state. Further, when the action potential has passed the peak and is in a repolarized state, the computer device 100 reduces the weighting between the nodes as the action potential swings from a positive potential to a negative potential and approaches the resting potential.

Then, when the action potential is swung back to a negative potential higher than the resting potential in a state of hyperpolarization between the nodes, the computer device 100 determines the weighting between the nodes to be "0", a negative value, or the like. In this case, although the information is transmitted between the nodes of the neural network, the weighting is "0", a negative value, or the like, which indicates that the information is not transmitted between the nodes. This is because, for example, the presynaptic cell and the postsynaptic cell in a nerve cell change the state of transmission in response to a chemical reaction, so that the information input to the presynaptic cell is not necessarily transmitted to the postsynaptic cell. It corresponds to the case.

In addition, the state between presynaptic cells and postsynaptic cells undergo fluctuations associated with chemical reactions. This fluctuation corresponds to the fluctuation component contained in the analog values x and y. Therefore, the computer device 100 can reproduce the fluctuation of the state (that is, weighting) of information transmission between the nodes by using the fluctuation components included in the analog values x and y. Then, the arithmetic processing system SYS shown in FIG. 1 can reproduce the state between the nodes in the neural network in the same manner as the actual neural circuit as compared with the conventional one.

In FIG. 5, the boundary between the binary digital values “0” and “1” is the position where the laser pointer PT irradiates the laser beam when the frame FR shown by the broken line is in the equilibrium state, but it is arbitrary. It may be set to a position. As a result, the computer device 100 can set the states between the nodes in the neural network to various states, and can give the neural network individuality.

Further, the digital value is not limited to a binary number and may be an N-ary number. In the case of an N-ary number, the range of angles corresponding to the movable range of the frame FR is divided into N pieces, and the values from "0" to "N-1" are set in each of the divided N-points range of angles.

FIG. 6 shows an example of arithmetic processing in the arithmetic processing system SYS shown in FIG. The process shown in FIG. 6 is realized by the arithmetic processing device of the computer device 100 receiving an execution instruction of the arithmetic processing by the user via the input device of the computer device 100 and executing the program stored in the storage device. Will be done.

In the process shown in FIG. 6, the computer device 100 describes the case of the neural network shown in FIG. 3, but the same applies to the case of a neural circuit or the like.

In step S100, the computer device 100 determines the operating conditions of the rotation speed, the rotation direction, and the tilt angle of each motor M1 in the auxiliary device 200. The computer device 100 may accept the operating conditions of the auxiliary device 200 by operating the input device of the computer device 100 by the user. Then, the computer device 100 outputs the determined operating conditions to the auxiliary device 200.

In step S110, the computer device 100 receives from the auxiliary device 200 a signal including information on the position of the laser beam emitted by the laser pointer PT in the auxiliary device 200 operated under the operating conditions determined in step S100.

In step S120, the computer device 100 acquires digital values and analog values x and y as shown in FIG. 5 using the positions received in step S110. Then, the computer device 100 determines whether the nodes in the neural network are in the ON state or the OFF state by using the acquired digital values and analog values x and y, and also determines the weighting between the nodes. The computer device 100 executes the arithmetic processing of the neural network by using the state and the weighting between the determined nodes.

In step S130, the computer device 100 determines whether or not the end instruction has been received by the operation of the input device of the computer device 100 by the user. When the end instruction is received, the process of the computer device 100 proceeds to step S140. On the other hand, if the end instruction is not received, the process of the computer device 100 proceeds to step S100. In this case, in step S100, the computer device 100 has the analog values x and y acquired in step S120, that is, the rotational speed of the motor M1 in the auxiliary device 200 based on the operating state of the auxiliary device 200 and the maintenance of homeostasis. , The operating conditions of the rotation direction and the tilt angle may be changed. Alternatively, the computer device 100 may accept changes in the operating conditions of the auxiliary device 200 by the operation of the input device of the computer device 100 by the user.

In step S140, the computer device 100 outputs a signal including the end instruction received in step S130 to the auxiliary device 200. Then, the computer device 100 ends the arithmetic processing.

In step S200, the auxiliary device 200 receives a signal from the computer device 100 including the operating conditions determined in step S100. Then, the auxiliary device 200 drives the motors M1 and M2 under the received operating conditions.

In step S210, the auxiliary device 200 causes the optical sensor LS to detect the position of the laser beam of the laser pointer PT accompanying the movement of the frame FR due to the operating conditions received in step S200. The auxiliary device 200 outputs a signal including position information detected by the optical sensor LS to the computer device 100.

In step S220, the auxiliary device 200 determines whether or not a signal including an end instruction has been received from the computer device 100. Upon receiving the termination instruction, the auxiliary device 200 stops the motor M1 and terminates. On the other hand, if the end instruction is not received, the process of the auxiliary device 200 proceeds to step S200.

In the embodiment shown in FIGS. 1 to 6, the auxiliary device 200 arranges a gyro of the motor M1 and the disk E at each end of the frame FR, and the gyro is based on a control instruction from the computer device 100. Each is rotated at the specified rotation speed, rotation direction and tilt angle. Then, although the equilibrium state is broken in the frame FR, the auxiliary device 200 can reproduce the same swing back as the waveform of the membrane potential shown in FIG. 4 due to the action of maintaining the equilibrium state. As a result, the arithmetic processing system SYS can reproduce the transmission of information in a neural circuit such as a neural network with higher accuracy than in the past.

<Second Embodiment>
FIG. 7 shows an example of the arithmetic processing system of the second embodiment.

The arithmetic processing system SYS1 shown in FIG. 7 has a computer device 100A and an auxiliary device 200. Elements having the same or similar functions as those described with reference to FIG. 1 are designated by the same or similar reference numerals, and detailed description thereof will be omitted.

The computer device 100A is a Neumann-type computer having an arithmetic processing unit such as a CPU and a storage device such as a hard disk device, similarly to the computer device 100 shown in FIG. The arithmetic processing device of the computer device 100A executes a program stored in the storage device by receiving an instruction of arithmetic processing by a user via, for example, an input device such as a keyboard included in the computer device 100A.

FIG. 8 shows an example of the computer device 100A shown in FIG. 7. That is, the computer device 100A executes the program to execute the memory 110a, 110b, the first FFT unit 120, the second FFT unit 130, the third FFT unit 140, the fourth FFT unit 150, the first calculation unit 160, and the second calculation unit 170. And functions as a state determination unit 180. The first FFT unit 120, the second FFT unit 130, the third FFT unit 140, the fourth FFT unit 150, the first calculation unit 160, the second calculation unit 170, and the state determination unit 180 may be realized by a dedicated circuit.

The memories 110a and 110b are non-volatile memories and the like included in the storage device of the computer device 100A. Each of the memories 110a and 110b stores each of the analog values x and y output from the auxiliary device 200.

The first FFT unit 120 reads the latest analog value x data at a predetermined time interval from the analog value x received from the auxiliary device 200 from the memory 110a, and fast Fouriers the read latest analog value x data. Transform (FFT: Fast Fourier Transform) processing is executed. The first FFT unit 120 outputs the spectrum of the latest converted analog value x to the first arithmetic unit 160 as the current state of the gyro of the motor M1 (1) and the disk E (1) of the auxiliary device 200. It is preferable that the predetermined time interval is appropriately determined according to the calculation target such as the neural network shown in FIG. 3, nerve cells, and synapses.

The second FFT unit 130 reads the data of the past analog value x before the latest data of the analog value x from the memory 110a, and executes the FFT process on the read data of the past analog value x. The second FFT unit 130 outputs the spectrum of the converted past analog value x to the first calculation unit 160. It is preferable that the time interval of the data of the past analog value x processed by the second FFT unit 130 is longer than the predetermined time interval of the latest analog value x processed by the first FFT unit 120. As a result, the average state at the past analog value x in the gyro of the motor M1 (1) and the disk E (1) of the auxiliary device 200, that is, the standard of homeostasis can be obtained.

Similar to the first FFT unit 120, the third FFT unit 140 reads the latest analog value y data at a predetermined time interval from the analog value y received from the auxiliary device 200 from the memory 110b, and reads the analog value y. FFT processing is executed for the data. The third FFT unit 140 outputs the converted spectrum of the latest analog value y to the second arithmetic unit 170 as the current state of the gyro of the motor M1 (2) and the disk E (2) of the auxiliary device 200.

Like the second FFT unit 130, the fourth FFT unit 150 reads the data of the past analog value y before the data of the latest analog value y from the memory 110b, and FFT with respect to the read data of the past analog value y. Execute the process. The fourth FFT unit 150 transfers the converted past analog value y spectrum to the second arithmetic unit 170 as a reference for homeostasis in the gyro between the motor M1 (2) and the disk E (2) of the auxiliary device 200. Output.

The first calculation unit 160 obtains the difference between the spectrum of the latest analog value x received from the first FFT unit 120 and the spectrum of the past analog value x received from the second FFT unit 130 within a common frequency range. Then, the first calculation unit 160 responds to the rotation speed and the like set for the gyro of the motor M1 (1) and the disk E (1) of the auxiliary device 200 and the digital value received from the auxiliary device 200. Correct the difference in the spectrum of the analog value x. The first calculation unit 160 outputs the difference in the spectrum of the corrected analog value x to the state determination unit 180. The difference in the spectrum of the analog value x corresponds to the amount of deviation from the standard of homeostasis in the gyro between the motor M1 (1) and the disk E (1) of the auxiliary device 200.

The second calculation unit 170 obtains the difference between the spectrum of the latest analog value y received from the third FFT unit 140 and the spectrum of the past analog value y received from the fourth FFT unit 150 within a common frequency range. Then, the second calculation unit 170 responds to the rotation speed and the like set for the gyro of the motor M1 (2) and the disk E (2) of the auxiliary device 200 and the digital value received from the auxiliary device 200. Correct the difference in the spectrum of the analog value y. The second calculation unit 170 outputs the difference in the spectrum of the corrected analog value y to the state determination unit 180. The difference in the spectrum of the analog value y corresponds to the amount of deviation from the homeostasis reference in the gyro between the motor M1 (2) and the disk E (2) of the auxiliary device 200.

The state determination unit 180 corresponds to the auxiliary device 200 according to the difference in the spectrum of the analog value x corrected by the first calculation unit 160 and the difference in the spectrum of the analog value y corrected by the second calculation unit 170. Determines the weighting of states between the nodes to be used. Then, the computer device 100A executes the arithmetic processing of the neural network by using the weighting between the nodes determined by the state determination unit 180.

The arithmetic processing in the arithmetic processing system SYS1 shown in FIG. 7 is the same as the processing shown in FIG. 6, and detailed description thereof will be omitted.

In the embodiment shown in FIGS. 7 and 8, the auxiliary device 200 arranges a gyro of the motor M1 and the disk E at each end of the frame FR, and the gyro is based on a control instruction from the computer device 100A. Each is rotated at the specified rotation speed, rotation direction and tilt angle. Then, although the equilibrium state is broken in the frame FR, the auxiliary device 200 can reproduce the same swing back as the waveform of the membrane potential shown in FIG. 4 due to the action of maintaining the equilibrium state. As a result, the arithmetic processing system SYS1 can reproduce the transmission of information in a neural circuit such as a neural network with higher accuracy than in the past.

Here, in the second embodiment, the computer device 100A virtually realizes a device corresponding to the auxiliary device 200 on the program by performing each arithmetic processing. The virtually realized device has the same configuration as the auxiliary device 200, in other words, a seesaw type or a balance having two arms in which the gyros of the motor M1 and the disk E are arranged at both ends of the frame FR. It may be an auxiliary device having a mold configuration, or may be, for example, a slider device 240 shown in FIG. As shown in FIG. 9, the slider device 240 has, for example, a rail track 250 on a straight line and a slider 260 that moves along the rail track 250 (in the left-right direction in FIG. 9). In the slider device 240 in FIG. 9, the distance from one end to the slider corresponds to the above-mentioned analog value x, and the distance from the other end to the slider corresponds to the above-mentioned analog value y.

In the first embodiment and the second embodiment, the system SYS (or SYS1) having the computer device 100 (or 100A) and the auxiliary device 200 is given as an example, but the auxiliary device 200 is virtually realized by a program. If possible, it can be realized by the computer device alone without constructing a system having the auxiliary device 200 and the computer device 100 (or 100A). In this case, as shown in FIG. 10A, the storage unit (memory, hard disk, etc.) 310 of the computer device 100 (or 100A) may store the program 320 that virtually realizes the auxiliary device 200. ..

Further, as shown in FIGS. 10 (b) and 10 (c), the storage unit 340 of another computer device (supercomputer or the like) 330 connected to the computer device 100A and the storage unit 360 of the server device 350 are described above. It is also possible to store the program 320. In these cases, by executing the program in another computer device 330 or server device 350, the auxiliary device of FIG. 1 can be virtually realized in these devices.

The arithmetic processing systems SYS and SYS1 shown in FIGS. 1 and 7 have been shown to be applied to a neural circuit such as a neural network, but the present invention is not limited thereto. For example, arithmetic processing systems SYS and SYS1 are applied to robots and automobiles, call centers, entertainment and Internet and mobile communication terminal applications and services, applications to search systems, financial credit management systems and behavior prediction, companies, schools, government agencies, etc. Information analysis in police, military, information gathering activities, psychological analysis leading to false detection, application to organizational group management, management of mental health and behavior prediction of organizational members, researchers, employees, managers, etc. Application to systems that control systems, applications to systems that control the environment such as homes and offices, airplanes and spacecraft, means for knowing the state of mind and behavior prediction of family and friends, application to music and movie distribution, general Information search, information analysis management, application to information processing, application to customer sensitivity preference market analysis, application of systems that manage these on a network or stand-alone, application to elucidation of quantum behavior in quantum computers, etc. It is also possible. Furthermore, it can be applied to transistors, quantum computer chips, etc., and can also be applied to devices that enable simultaneous calculation of digital and analog and function output from continuous quantity signal input such as current.

The above detailed description will clarify the features and advantages of the embodiments. It is intended to extend the features and advantages of the embodiments as described above to the extent that the claims do not deviate from their spirit and scope of rights. Also, anyone with normal knowledge in the art should be able to easily come up with any improvements or changes. Therefore, there is no intention to limit the scope of the embodiments having the invention to the above-mentioned ones, and it is possible to rely on appropriate improvements and equivalents included in the scope disclosed in the embodiments.

100, 100A ... Computer device; 110a, 110b ... Memory; 120 ... 1st FFT section; 130 ... 2nd FFT section; 140 ... 3rd FFT section; 150 ... 4th FFT section; 160 ... 1st calculation section; 170 ... 2nd calculation section 180 ... state determination unit; 200 ... auxiliary device; AX ... bearing; CL (1), CL (2) ... clamp; E (1), E (2) ... disk E; FR ... frame; HG (1), HG (2) ... Hinge; LS ... Optical sensor; M1 (1), M1 (2), M2 (1), M2 (2) ... Motor; ST (1), ST (2) ... Support; PT ... Laser pointer SYS, SYS1 ... Arithmetic processing system

Claims (5)

Computer equipment and
A plurality of gyros that rotate according to an auxiliary process among the arithmetic processes executed by the computer device, a frame in which the plurality of gyros are arranged in equilibrium in an initial state, and a frame in which the plurality of gyros are arranged in the frame. An auxiliary device having an adjusting unit for adjusting each angle and a detecting unit for detecting the movement of the frame is provided.
The detection unit is an arithmetic processing system characterized in that the detected movement of the frame is output to the computer device as a result of the auxiliary processing. In the arithmetic processing system according to claim 1,
The detection unit is an arithmetic processing system characterized in that a digital value corresponding to the movement of the detected frame and an analog value including a fluctuation component are output to the computer device as the result. In the arithmetic processing system according to claim 1 or 2.
The assisting process is an arithmetic processing system characterized in that each of the plurality of gyros calculates a state between a plurality of corresponding objects. In the arithmetic processing system according to any one of claims 1 to 3.
An arithmetic processing system including a plurality of the auxiliary devices. It is an auxiliary device that assists the arithmetic processing executed by the computer device.
Multiple gyros that rotate according to the auxiliary processing,
A frame in which the plurality of gyros are arranged in equilibrium in the initial state,
An adjustment unit arranged on the frame and adjusting the angle of each of the plurality of gyros,
It is equipped with a detection unit that detects the movement of the frame.
The detection unit is an auxiliary device characterized in that the detected movement of the frame is output to the computer device as a result of the auxiliary processing.

Images