Oscillating neural network circuit arrangement using gated diode based oscillator

Abstract

To provide a vibration neural network circuit device for solving a graphcoloringproblem by allowing a plurality of oscillators coupled by a capacitor or a resistor to detect an edge or to indicate colors of at least two graphs for an input graph.SOLUTION: In the vibration neural network circuit device 700 using a gated diode-based oscillator, phase differences between the voltage VIN1 and the voltage VIN2 may occur due to time differences between the voltage VOUT1 and the voltage VOUT2 input to at least two gated diodes constituting at least two oscillators 701 and 702, and thus colors of at least two graphs with respect to input graphs of a graph coloring problem may be represented based on the phase differences over time.SELECTED DRAWING: Figure 7A

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of priority based on Korean Patent Application No. 10-2024-0106829, filed on August 9, 2024, the entire contents of which are incorporated herein by reference.

[0002] The present invention relates to an oscillatory neural network circuit device using a gated diode-based oscillator, and more specifically to a technology for realizing an oscillator using a gated diode and a resistor that repeatedly generate and eliminate a positive feedback loop, and for solving combinatorial optimization problems with an image processor based on the coupling and synchronization characteristics between the oscillators. [Background technology]

[0003] In conventional von Neumann-based computers, the processor and memory are physically separated, causing a bottleneck in the transmission of data signals when processing large amounts of data.

[0004] Although various technologies have been researched to overcome this problem, various problems, such as a decrease in performance due to structural issues, remain unresolved.

[0005] Meanwhile, research into new computing architectures, such as artificial neural networks that mimic the human brain, is accelerating.

[0006] Among these, oscillatory neural networks are a technology that directly mimics the cognitive system that operates through neuronal oscillations (brain waves) and their interactions, using oscillators and couplings.

[0007] Oscillatory neural networks can be applied to various fields, and in particular, they are attracting more attention because they can easily realize combinatorial optimization problem (COP) solvers and Ising machines.

[0008] Conventional oscillators implemented using CMOS circuits, such as ring oscillators or inductor-capacitor (LC) circuits, have high reliability, but their area and energy efficiency are low, making it difficult to implement large-scale artificial neural networks.

[0009] On the other hand, conventional oscillator technologies using metal-insulator transition devices (MITs), resistive switching devices, and 2D materials, which have simple structures and high integration, cannot be applied to existing CMOS processes, and therefore suffer from poor device uniformity and stability and can be unstable due to the device's operating principle.

[0010] Therefore, there is a need to develop a highly integrated oscillator composed of gated diodes, which can be called silicon nanosheet gated diodes, that can be applied to CMOS processes and operate in a positive feedback loop with high reliability, as well as oscillatory neural network technology using this. Summary of the Invention [Problem to be solved by the invention]

[0011] The present invention aims to provide an oscillatory neural network circuit device that uses gated diodes and resistors that repeatedly generate and eliminate a positive feedback loop to realize an oscillator, and that solves combinatorial optimization problems with an image processor based on the coupling and synchronization characteristics between oscillators.

[0012] The present invention aims to provide an oscillator configured with a gated diode that overcomes the limitations on processing speed and performance degradation caused by the separation of memory and processor, and an oscillatory neural network circuit device using the same.

[0013] The present invention aims to provide an oscillatory neural network circuit device that solves the graph coloring problem by using multiple oscillators coupled by capacitors or resistors to detect edges or indicate at least two graph colors for an input graph.

[0014] The present invention aims to realize an oscillatory neural network circuit device that can be used as a solver for combinatorial optimization problems (COPs) with an image processor by using a highly integrated oscillator that is highly reliable and can be applied to existing CMOS processes. [Means for solving the problem]

[0015] The oscillatory neural network circuit device using a gated diode-based oscillator according to one embodiment of the present invention has a p + -in + The diode structure is located + -in + The oscillator includes a gate insulating film on an intrinsic region of the diode structure, and a gated diode having two gate terminals on the gate insulating film, and a resistor connected in series to the gated diode, and the oscillator is configured to generate a voltage such that the intrinsic region is n * Channel region and p *As the channel region is changed to one of the channel regions, a potential barrier is formed in the intrinsic region by electrostatic doping, and a positive feedback loop related to the formed potential barrier is repeatedly generated and annihilated due to fluctuations in electron injection based on the injection of majority carriers, thereby outputting an output voltage that oscillates with a certain period. The oscillator is composed of at least two oscillators, and output terminals corresponding to portions where resistors are connected to at least two gated diodes are coupled to each other by either a capacitor or a resistor, so that a phase difference with respect to the output voltage occurs depending on the time difference of the input voltages input to the at least two gated diodes, and at least two graph colors can be displayed for the input graph based on the phase difference with time.

[0016] The gated diode has a p-type gate electrode between a first gate terminal of the two gate terminals and the drain terminal. + When a positive voltage is applied to the first gate terminal, the region under the first gate terminal in the intrinsic region becomes n * a second gate terminal of the two gate terminals and the source terminal; + When a negative voltage is applied to the second gate terminal, the region under the second gate terminal in the intrinsic region becomes p * It may be included as a channel region.

[0017] When the resistor is connected to the drain terminal, the gated diode is + The height of the potential barrier of holes injected from the drain region is controlled, and the amplitude and frequency of the oscillator can be changed by moving the latch-up voltage in the positive direction, which generates a positive feedback loop proportional to the first gate voltage.

[0018] When the resistor is connected to the source terminal, the gated diode is turned on by a second gate voltage input to the second gate terminal. + The height of the potential barrier of electrons injected from the source region is controlled, and the amplitude and frequency of the oscillator can be changed by moving the latch-up voltage in the negative direction in proportion to the second gate voltage.

[0019] The oscillatory neural network circuit device may determine a first output pixel value based on the output voltage output out of phase, which exhibits opposite movement as time passes, based on a time difference between when a first input voltage and a second input voltage are applied to each of the at least two gated diodes and an influence is generated between the at least two gated diodes based on the capacitor, or may determine a second output pixel value opposite to the first output pixel value based on the output voltage output in phase, in which the first input voltage and the second input voltage are applied at the same time so that the time difference does not occur.

[0020] In the vibration neural network circuit device, each of the at least two gated diodes can output the output voltage of a vertical edge component when horizontally coupled via the capacitor, and can output the output voltage of a horizontal edge component when vertically coupled via the capacitor.

[0021] In the case where the oscillatory neural network circuit device is configured with four oscillators, output terminals of a first oscillator, a second oscillator, a third oscillator, and a fourth oscillator constituting the four oscillators are coupled horizontally or vertically via the capacitors and diagonally coupled via resistors, and the output phases at the output terminals coupled via the resistors are the same, and the phase difference may occur with respect to the phase coupled via the capacitors.

[0022] In the vibration neural network circuit device, the first oscillator is horizontally coupled to the second oscillator via the capacitor and vertically coupled to the third oscillator to output an output voltage having a phase difference with respect to output voltages of the second oscillator and the third oscillator, and is diagonally coupled to the fourth oscillator via the resistor to output an output voltage having the same phase; the second oscillator is horizontally coupled to the first oscillator via the capacitor and vertically coupled to the fourth oscillator to output an output voltage having a phase difference with respect to output voltages of the first oscillator and the fourth oscillator, and is diagonally coupled to the third oscillator via the resistor to output an output voltage having the same phase.

[0023] The vibration neural network circuit device can display the colors of the at least two graphs, each of which is composed of a first color corresponding to the output voltages of the same phase and a second color corresponding to the output voltages of the phases where the phase difference exists.

[0024] The oscillatory neural network circuit device considers the number of nodes constituting the input graph, the connection structure between the nodes, and the colors of the nodes, and determines a color separation criterion for the phase difference through a time phase change of the oscillator based on the phase difference using an oscillator corresponding to the number of nodes, and can display the colors of the at least two graphs with colors separated based on the color separation criterion. [Effects of the Invention]

[0025] The present invention provides an oscillatory neural network circuit device that uses gated diodes and resistors that repeatedly generate and eliminate a positive feedback loop to realize an oscillator, and solves combinatorial optimization problems with an image processor based on the coupling and synchronization characteristics between oscillators.

[0026] The present invention can provide an oscillator configured with a gated diode that overcomes the limitations on processing speed and performance degradation caused by the separation of memory and processor, and an oscillatory neural network circuit device using the same.

[0027] The present invention can provide an oscillatory neural network circuit device that solves the graph coloring problem by using multiple oscillators coupled by capacitors or resistors to detect edges or indicate at least two graph colors for an input graph.

[0028] The present invention can realize an oscillatory neural network circuit device that can be used as a solver for combinatorial optimization problems (COPs) with an image processor by using a highly integrated oscillator that is highly reliable and can be applied to existing CMOS processes. [Brief explanation of the drawings]

[0029] [Figure 1A] 1A and 1B are diagrams illustrating the structure and switching characteristics of a gated diode that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. [Figure 1B] 1A and 1B are diagrams illustrating the structure and switching characteristics of a gated diode that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. [Figure 1C] 1A and 1B are diagrams illustrating the structure and switching characteristics of a gated diode that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. [Figure 2A] 1A and 1B are diagrams illustrating the structure and operating characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. [Figure 2B] 1A and 1B are diagrams illustrating the structure and operating characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. [Figure 2C]1A and 1B are diagrams illustrating the structure and operating characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. [Figure 3] 1A and 1B are diagrams illustrating changes in the characteristics of an oscillator that constitutes an oscillation neural network circuit device according to an embodiment of the present invention. [Figure 4A] 1 is a diagram illustrating tuning characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. FIG. [Figure 4B] 1 is a diagram illustrating tuning characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention. FIG. [Figure 5] 10A and 10B are diagrams illustrating changes in the characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention when the oscillator uses electrons as majority carriers. [Figure 6A] 10A and 10B are diagrams illustrating tuning characteristics of an oscillator constituting an oscillatory neural network circuit device according to an embodiment of the present invention when the oscillator uses electrons as majority carriers. [Figure 6B] 10A and 10B are diagrams illustrating tuning characteristics of an oscillator constituting an oscillatory neural network circuit device according to an embodiment of the present invention when the oscillator uses electrons as majority carriers. [Figure 7A] 1A and 1B are diagrams illustrating an oscillatory neural network circuit device according to an embodiment of the present invention and an edge detection simulation using the same; [Figure 7B] 1A and 1B are diagrams illustrating an oscillatory neural network circuit device according to an embodiment of the present invention and an edge detection simulation using the same; [Figure 7C] 1A and 1B are diagrams illustrating an oscillatory neural network circuit device according to an embodiment of the present invention and an edge detection simulation using the same; [Figure 8A] 1A and 1B are diagrams illustrating the structure and vibration characteristics of a vibration neural network circuit device according to an embodiment of the present invention; [Figure 8B]1A and 1B are diagrams illustrating the structure and vibration characteristics of a vibration neural network circuit device according to an embodiment of the present invention; [Figure 9A] 10 is a diagram illustrating a configuration for implementing the color of a graph in an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention; FIG. [Figure 9B] 10 is a diagram illustrating a configuration for implementing the color of a graph in an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention; FIG. [Figure 9C] 10 is a diagram illustrating a configuration for implementing the color of a graph in an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention; FIG. [Figure 9D] 10 is a diagram illustrating a configuration for implementing the color of a graph in an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention; FIG. [Figure 10] 10 is a diagram illustrating a configuration for implementing the color of a graph in an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention; FIG. DETAILED DESCRIPTION OF THE INVENTION

[0030] Various embodiments of the present specification will now be described with reference to the accompanying drawings.

[0031] The examples and the terms used therein are not intended to limit the technology described herein to a particular embodiment, but should be understood to include various modifications, equivalents, and / or alternatives to the examples.

[0032] In the following description of various embodiments, if it is determined that a detailed description of related publicly known functions or configurations may obscure the gist of the invention, the detailed description will be omitted.

[0033] The terms used below are defined in consideration of the functions in various embodiments, and may vary depending on the intentions or practices of users or operators. Therefore, the definitions should be based on the overall content of this specification.

[0034] In connection with the description of the drawings, like reference numerals may be used for like components.

[0035] A singular expression can include a plural expression unless the context clearly indicates otherwise.

[0036] As used herein, phrases such as "A or B" or "at least one of A and / or B" can include all possible combinations of the items listed together.

[0037] Expressions such as "first," "second," "initial," or "second" may modify the components in question regardless of order or importance, and are used only to distinguish one component from other components, and do not limit the components in question.

[0038] When a (e.g., first) component is referred to as being "(functionally or communicatively) coupled" or "connected" to another (e.g., second) component, the component may be directly coupled to the other component or may be coupled through another component (e.g., third component).

[0039] As used herein, "configured to" may be used interchangeably with, for example, hardware or software terms such as "suitable for," "capable of," "modified to," "made to," "capable of," or "designed to," depending on the context.

[0040] In some circumstances, the phrase "a device configured to" can mean that the device is "capable of" in conjunction with other devices or components.

[0041] For example, a phrase such as "a processor configured (or set) to perform A, B, and C" may refer to a processor that is dedicated to performing those operations (e.g., an embedded processor), or to a general-purpose processor (e.g., a CPU or application processor) that can perform those operations by executing one or more software programs stored in a memory device.

[0042] Also, the term "or" means an inclusive or rather than an exclusive or.

[0043] That is, unless otherwise stated or clear from the context, the phrase "x uses a or b" means any one of the natural inclusive permutations.

[0044] As used below, terms such as "...module" and "...device" refer to a unit that processes at least one function or operation, which may be embodied in hardware, software, or a combination of hardware and software.

[0045] 1A to 1C are diagrams illustrating the structure and switching characteristics of a gated diode that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0046] FIG. 1A illustrates the structure of a gated diode that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0047] Referring to FIG. 1A, a gated diode 100 constituting an oscillatory neural network circuit device according to an embodiment of the present invention is p + -in +It has a doping structure and two gate electrodes are located above the intrinsic channel region.

[0048] The gated diode 100 according to one embodiment of the present invention has two gate voltages (V G1 ,V G2 ) electrostatic doping changes the energy band structure of the channel region to p + -n * -p * -n + By varying the input voltage, a positive feedback loop is created and destroyed, thereby exhibiting latch-up and latch-down characteristics.

[0049] More specifically, the gated diode 100 according to the embodiment of the present invention has a p-type gate between a drain terminal and a source terminal. + -in + A diode structure is located.

[0050] The gated diode 100 is p + -in + A gate insulating film is located on the intrinsic region of the diode structure, and two gate terminals are located on the gate insulating film.

[0051] The gated diode 100 according to the embodiment of the present invention is G1 and the second gate voltage V G2 In response to the application of , a positive feedback loop is created and destroyed over the intrinsic region.

[0052] The supply voltage V DS is applied to the source terminal, and the source terminal can be connected to the ground terminal GND.

[0053] FIG. 1B illustrates an optical image of a gated diode that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0054] Referring to FIG. 1B, an optical image 110 shows a gated diode with a polysilicon gate electrode and an oxide silicon gate oxide layer on a buried oxide. + -in + It is surrounded by a silicon nanosheet (NS).

[0055] FIG. 1C illustrates the switching characteristics of a gated diode that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0056] Referring to FIG. 1C, a graph 120 shows the relationship between the first gate voltage V and the gated diode constituting the oscillatory neural network circuit device according to one embodiment of the present invention. G1 is applied as a positive voltage, and the second gate voltage V G2 When is applied as a negative voltage, I DS This shows that the characteristics show abrupt switching characteristics.

[0057] That is, the gated diode according to one embodiment of the present invention exhibits latch-up and latch-down characteristics due to the generation and disappearance of a positive feedback loop. 2A to 2C are diagrams illustrating the structure and operating characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0058] FIG. 2A illustrates the structure of an oscillator that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0059] Referring to FIG. 2A, an oscillator 200 constituting an oscillatory neural network circuit device according to an embodiment of the present invention has a p + -in + The diode structure is located + -in +The diode structure has a gate insulating film on the intrinsic region of the diode structure, and a gated diode with two gate terminals located on the gate insulating film as the base, with a resistor connected in series to the drain terminal of the gated diode.

[0060] For example, a gated diode has two gate terminals and a p-type transistor between the first gate terminal and the drain terminal. + When a positive voltage is applied to the first gate terminal, the region under the first gate terminal in the intrinsic region becomes n * It may be included as a channel region.

[0061] The gated diode has two gate terminals, and the second gate terminal is connected to the source terminal. + When a negative voltage is applied to the second gate terminal, the region under the second gate terminal in the intrinsic region becomes p * It may be included as a channel region.

[0062] The oscillator 200 is configured such that the gated diode is connected to the n * Channel region and p * While one of the channel regions is being changed into a channel region, a potential barrier is formed in the intrinsic region by electrostatic doping.

[0063] Furthermore, the oscillator 200 can output an output voltage that oscillates with a constant period by repeatedly generating and annihilating a positive feedback loop associated with the potential barrier formed by fluctuations in electron injection based on the injection of majority carriers.

[0064] FIG. 2B illustrates vibration characteristics based on the operation of an oscillator that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0065] Referring to FIG. 2B, graph 210 shows the output voltage over time, which is related to the vibration characteristics based on the operation of an oscillator that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0066] Looking at graph 210, the oscillator output voltage V OUT It can be confirmed that the characteristic of the oscillating frequency is constant.

[0067] FIG. 2C explains the operating characteristics of an oscillator constituting an oscillatory neural network circuit device according to one embodiment of the present invention in terms of changes in the energy band structure.

[0068] Referring to FIG. 2C, energy band 220 shows the case where a positive feedback loop is created, and energy band 221 shows the case where the positive feedback loop is eliminated.

[0069] The energy band 220 and the energy band 221 oscillate alternately, thereby producing an oscillation characteristic.

[0070] In summary, graph 210, energy band 220, and energy band 221 confirm that the oscillator output voltage oscillates with a constant period, and this oscillation characteristic is manifested by the repeated generation and disappearance of a positive feedback loop.

[0071] FIG. 3 is a diagram illustrating the change in characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention.

[0072] Referring to FIG. 3, a graph 300 shows the input voltage V IN Graph 310 shows the change in oscillator frequency due to resistor R S The figure shows the frequency change of the oscillator due to

[0073] Graph 300 and graph 310 are graphs showing the relationship between the input voltage V IN and resistance RS This shows that the frequency changes in a proportional or inversely proportional relationship depending on the frequency, which can be proven through the following equation 1.

[0074]

number

[0075] Equation 1 states that the oscillator frequency is proportional to the input voltage V IN is proportional to the resistance R S It is shown that the oscillation amplitude and the capacitance of the oscillator (C total ) is considered additionally.

[0076] 4A and 4B are diagrams illustrating tuning characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention.

[0077] FIG. 4A shows the relationship between the first gate voltage V and the tuning characteristic of the oscillator constituting the oscillatory neural network circuit device according to one embodiment of the present invention. G1 This shows the changes caused by

[0078] Referring to FIG. 4A, a schematic diagram 400 of an oscillator is shown, in which a resistor is connected to the drain terminal and a first gate voltage V G1 Graph 401 shows the switching characteristics due to the change in the output voltage over time, and graphs 402 and 403 show the change in amplitude and frequency due to the change in the output voltage over time.

[0079] The operating characteristics of the oscillator according to the structural diagram 400 are as follows: G1 This shows that the height of the potential barrier for holes injected from the drain region can be adjusted by

[0080] This means that, according to Equation 1, the amplitude changes in proportion to the latch-up voltage and the latch-down voltage, and the frequency is inversely proportional to the amplitude.

[0081] According to one embodiment of the present invention, when a resistor is connected to the drain terminal of the gated diode, a first gate voltage input to the first gate terminal causes p + The height of the potential barrier for holes injected from the drain region is controlled.

[0082] In addition, the gated diode generates a positive feedback loop proportional to the first gate voltage, and as the latch-up voltage moves in the positive direction, the amplitude and frequency of the oscillator can be changed.

[0083] FIG. 4B shows the relationship between the tuning characteristics of the oscillator constituting the oscillatory neural network circuit device according to one embodiment of the present invention and the second gate voltage V G2 This shows the changes caused by

[0084] Referring to FIG. 4B, a schematic diagram 410 of an oscillator is shown with a resistor connected to the drain terminal and a second gate voltage V G2 Graph 411 shows the switching characteristics as the voltage changes, and graphs 412 and 413 show that the amplitude and frequency do not change as the output voltage changes over time.

[0085] That is, the second gate voltage V G2 adjusts the depth of the potential well in which the injected holes accumulate, and shows that this does not affect the switching characteristics of the device and the oscillation characteristics of the oscillator.

[0086] The gated diode of the present invention can adjust the energy band by adjusting the gate voltage, thereby enabling tuning of oscillator characteristics. The tuning characteristics can correct characteristic deviations of the fabricated device, thereby realizing a highly reliable oscillator.

[0087] FIG. 5 is a diagram illustrating the change in characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to an embodiment of the present invention when the oscillator uses electrons as the majority carriers.

[0088] Referring to FIG. 5, a graph 500 shows that when a resistor is connected in series with the source region of a gated diode, it can exhibit oscillatory characteristics using electrons as the majority carriers.

[0089] Graph 500 shows the input voltage V IN Similarly, when a negative voltage is applied, the output voltage V OUT This shows that the vibration characteristic is exhibited.

[0090] Graphs 501 and 502 show the change in frequency that is proportional or inversely proportional to the input voltage and resistance, in accordance with Equation 1 above.

[0091] In summary, graphs 500 to 502 show the results when a resistor is connected to the source terminal and the second gate voltage V G2 It is shown that vibration characteristics can be tuned by controlling the

[0092] 6A and 6B are diagrams illustrating the tuning characteristics of an oscillator that constitutes an oscillatory neural network circuit device according to one embodiment of the present invention when the oscillator uses electrons as the majority carriers.

[0093] FIG. 6A shows the relationship between the first gate voltage V and the tuning characteristic of the oscillator constituting the oscillatory neural network circuit device according to one embodiment of the present invention. G1 This shows the changes caused by

[0094] Referring to FIG. 6A, a schematic diagram 600 of an oscillator is shown, in which a resistor is connected to the source terminal and a first gate voltage V G1 Graph 601 shows the switching characteristics as the voltage changes, and graphs 602 and 603 show that the amplitude and frequency do not change as the output voltage changes over time.

[0095] That is, in the oscillator structural diagram 600, the first gate voltage V G1 It can be confirmed that the switching characteristics of the gated diode and the oscillation characteristics of the oscillator are not changed because the depth of the potential well for the injected electrons is adjusted.

[0096] FIG. 6B shows the relationship between the second gate voltage V and the tuning characteristic of the oscillator constituting the oscillatory neural network circuit device according to one embodiment of the present invention. G2 This shows the changes caused by

[0097] Referring to FIG. 6B, a schematic diagram 610 of an oscillator is shown with a resistor connected to the source terminal and a second gate voltage V G2 Graph 611 shows the switching characteristics due to the change in , and graphs 612 and 613 show the change in amplitude and frequency through the change in output voltage over time.

[0098] The operating characteristics of the oscillator according to the structural diagram 610 are as follows: G2 This shows that the height of the potential barrier for holes injected from the drain region can be adjusted by

[0099] This means that, according to Equation 1, the amplitude changes in proportion to the latch-up voltage and the latch-down voltage, and the frequency is inversely proportional to the amplitude.

[0100] According to one embodiment of the present invention, when a resistor is connected to the source terminal of the gated diode, a second gate voltage input to the second gate terminal causes n + The height of the potential barrier for electrons injected from the source region is controlled, and the amplitude and frequency of the oscillator can be changed by moving the latch-up voltage in the negative direction in proportion to the second gate voltage.

[0101] This shows the opposite trend to the oscillator characteristic changes described in FIG. 4A.

[0102] That is, as the latch-up voltage moves in the negative direction as shown in graph 611, the amplitude and frequency of the oscillator change.

[0103] Therefore, the oscillator according to an embodiment of the present invention can determine the type of majority carrier depending on the configuration method, and can tune the characteristics by the gate voltage.

[0104] 7A to 7C are diagrams illustrating an oscillatory neural network circuit device according to an embodiment of the present invention and an edge detection simulation using the same.

[0105] FIG. 7A is a diagram illustrating the structure of an oscillatory neural network circuit device using a gated diode-based oscillator according to one embodiment of the present invention.

[0106] Referring to FIG. 7A, an oscillatory neural network circuit device 700 using a gated diode-based oscillator according to one embodiment of the present invention has a structure in which a first oscillator 701 and a second oscillator 702 are coupled by a coupling capacitor.

[0107] As an example, the first oscillator 701 and the second oscillator 702 are gated diodes whose intrinsic regions are n based on different voltages applied via two gate terminals. * Channel region and p * While the channel region is changed to one of the channel regions, a potential barrier is formed in the intrinsic region by electrostatic doping, and a positive feedback loop related to the formed potential barrier is repeatedly generated and annihilated by fluctuations in electron injection based on the injection of majority carriers, thereby outputting an output voltage that oscillates with a certain period.

[0108] The output terminals of the first oscillator 701 and the second oscillator 702 are coupled to each other by a capacitor 703, and a phase difference in the output voltage may occur depending on the time difference between the input voltages input to the gated diodes. For example, the capacitor 703 may be referred to as a coupling capacitor.

[0109] FIG. 7B shows the input voltage V applied to each of the first oscillator and the second oscillator constituting the oscillatory neural network circuit device according to one embodiment of the present invention. IN1 and the input voltage V IN2 Time difference ΔT IN The vibration characteristics are shown.

[0110] Referring to FIG. 7B, a graph 710 shows the relationship between ΔT IN = 0.0ms, the oscillations of the two oscillators will be in-phase over time.

[0111] On the other hand, ΔT corresponding to graph 711 IN = 0.5 ms and ΔT corresponding to graph 712 IN = 1.0 ms, the oscillation of the two oscillators will be C They influence each other but move in opposite directions, out of phase.

[0112] For example, the time difference ΔT IN may indicate pixel values ​​of the input image for edge detection.

[0113] The oscillatory neural network circuit device according to an embodiment of the present invention is configured to detect the difference between two input pixel values ​​(ΔT IN =0.0ms), the output pixel value is black because they are in-phase. Conversely, when there is a difference between the two pixel values ​​(ΔT IN =0.5, 1.0 ms) are in opposite phase, so the output pixel value can be specified as white.

[0114] FIG. 7C illustrates an edge detection simulation in an oscillatory neural network circuit arrangement using a gated diode-based oscillator according to one embodiment of the present invention.

[0115] Referring to FIG. 7C, in accordance with one embodiment of the present invention, in a simulation 720, the time difference ΔT illustrated in FIG. IN 7 shows an example of the results of horizontal edge detection 722 and vertical edge detection 723 performed through an oscillatory neural network circuit device 721 in which two oscillators are coupled.

[0116] When processing the same window image, it can be seen that when the oscillators are coupled vertically, edge components in the horizontal direction 722 are detected, and conversely, when they are coupled horizontally, edge components in the vertical direction 723 are detected.

[0117] That is, each of the at least two gated diodes can output the output voltage for a vertical edge component when coupled laterally via a capacitor, and can output the output voltage for a horizontal edge component when coupled vertically via a capacitor.

[0118] In conclusion, the oscillator of the present invention is capable of selective edge detection depending on the coupling method.

[0119] Therefore, the present invention provides an oscillatory neural network circuit device that uses gated diodes and resistors that repeatedly generate and eliminate a positive feedback loop to implement oscillators, and solves combinatorial optimization problems with an image processor based on the coupling and synchronization characteristics between oscillators.

[0120] 8A and 8B are diagrams illustrating the structure and vibration characteristics of an oscillatory neural network circuit device according to one embodiment of the present invention.

[0121] FIG. 8A illustrates vibration characteristics in a structure in which a vibration neural network circuit device according to an embodiment of the present invention is coupled by a capacitor.

[0122] Referring to FIG. 8A, an oscillator constituting an oscillatory neural network circuit device 800 according to an embodiment of the present invention is coupled with a capacitor.

[0123] Graph 801 shows the vibration characteristics of a vibration neural network circuit device 800 according to one embodiment of the present invention.

[0124] FIG. 8B illustrates vibration characteristics in a structure in which a vibration neural network circuit device according to an embodiment of the present invention is coupled with a resistor.

[0125] Referring to FIG. 8B, the oscillators constituting the oscillatory neural network circuit device 810 according to one embodiment of the present invention are coupled with resistors.

[0126] Graph 811 shows the vibration characteristics of a vibration neural network circuit device 810 according to one embodiment of the present invention.

[0127] An oscillatory neural network circuit device according to an embodiment of the present invention can solve graph coloring problems based on combinatorial optimization.

[0128] The oscillatory neural network circuit device 800 includes a capacitor (C C =100pF), the oscillations of the two oscillators are synchronized over time to an out-of-phase state where the phase difference between them is 180 degrees.

[0129] In the graph 801, when the gated diode of the second oscillator located on the left side of the first and second oscillators is turned on, the charge accumulated in the output terminal is discharged to the ground through the gated diode, and the output voltage V OUT1decreases rapidly.

[0130] At the same time, the output voltage V of the first oscillator OUT2 also decreases slightly, but this is because some of the charge stored at the output terminal of the right oscillator by the capacitor also leaks out to ground via the left diode.

[0131] The oscillatory neural network circuit device 810 is the opposite of the oscillatory neural network circuit device 800, and the oscillatory neural network circuit device 810 has a resistance (R C It consists of two oscillators coupled with a resistor (=100kΩ).

[0132] As can be seen from the graph 810, the phases are synchronized to an in-phase state, where the phases are completely the same.

[0133] This is due to the coupling resistance R C is the series resistance R S Since the resistance is small compared to the resistance of the gated diode, the output terminals of the two oscillators may share carriers through voltage distribution.

[0134] Therefore, the coupling strength of our oscillator can be expressed as a resistor (strong coupling) and a capacitor (weak coupling), which can be used to map the input graph of a graph coloring problem to an oscillatory neural network.

[0135] Therefore, the present invention can realize an oscillator configured with a gated diode, which overcomes the limitations in processing speed and performance degradation caused by the separation of memory and processor, and an oscillatory neural network circuit device using the same.

[0136] 9A to 10 are diagrams illustrating a configuration for implementing the color of a graph of an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention.

[0137] 9A to 9D illustrate an example of a configuration in which an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention is configured with four oscillators to embody the colors of the graph.

[0138] Referring to FIG. 9A, an input graph 900 and a corresponding oscillatory neural network circuit device 901 are shown in relation to a configuration in which an oscillatory neural network circuit device for an input graph according to one embodiment of the present invention is configured with four oscillators to embody the color of the graph.

[0139] The nodes 1 to 4 of the input graph 900 correspond to the oscillators DO1 to DO4 of the vibration neural network circuit device, and coupling capacitors and resistors are used to connect the nodes.

[0140] For example, when four oscillators are used, the output terminals of the first, second, third, and fourth oscillators constituting the four oscillators are coupled horizontally or vertically via the capacitors and diagonally coupled via resistors, and the output terminals coupled via resistors have the same phase, but a phase difference may occur with respect to the phase coupled via the capacitors.

[0141] For example, horizontal may be the same as widthwise, and vertical may be the same as lengthwise.

[0142] Referring to FIG. 9B, a graph 910 shows the operating characteristics of the four oscillators that make up the oscillatory neural network circuit arrangement.

[0143] The phases of the first oscillator DO1, the second oscillator DO2, the third oscillator DO3 and the fourth oscillator DO4 change over time, and the first oscillator DO1 and the fourth oscillator DO4 exhibit the same phase, and the second oscillator DO2 and the third oscillator DO3 exhibit the same phase.

[0144] The first oscillator DO1 and the fourth oscillator DO4 have a strong coupling strength R C14 , the phases of the first and second oscillators DO2 and DO3 are the same, and the second and third oscillators DO2 and DO3 exhibit a constant phase difference.

[0145] The second oscillator DO2 and the third oscillator DO3 have a strong coupling strength R C23 , the phases of the first oscillator DO1 and the fourth oscillator DO4 are the same, and the first oscillator DO1 and the fourth oscillator DO4 exhibit a constant phase difference.

[0146] That is, the first oscillator is horizontally coupled to the second oscillator via a capacitor and vertically coupled to the third oscillator, outputting an output voltage with a phase difference from the output voltages of the second and third oscillators, and is diagonally coupled to the fourth oscillator via a resistor, outputting an output voltage with the same phase.

[0147] The second oscillator is horizontally coupled to the first oscillator via a capacitor and vertically coupled to the fourth oscillator, and outputs an output voltage having a phase difference with the output voltages of the first and fourth oscillators. The second oscillator is diagonally coupled to the third oscillator via a resistor, and outputs an output voltage having the same phase.

[0148] Referring to FIG. 9C, a diagram showing a phase change over time in relation to an output voltage output by an oscillatory neural network circuit device for an input graph according to one embodiment of the present invention.

[0149] Graph 920 shows the phase change for the output voltages of the first and fourth oscillators over time.

[0150] In graph 920, the radius axis of the circle indicates the passage of time, with the center and circumference indicating 10 ms and 55 ms, respectively.

[0151] The phase corresponding to the output voltage of the first oscillator may be set as a reference (0°).

[0152] The coupling configuration of the oscillatory neural network causes the output voltages of the oscillators to interact and, over time, reach a steady state.

[0153] In the graph coloring problem, an oscillatory neural network circuit device for an input graph according to one embodiment of the present invention can be configured to color in different colors (e.g., red and blue) when the phase difference is greater than or equal to ±90° in relation to the phase of the oscillator output voltage.

[0154] Referring to FIG. 9D, an example is shown in which the color of the graph is output according to the phase difference over time in relation to the output voltage output by the oscillatory neural network circuit device for the input graph according to one embodiment of the present invention.

[0155] Graph 930 shows that the colors based on the output voltages of the first and fourth oscillators are the same, and the colors based on the output voltages of the second and third oscillators are the same, as confirmed by graph 920 in FIG. 9C.

[0156] Graph 930 shows the oscillator phase difference over 50 ms, showing that the phase difference between second oscillator 2 and fourth oscillator 4 is approximately 120°.

[0157] The oscillatory neural network device according to one embodiment of the present invention can solve the graph coloring problem by coloring the graph in a solution where nodes 1 and 4 of the input graph are red and nodes 2 and 3 are blue through phase difference.

[0158] Thus, the present invention can provide an oscillatory neural network circuit device that solves the graph coloring problem by using multiple oscillators coupled by capacitors or resistors to detect edges or indicate at least two graph colors for an input graph.

[0159] FIG. 10 illustrates a configuration for implementing various graph colors of an oscillatory neural network circuit device for an input graph according to an embodiment of the present invention.

[0160] Referring to FIG. 10, as in the first embodiment 1000, the second embodiment 1010, the third embodiment 1020, and the fourth embodiment 1030, the graph coloring problem can be solved by checking the phase change over time for various input graphs and displaying the color of the graph based on the phase difference.

[0161] For an input graph, over time the oscillators interact and the phase stabilizes.

[0162] Unlike conventional oscillatory neural networks that must apply an appropriate coloring algorithm depending on the input graph, the oscillatory neural network of the present invention can solve all coloring problems based on a phase difference of ±45°, regardless of the input graph, as can be seen from the phase difference graph.

[0163] In other words, at least two graph colors can be displayed, consisting of a first color corresponding to output voltages of the same phase and a second color corresponding to output voltages of phases where the phase difference exists.

[0164] According to one embodiment of the present invention, the oscillatory neural network circuit device considers the number of nodes constituting an input graph, the connection structure between the nodes, and the colors of the nodes, and determines a color separation criterion for the phase difference through a time-dependent phase change of the oscillator based on the phase difference using an oscillator corresponding to the number of nodes, and can display the colors of the at least two graphs with colors separated based on the color separation criterion.

[0165] Therefore, the present invention can realize an oscillatory neural network circuit device that can be used as a solver for combinatorial optimization problems (COPs) with an image processor by using a highly integrated oscillator that is highly reliable and can be applied to existing CMOS processes.

[0166] In the above-described specific embodiments, elements included in the invention are expressed as singular or plural by the specific embodiments presented.

[0167] However, the expressions "singular" or "plural" are selected to suit the circumstances presented for the convenience of explanation, and the above-described embodiments are not limited to singular or plural components, and elements expressed as plural may be composed of singular, and elements expressed as singular may be composed of plural.

[0168] Meanwhile, although specific embodiments have been described in the explanation of the invention, it goes without saying that various modifications are possible as long as they do not deviate from the scope of the technical ideas contained in the various embodiments.

[0169] Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the following claims as well as equivalents to these claims.

Claims

1. Between the drain terminal and the source terminal, ＋ -i-n ＋ The diode structure is located ＋ -i-n ＋ an oscillator including a gated diode having a gate insulating film on an intrinsic region of the diode structure and two gate terminals on the gate insulating film, the gated diode having a resistor connected in series with the gated diode; The oscillator is configured such that the gated diode is connected to the intrinsic region of the n-type gate electrode based on different voltages applied to the gated diode via the two gate terminals. ＊ The channel region and p ＊ While being changed to one of the channel regions, a potential barrier is formed in the intrinsic region by electrostatic doping, and a positive feedback loop related to the formed potential barrier is repeatedly generated and annihilated by fluctuations in majority carrier injection, thereby outputting an output voltage that oscillates with a certain period; The oscillator is configured with at least two oscillators, and output terminals corresponding to portions where resistors are connected to the at least two gated diodes are coupled to each other by either a capacitor or a resistor, and a phase difference with respect to the output voltage occurs depending on a time difference between input voltages input to the at least two gated diodes, thereby displaying the colors of at least two graphs with respect to the input graph based on the phase difference over time.

2. The gated diode has a p-type gate electrode between a first gate terminal of the two gate terminals and the drain terminal. ＋ When a positive voltage is applied to the first gate terminal, the region under the first gate terminal in the intrinsic region becomes n ＊ a second gate terminal of the two gate terminals and the source terminal; ＋ When a negative voltage is applied to the second gate terminal, the region under the second gate terminal in the intrinsic region becomes p ＊ 10. An oscillatory neural network circuit device using a gated diode-based oscillator according to claim 1, comprising as a channel region.

3. When the resistor is connected to the drain terminal, the gated diode is ＋ 3. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 2, wherein the amplitude and frequency of the oscillator are changed by controlling the height of a potential barrier of holes injected from the drain region and by shifting a latch-up voltage in a positive direction, which generates a positive feedback loop proportional to the first gate voltage.

4. When the resistor is connected to the source terminal, the gated diode is turned on by a second gate voltage input to the second gate terminal. ＋ 3. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 2, wherein the height of a potential barrier of electrons injected from the source region is controlled, and the amplitude and frequency of the oscillator are changed by shifting a latch-up voltage in a negative direction in proportion to the second gate voltage.

5. 2. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 1, wherein a first output pixel value is determined based on the output voltages output in an out-of-phase state, which exhibits opposite movement as time passes, as an effect occurs between the at least two gated diodes based on the capacitor, based on a time difference between when a first input voltage and a second input voltage are applied to each of the at least two gated diodes. Alternatively, a second output pixel value opposite to the first output pixel value is determined based on the output voltages output in-phase, as the first input voltage and the second input voltage are applied at the same time so that the time difference does not occur.

6. 2. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 1, wherein each of the at least two gated diodes outputs the output voltage of a vertical edge component when coupled laterally via the capacitor, and outputs the output voltage of a horizontal edge component when coupled vertically via the capacitor.

7. 2. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 1, wherein when the oscillator is configured with four oscillators, output terminals of a first oscillator, a second oscillator, a third oscillator, and a fourth oscillator constituting the four oscillators are coupled horizontally or vertically via the capacitor and diagonally via a resistor, and the output terminals coupled via the resistor have the same phase, and the phase difference occurs with respect to the phase coupled via the capacitor.

8. the first oscillator is horizontally coupled to the second oscillator via the capacitor and vertically coupled to the third oscillator, and outputs an output voltage having a phase difference with respect to output voltages of the second oscillator and the third oscillator; and is diagonally coupled to the fourth oscillator via the resistor, and outputs an output voltage having the same phase; 8. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 7, wherein the second oscillator is horizontally coupled to the first oscillator via the capacitor and vertically coupled to the fourth oscillator to output an output voltage having a phase difference with respect to the output voltages of the first oscillator and the third oscillator, and is diagonally coupled to the third oscillator via the resistor to output an output voltage having the same phase.

9. 9. The oscillatory neural network circuit device using a gated diode-based oscillator according to claim 8, wherein the colors of the at least two graphs are configured as a first color corresponding to the output voltages of the same phase, and a second color corresponding to the output voltages of the phases where there is a phase difference.

10. 9. The oscillatory neural network circuit device using gated diode-based oscillators according to claim 8, wherein the oscillators corresponding to the number of nodes are used in consideration of the number of nodes constituting the input graph, the connection structure between the nodes, and the colors of the nodes, and a color separation criterion for the phase difference is determined through a phase change over time of the oscillators based on the phase difference, and the colors of the at least two graphs are displayed in colors separated based on the color separation criterion.

Images