Solution device and method for sampling qubits in quantum annealer

Abstract

The present invention relates to a solution device and method for sampling qubits in a quantum annealer. The solution device according to the present invention finds optimized qubits through an iterative operation on sampling qubits for a problem input in the quantum annealer, wherein the optimized qubits can be effectively found by reducing the size of the operation by fixing highly reliable qubit values by various fixing methods considering bias and system energy.

Description

Solution device and method for sampling qubits in a quantum annealer

[0001] The present invention relates to a solution device and method for finding a highly reliable optimal value for a sampling qubit (quantum bit) of a quantum annealer.

[0002] Quantum annealing utilizes superconductor-based quantum annealers that practically implement the Ising model. However, technical limitations hinder the initialization and manipulation of each qubit sampled from the target problem for which the quantum annealer seeks a solution. The resulting errors prevent the annealer from accurately calculating the desired quantum computing model from the sampled qubits, resulting in inaccurate results.

[0003] Therefore, a highly reliable solution device and method are required to correct these errors.

[0004] As related prior literature, reference may be made to Patent Registration No. 10-2361858 (February 8, 2022), Patent Publication No. 10-1768066 (August 8, 2017), and Patent Publication No. 10-2018-0022925 (March 6, 2018).

[0005] Accordingly, the present invention has been devised to solve the above-described problem, and the purpose of the present invention is to provide a solution device and method capable of effectively finding optimized qubits by reducing the size of the operation by fixing reliable qubit values ​​using various fixing methods that take bias and system energy into account while finding optimized qubits through repeated operations on the sampled qubits of the quantum annealer for the input of the problem.

[0006] First, to summarize the features of the present invention, a solution method performed in a processor of a device according to one aspect of the present invention for achieving the above object may include the steps of: generating a plurality of sample sets in a quantum annealer for an input; setting a threshold to be referred to for qubit fixation; calculating a bias for a target qubit in each sample set of the plurality of sample sets and determining candidates having the bias greater than or equal to the threshold; recalculating the bias of the target qubits for a sample set of a corresponding subset consisting of qubits having the same fixed expected qubit value as the candidates; calculating a local system energy based on the sum of the energy of each of the candidates and the coupling energy of each of the candidates according to the fixed expected qubit value of each of the candidates and the recalculated bias; and classifying the candidates into fixed qubits and qubits to be re-operated depending on whether the local system energy decreases.

[0007] The above solution method is to find a solution for the fixed qubits while reducing the size of the operation by repeatedly generating the qubits to be re-operated as the input to the quantum annealer and excluding the fixed qubits again.

[0008] When excluding the fixed qubit and repeating the process by inputting the qubit to be re-operated into the quantum annealer, in the step of setting the threshold, the set value of the threshold can be set to gradually increase or decrease compared to the threshold of the previous routine.

[0009] In the step of setting the threshold, the threshold may be set so that the number of qubits in each sample set generated by the annealer (current problem size) is proportional to the number of qubits exceeding the threshold, and the bias is proportional to the number of qubits exceeding the threshold.

[0010] The number of fixed qubits in the above classification step is predetermined, and among the qubits whose bias is greater than or equal to the threshold, the predetermined number of qubits can be determined as the fixed qubits in order of the difference between them.

[0011] In the determination of the fixed qubit in the above classification step, among the qubits whose bias is greater than or equal to the threshold, a qubit whose number of coupling qubits is greater than or equal to a predetermined number can be determined as the fixed qubit.

[0012] In the determination of the fixed qubit in the above classification step, among the qubits whose bias is greater than or equal to the threshold, one or more qubits whose local system energy is greater than or equal to a predetermined value can be determined as the fixed qubit.

[0013] When excluding the fixed qubit and repeating the process by inputting the qubit to be re-operated into the quantum annealer, in the determination of the fixed qubit in the classification step, it can be performed so that one or more qubits whose bias is greater than the threshold and whose local system energy is reduced compared to the previous routine are not determined as the fixed qubit.

[0014] When excluding the fixed qubits and repeating the process by inputting the qubits to be re-operated into the quantum annealer, in the determination of the fixed qubits in the classification step, one or more qubits whose bias is greater than the threshold and whose local system energy has increased compared to the previous routine can be determined as the fixed qubits.

[0015] When excluding the fixed qubits and repeating the process by inputting the qubits to be re-operated into the quantum annealer, in the classification step, one or more qubits whose bias is less than the threshold and whose local system energy has increased compared to the previous routine can be determined as the fixed qubits.

[0016] The above solution method may further include, before the step of calculating the bias, a step of determining a combination of qubits on a qubit sequence for the plurality of sample sets as the target qubit.

[0017] In the step of determining the above qubit combination as the target qubit, the combination may be determined based on the number of coupled qubits that the two coupled qubits have in common, or the combination may be determined with qubits that belong to a value greater than the standard deviation in the distribution of the bias coefficient or coupling strength coefficient of all qubits among the two coupled qubits.

[0018] In the step of determining the above qubit combination as the target qubit, the combination may be determined based on the number of coupled qubits that two uncoupled qubits have in common, or the combination may be determined with qubits that belong to a value greater than the standard deviation in the distribution of the bias coefficient or coupling strength coefficient of all qubits among the two uncoupled qubits.

[0019] In the step of determining the above qubit combination as the target qubit, three or more qubits belonging to a cluster classified by a predetermined clustering method, regardless of coupling, may be determined as the combination.

[0020] In addition, a solution device according to another aspect of the present invention may include a quantum annealer that generates a plurality of sample sets for an input; a threshold setting unit that sets a threshold to be referred to for qubit fixation; a bias calculation and combination setting unit that calculates a bias for a target qubit in each sample set of the plurality of sample sets and determines candidates having the bias greater than or equal to the threshold; an adjacent qubit bias recalculation unit that recalculates the bias of target qubits with respect to a sample set of a corresponding subset consisting of qubits having the same fixed expected qubit value as the candidates; a system energy calculation unit that calculates local system energy based on the sum of the energy of each of the candidates and the coupling energy of each of the candidates according to the fixed expected qubit value of each of the candidates and the recalculated bias; and a fixation unit that classifies the candidates into fixed qubits and qubits to be re-operated depending on whether the local system energy decreases.

[0021] And, a computer-readable code for performing a solution function performed in a processor of a device according to another aspect of the present invention may be recorded on a recording medium having recorded thereon a function of generating a plurality of sample sets in a quantum annealer for an input; a function of setting a threshold to be referred to for fixing a qubit; a function of calculating a bias for a target qubit in each sample set of the plurality of sample sets and determining candidates having the bias greater than or equal to the threshold; a function of recalculating the bias of the target qubits for a sample set of a corresponding subset consisting of qubits having the same fixed expected qubit value as the candidates; a function of calculating local system energy based on the sum of the energy of each of the candidates and the coupling energy of each of the candidates according to the fixed expected qubit value of each of the candidates and the recalculated bias; and a function of classifying the candidates into fixed qubits and qubits to be re-operated depending on whether the local system energy decreases.

[0022] According to the solution device and method according to the present invention, by applying various fixation methods in consideration of bias and system energy by statistical analysis for the sampling qubits of the quantum annealer for the input of the problem, the size of the operation can be reduced by fixing highly reliable qubit values, and the solution of optimized qubits that is faster and more reliable than before can be effectively obtained.

[0023] The accompanying drawings, which are included as part of the detailed description to aid understanding of the present invention, provide examples of the present invention and, together with the detailed description, explain the technical idea of ​​the present invention.

[0024] FIG. 1 is a block diagram illustrating a solution device according to one embodiment of the present invention.

[0025] FIG. 2 is an exemplary diagram of a set of qubits generated by a quantum annealer according to one embodiment of the present invention.

[0026] Figure 3a is an example of a condition for a combination of two coupled qubits of the present invention.

[0027] Figure 3b is an example of a condition for a combination of two uncoupled qubits of the present invention.

[0028] FIG. 3c is an example of qubit clustering conditions for a combination of three or more qubits of the present invention.

[0029] FIG. 4a shows examples of conditions for selectively fixing qubits having a bias greater than a threshold value according to the present invention.

[0030] Figure 4b shows examples of conditions for selectively fixing the bias threshold and the increase or decrease in local system energy in the iterative routine of the present invention.

[0031] FIG. 5 is an example of a graph showing the change trend of system energy as the number of repetition routines increases in the annealing of the present invention.

[0032] FIG. 6 is a drawing for explaining an example of a method for implementing a solution device according to one embodiment of the present invention.

[0033] Hereinafter, the present invention will be described in detail with reference to the attached drawings. In this case, the same components are indicated by the same reference numerals in each drawing, where possible. In addition, detailed descriptions of functions and / or configurations already known will be omitted. The content disclosed below focuses on parts necessary for understanding the operation according to various embodiments, and descriptions of elements that may obscure the gist of the description will be omitted. In addition, some components in the drawings may be exaggerated, omitted, or schematically illustrated. The size of each component does not entirely reflect the actual size, and therefore, the contents described herein are not limited by the relative sizes or spacing of components drawn in each drawing.

[0034] In describing embodiments of the present invention, if a detailed description of a known technology related to the present invention is judged to unnecessarily obscure the gist of the present invention, the detailed description will be omitted. In addition, the terms described below are terms defined in consideration of their functions in the present invention, and this may vary depending on the intention or custom of the user or operator. Therefore, the definitions should be made based on the contents throughout this specification. The terminology used in the detailed description is only for the purpose of describing embodiments of the present invention and should not be limited in any way. Unless clearly used otherwise, the singular form includes the plural form. In this description, expressions such as "comprises" or "having" are intended to indicate certain features, numbers, steps, operations, elements, parts, or combinations thereof, and should not be construed to exclude the presence or possibility of one or more other features, numbers, steps, operations, elements, parts, or combinations thereof other than those described.

[0035] Additionally, although terms such as first, second, etc. may be used to describe various components, the components are not limited by the terms, and the terms are used only for the purpose of distinguishing one component from another.

[0036] First, a method for performing a solution to find optimized qubits through repeated operations on the sampled qubits of a quantum annealer for the input of a predetermined problem (or signal or information) to be optimized for qubit values ​​(e.g., binary bit values) according to an embodiment of the present invention will be briefly described.

[0037] A qubit (or quantum bit) is the fundamental unit of information in quantum computing. Binary bits, the fundamental unit of information in classical computing, represent either -1 or 1 (or 0 or 1). In contrast, qubits utilize quantum mechanical phenomena to represent linear combinations of two states. A qubit can represent any ratio of 0 to 1 by superposing two states: a state with a certain probability of 0 and a state with a certain probability of 1. Qubits can be implemented using trapped ions, photons, artificial or real atoms, quasiparticles, etc. Quantum annealing using a quantum annealer effectively solves optimization problems to find optimal qubits. "Hill climbing" in classical computing is an approach that iteratively changes parameters to find a solution that is slightly better than the current solution. The name quantum annealing is derived from a classical computing technique called simulated annealing.

[0038] For example, a quantum annealer may include a metal ring formed by processing the metal niobium (Nb). Multiple rings are used for the quantum annealer. Depending on the direction of the current flowing through the metal ring, the qubit value can be 1 or 0. For example, if the current flows counterclockwise, the qubit value can be 1, and if the current flows clockwise, the qubit value can be 0. When the temperature of the metal ring is reduced to absolute zero, the metal ring becomes a superconductor, allowing current to flow in both directions, allowing a qubit in a superposition state of 0 and 1 to be expressed. When the metal rings are in a superposition state, if a magnetic field is applied transversely, the temperature of the metal ring increases, eliminating the superconducting effect. When the magnetic field is gradually weakened, an interaction occurs between each metal ring. Depending on the interaction between each metal ring, the current flows in either a clockwise or counterclockwise direction through each metal ring. By controlling the interactions between each metal ring, the direction of the current within each metal ring can be controlled. The more metal rings with current flowing in the same direction, the higher the signal measured. By repeating the signal measurement while varying the combination of interactions between the metal rings, an optimized combination of qubits (sequences) can be generated.

[0039] Heuristic methods for finding optimized qubits using quantum annealers like this can be used in real-world applications that require complex computations, such as optimizing logistics routes, optimizing traffic signal systems, and optimizing scenarios for investment portfolios, as a way to find the best solution among a set of feasible options.

[0040] The present invention estimates the optimal value of a qubit through statistical analysis of quantum annealing results, and then fixes highly reliable qubit values ​​during an iterative annealing process, thereby reducing the size of the calculation and effectively obtaining optimized qubit solutions that are faster and more reliable than before. The method of the present invention gradually reduces the size of the problem in a repetitive routine, thereby reducing the search space of the quantum annealer, thereby improving not only the computational time but also the quality of the computational results.

[0041] Hereinafter, with reference to FIGS. 1 to 6, a solution device and method for sampling qubits of the quantum annealer of the present invention will be described in detail.

[0042] FIG. 1 is a block diagram illustrating a solution device (100) according to one embodiment of the present invention.

[0043] Referring to FIG. 1, a solution device (100) according to one embodiment of the present invention may include a quantum annealer (110), a threshold setting unit (115), a bias calculation and combination setting unit (120), an adjacent qubit bias recalculation unit (130), a system energy calculation unit (140), and a fixing unit (150).

[0044] According to an embodiment of the present invention, the quantum annealer (110), the bias calculation and combination setting unit (120), the adjacent qubit bias recalculation unit (130), the system energy calculation unit (140), the threshold setting unit (115), and the fixing unit (150) constituting the solution device (100) may be implemented not only by hardware such as a semiconductor processor, but also by being combined with software such as an application program. Alternatively, each of the above devices may be implemented only as software composed of computer-readable codes on a computer and may be a device that operates virtually. In addition, the solution device (100) of the present invention may include a memory for storing data or setting information necessary for performing the solution processing process (see FIG. 6). Here, the functions of the bias calculation and combination setting unit (120), the adjacent qubit bias recalculation unit (130), the system energy calculation unit (140), the threshold setting unit (115), and the fixing unit (150) will be described, but two or more of these components may also be combined and implemented as a single block.

[0045] Hereinafter, each function of the quantum annealer (110), bias calculation and combination setting unit (120), adjacent qubit bias recalculation unit (130), system energy calculation unit (140), threshold setting unit (115), and fixing unit (150) performed in the processor of the solution device (100) will be described in detail.

[0046] FIG. 2 is an exemplary diagram of a set of qubits generated by a quantum annealer (110) according to one embodiment of the present invention.

[0047] First, the quantum annealer (110) generates sampling qubits (S20) for a given problem input (S10) to optimize qubit values ​​(e.g., binary bit values). That is, the quantum annealer (110) generates a plurality of sample sets (e.g., M sample sets) (the number of qubit sequences in a horizontal line in S20 of FIG. 1), each sample set consisting of a given qubit sequence (the number of which gradually decreases with the generation of a fixed qubit in each iteration routine).

[0048] For example, the input problem (H F ) can be expressed as in [Mathematical Formula 1] below, and this may correspond to a signal or information for causing the quantum annealer (110) to generate each sample set included in the qubit set as in FIG. 2. As in FIG. 2, the qubit set may include qubit clusters clustered by a clustering algorithm based on the Euclidean distance, etc.

[0049] [Mathematical Formula 1]

[0050]

[0051] Here, i, j are qubit indices, h i is the bias coefficient of qubit i, J i,j is the coupling strength coefficient between qubits i and j, , is the Pauli-Z operator applied to each qubit i and j. The bias coefficient (h) i ) and coupling strength coefficient (J i,j ) is a value preset by the user. This also applies when the target qubit is a combination of two or more.

[0052] The bias calculation and combination setting unit (120) determines a qubit combination on the qubit sequence for the plurality of sample sets generated by the quantum annealer (110). The qubit combination determines whether the qubits are coupled, or if the qubits are not coupled but have a high degree of bias (the degree of bias toward either binary value - 1 / 1) or coupling strength, the qubits are grouped into combinations (e.g., 00, 10, 01, 11 in two combinations, or 000, 001, 010, 011, 100, 101, 110, 111 in three combinations, etc.) and applied with computational processing as a unit qubit to analyze meaningful correlations between the qubits. The coupling (or proximity) refers to a case where the qubits are adjacent to each other so as to interact with each other. The physical implementation of coupling is possible based on Compound Josephson Junctions. Coupling strength factor J as described above i,j Using this, we can express the local system energy as follows. J i,j ≠0 indicates that the qubits are coupled or adjacent, and J i,j =0 indicates that the qubits are not coupled, or that the qubits are not adjacent.

[0053] In the case where each qubit is treated as a qubit to be processed as described below, the configuration related to the combination setting in the bias calculation and combination setting unit (120) may be omitted. However, in the case where the qubit combination consisting of two or more related qubit combinations is used as a qubit to be processed, the functions and operation methods in the subsequent adjacent qubit bias recalculation unit (130), system energy calculation unit (140), and fixing unit (150) may be calculated and processed in units of the corresponding qubit combinations instead of in units of qubits.

[0054] Figure 3a is a coupled (J) of the present invention i,j≠0) This is an example of a condition for the combination of two qubits.

[0055] As shown in Fig. 3a, for example, the bias calculation and combination setting unit (120) is coupled (J i,j ≠0) The combination of the qubits can be determined based on the number of coupled qubits that the two qubits have in common. That is, if the two qubits are coupled to each other and the number of qubits that are commonly coupled among the qubits coupled to each qubit is greater than a predetermined threshold, the combination of the qubits can be determined.

[0056] In addition, as in Fig. 3a, for example, the bias calculation and combination setting unit (120) is coupled (J i,j ≠0) Among two qubits, the combination of the qubits can be determined by qubits that belong to a value that is more than a standard deviation (e.g., 3 σ, etc.) in the distribution of the bias coefficient (h) or the coupling strength coefficient (J) of all qubits. For example, when the binary values ​​of the qubits are -1 and 1, and qubits having a bias coefficient (h) / coupling strength coefficient (J) of 0 as an average value, qubits having a bias coefficient (h) / coupling strength coefficient (J) smaller than -1, and qubits having a bias coefficient (h) / coupling strength coefficient (J) greater than 1, etc. form a certain (normal) distribution, the qubits whose bias coefficient (h) or coupling strength coefficient (J) is distributed in a value that is more than a standard deviation (e.g., 3 σ, etc.) than the average value (0) can be determined as the combination of the qubits.

[0057] Figure 3b is an example of a condition for a combination of two uncoupled qubits of the present invention.

[0058] As in Fig. 3b, for example, the bias calculation and combination setting unit (120) is uncoupled (J i,j=0) The combination can be determined based on the number of coupled qubits that two qubits have in common. That is, even if two qubits are not coupled to each other, if the number of qubits that are commonly coupled among the qubits coupled to each qubit is greater than a predetermined threshold, the combination of the qubits can be determined.

[0059] In addition, the bias calculation and combination setting unit (120) is uncoupled (J i,j =0) Among the two qubits, the combination can be determined with qubits that belong to a value greater than or equal to the standard deviation (e.g., 3σ, etc.) of the distribution of the bias coefficient (h) or coupling strength coefficient (J) of all qubits. Similarly to the above, when the binary values ​​of the qubits are -1 and 1, and qubits having a bias coefficient (h) / coupling strength coefficient (J) of 0 as the average value, qubits having a bias coefficient (h) / coupling strength coefficient (J) smaller than -1, and qubits having a bias coefficient (h) / coupling strength coefficient (J) greater than 1, etc. form a certain (normal) distribution, qubits whose bias coefficient (h) or coupling strength coefficient (J) is distributed in a value greater than or equal to the standard deviation (e.g., 3σ, etc.) of the average value (0) can be determined as the combination of the qubits.

[0060] FIG. 3c is an example of qubit clustering conditions for a combination of three or more qubits of the present invention.

[0061] As shown in FIG. 3c, for example, the bias calculation and combination setting unit (120) may determine three or more qubits belonging to a cluster classified by a predetermined clustering method, regardless of coupling, as the combination. As shown in FIG. 2, the qubit set may include qubit clusters clustered by a clustering algorithm based on a Euclidean distance or the like. Among the qubits clustered in this way, three or more qubits within the same cluster group may be determined as the combination. The three or more qubits may be randomly selected, or may be selected using the coupling relationship, the number of common coupling qubits, and the distribution of the bias coefficient (h) or the coupling strength coefficient (J), as shown in FIG. 3a and FIG. 3.

[0062] The threshold setting unit (115) sets the threshold (z) of the bias (z) to be referenced for qubit fixation. f ) is set. For example, the bias (z) calculated by the bias calculation and combination setting unit (120) and the adjacent qubit bias recalculation unit (130) can be between 1 and -1, and the first threshold (-z f ) is less than -0.8 second critical value (z) f ) can be set to a range such as 0.8 or more. That is, here, the bias (z) of -0.8 or less is fixed to -1, and the bias (z) of +0.8 or more is fixed to +1. In a similar way, when the target qubit is a qubit combination, the threshold (z) for each target qubit to be fixed is similarly set for the qubit combination. f ) can be set. The threshold (z) in the threshold setting unit (115) f ) The setup method will be explained in more detail below.

[0063] In addition, the bias calculation and combination setting unit (120) calculates the bias (z) for each target qubit in each sample set of the plurality of sample sets, and the bias (z) is equal to the threshold (z f) can determine candidates that are more than 1. In the sample sets (S20) of Fig. 1, the qubit values ​​in the same column become sample values ​​for the qubits in the same position in Fig. 2, and the bias (z) corresponds to the expected value for the qubit value at each position. The bias (z) is an operator It can be calculated by applying a predetermined probability state function to it. For example, in Fig. 1, the value obtained by dividing the number of qubits having '1' among the qubit values ​​of the first column by the number of sample sets (e.g., M) can represent the bias (z). Similarly, the probability state function can be applied to the qubit combination, and in a similar way, when the target qubit is a qubit combination, the bias (z) of each value for 00, 10, 01, 11 in 2 combinations, or 000, 001, 010, 011, 100, 101, 110, 111 in 3 combinations can be calculated.

[0064] In other words, the bias calculation and combination setting unit (120) calculates the bias (z) for each qubit or target qubit of the qubit combination and sets the initially set threshold (z f ) is the bias (z) of the target qubit based on the threshold (z f ) is exceeded, the candidates are determined to check whether the target qubit is fixed to the expected optimal value ( / z), that is, the fixed expected qubit value (-1 or 1).

[0065] The adjacent qubit bias recalculation unit (130) recalculates the bias (z) of the target qubits for a sample set of the subset consisting of qubits having the same fixed expected qubit value (-1 or 1) as the candidates determined by the bias calculation and combination setting unit (120). The bias in the entire qubit set and the subset may appear differently. The recalculation of the bias (z) can be performed in a similar manner to the calculation of the bias (z) in the bias calculation and combination setting unit (120).

[0066] That is, the adjacent qubit bias recalculation unit (130) is a part that recalculates only the bias of the qubits that are coupled or associated with the target qubit. The reason for recalculating the bias (z) here is that among the M samples, only the subset in which the value of the candidate qubit (i) is the expected optimal value ( / z), i.e., the fixed expected qubit value (-1 or 1), must be considered.

[0067] The system energy calculation unit (140) calculates the energy (h) of each of the candidates, as in [Mathematical Formula 2], according to the fixed expected qubit value of each of the candidates determined by the bias calculation and combination setting unit (120) and the bias (z) recalculated by the adjacent qubit bias recalculation unit (130). i / z (i) ) and the coupling energy (J) of each of the above candidates i,j / z (i) z (j) ) based on the sum (Σ) of the local system energy (δE i ) can be produced. Here, is a constant value -1 or +1 to be fixed for qubit i, and z (j) is the bias of qubit j, and neighbor(i), which is the neighbor of qubit i, represents qubit j coupled with i. Similarly, if the target qubit is a combination of qubits, it can be similarly calculated between qubit combinations as in [Mathematical Formula 2].

[0068] [Mathematical Formula 2]

[0069]

[0070] That is, in the (Σ) term of the above equation, the bias z (j) is the coupling energy (J i,j / z (i) z (j)) is applied as a weight when obtaining. Here, for example, the corresponding subset consisting of identical qubits to be fixed to '1' in the first column in S20 of Fig. 1 may be the sample set from the first row to the fifth row. In each repeated routine, the local system energy (δE i ) decreases the probability of fixation, and conversely, the local system energy (δE i ) decreases the probability of fixation.

[0071] That is, the system energy calculation unit (140) calculates the local system energy (δE) based on the expected optimal value ( / z(i)) of the candidates determined by the bias calculation and combination setting unit (120), i.e., the fixed expected qubit value (-1 or 1), and the bias (z(j)) of the qubits coupled with the candidates by the adjacent qubit bias recalculation unit (130), and when the target qubit is fixed to the expected optimal value, i.e., the fixed expected qubit value (-1 or 1), it checks and fixes whether the energy (E) of the entire system is lowered (whether δE is negative).

[0072] The fixed part (150) is the local system energy (δE i ) decreases, the candidate among the above candidates is fixed to 1 or -1 and classified as a fixed qubit (fixed to 1 or -1), and all other qubits that are not fixed are classified as qubits to be operated on again. In a similar manner, if the target qubit is a qubit combination, the qubit combination can be fixed to 00, 01, 10, or 11, etc. in a similar manner.

[0073] The qubit (H) to be re-calculated above F SQF ) can be expressed as in [Mathematical Formula 3], and accordingly, the qubit (H) to be re-operated generated in the fixed part (150) F SQF) is input to the quantum annealer (110) instead of the input of [Mathematical Formula 1], and the routine is repeated, so that the fixed qubits are gradually excluded from each routine in the fixed unit (150), thereby reducing the size of the operation and finding a solution for the fixed qubits, and generating optimized qubits through the quantum annealer (110).

[0074] [Mathematical Formula 3]

[0075]

[0076] The qubit (energy) to be re-calculated according to the result of statistical qubit freezing (SQF) (H F SQF ) in [Mathematical Formula 3], when compared with [Mathematical Formula 1], the bias coefficient (h) in the first term i ) instead The coupling relationship between qubits (i, j) is reflected by applying the third term. In , an offset is applied to compensate for the input of the initial problem to be optimized so that it does not change due to the exclusion of the fixed qubits.

[0077] Meanwhile, in the threshold setting section (115) as above, the threshold (z) is set in each routine. f ) can be set to be fixed without change, but in the present invention, it can also be applied to be dynamically changed in each routine as follows.

[0078] For example, as above, the qubits to be re-operated (H) are excluded from the fixed qubit(s). F SQF ) is input to the quantum annealer (110) and the routine is repeated, the threshold setting unit (115) sets the threshold (z f) can be set to gradually increase or decrease from the threshold applied in the previous routine. For example, the bias (z) calculated by the adjacent qubit bias recalculation unit (130) can be between 1 and -1, and the threshold (z) for fixing it to 1 in each routine f ) is decreased as 0.8->0.79,->0.78,... or the threshold (z) to be fixed to 1 f ) can increase as 0.8->0.81->0.82,... Also, the threshold (z) to fix to -1 in each routine f ) can also increase or decrease similarly.

[0079] Alternatively, the threshold setting unit (115) sets the bias (z) calculated by the adjacent qubit bias recalculation unit (130) to the threshold (z f ) is proportional to the number of qubits exceeding the threshold (z). f ) can also be set. For example, if each sample set has M=10000, the bias (z) to fix to 1 is the threshold (z) of the previous routine. f )=0.8, the number of qubits exceeding 3000 in the previous routine and 4000 in the current routine is the threshold (z f ) can be set in the direction of increasing to 0.85, etc. At this time, the threshold setting unit (115) sets the threshold (z) in proportion to the number of qubits (current problem size) in each sample set in each routine generated by the annealer (110). f ) can be set.

[0080] In addition, the fixed part (150) has a threshold value (z f ) based on qubits (fixed to 1 or -1), various fixing conditions can be applied to fix qubits that meet the conditions.

[0081] The fixing conditions of the present invention will be described in more detail with reference to FIGS. 4a and 4b below.

[0082] Figure 4a shows that the bias (z) of the present invention is a threshold (z f ) are examples of conditions for selectively fixing among the qubits.

[0083] Referring to FIG. 4a, the fixed part (150) can apply a predetermined value (n) to the number of fixed qubits in each routine, for example, when the bias (z) is a threshold (z f ) Among the qubits having a difference of greater or lesser order, the predetermined number of qubits (n) can be determined as the fixed qubits.

[0084] In addition, the fixed part (150) has a bias (z) of a threshold (z) in each routine. f ) Among the qubits having a number of coupling qubits greater than or equal to a predetermined number (m), a qubit can be determined as the fixed qubit.

[0085] Alternatively, the fixed part (150) has a bias (z) of a threshold (z) in each routine f ) among the qubits with local system energy (δE i ) can be determined as the fixed qubit (which may be one or more) whose value (e.g., e) is greater than or equal to a predetermined value. Here, the local system energy (δE i ) Instead, the qubit (which may be one or more) whose value is greater than or equal to a predetermined value or whose value is the sum of the bias coefficient (h) and the coupling strength coefficient (J) that are the components thereof, or one of the bias coefficient (h) and the coupling strength coefficient (J) can be determined as the fixed qubit.

[0086] Figure 4b shows the bias (z) threshold (z) in the iterative routine of the present invention. f ) and local system energy (δE i ) are examples of conditions for selectively fixing the increase or decrease.

[0087] Referring to FIG. 4b, the fixed part (150) has a bias (z) of a threshold (z) in each routine. f) and the local system energy (δE i ) can be performed such that the qubits (which may be more than one) are not determined as fixed qubits, which is reduced from the previous routine, and the local system energy (δE i ) can fix qubits with values ​​greater than or equal to those in the previous routine.

[0088] In addition, the fixed part (150) has a bias (z) of a threshold (z) in each routine. f ) and the local system energy (δE i ) may be determined as the fixed qubit, which may be more than one, compared to the previous routine.

[0089] Alternatively, the fixed part (150) has a bias (z) of a threshold (z) in each routine f ) is less than the local system energy (δE i ) may be determined as the fixed qubit, which may be more than one, compared to the previous routine.

[0090] Figure 5 shows the increase in the number of target qubits (H) according to the iteration routine in the annealing of the present invention. F SQF ) is an example of a graph showing the change trend of the overall system energy (E). As in the example above, when there are M qubit samples, each sample has energy according to the value of the qubit resulting from each iteration routine, and the overall system energy of the qubits (H F SQF ) corresponds to the result of adding them as in [Mathematical Formula 3].

[0091] Referring to Fig. 5, the purple graph located at the very back represents the energy values ​​of the samples when M samples are obtained by calculating the initially given S10 with a quantum annealer (110) without applying SQF, and shows the sky blue, yellow, and red histograms in the front (the 2nd, 3rd, and 4th histograms from the back) in order according to the increase of the iteration routine 1, 2, 3,,,.

[0092] As described above, as the SQF iteration progresses, the number of qubits that reduce the system energy increases, and as the qubits are fixed, the number of qubits to be re-calculated gradually decreases. As shown in FIG. 5, when a quantum annealing operation using the SQF of the present invention is performed, a solution with a lower energy value can be obtained. Therefore, as the iterative routine of the present invention increases, the qubits are fixed to the expected optimal values, so it can be confirmed that the problem size and error are reduced, and the solution for finding the optimal qubit is performed well. In other words, as the number of qubits increases, errors may overlap, resulting in a larger error in the calculation result. However, the present invention solves this problem by gradually reducing the problem size during the execution of the iterative routine, thereby reducing the error and enabling the performance of a high-performance solution for finding the optimal qubit.

[0093] FIG. 6 is a drawing for explaining an example of a method for implementing a solution device (100) according to one embodiment of the present invention.

[0094] Referring to FIG. 6, each component of the solution device (100) according to one embodiment of the present invention may be implemented using hardware, software, or a combination thereof. For example, each of the devices may be implemented in the form of a computing system (1000) having at least one processor for performing the functions / steps / processes described above, as shown in FIG. 6, or as a server on the Internet. Furthermore, the entire solution device (100) may be implemented in the form of a computing system (1000) having one processor, as shown in FIG. 6, or as a server on the Internet.

[0095] A computing system (1000) may include at least one processor (1100), memory (1300), a user interface input device (1400), a user interface output device (1500), storage (1600), and a network interface (1700) connected via a bus (1200). The processor (1100) may be a central processing unit (CPU) or a semiconductor device that executes processing on instructions stored in the memory (1300) and / or storage (1600). The memory (1300) and storage (1600) may include various types of volatile or non-volatile storage media. For example, the memory (1300) may include a read-only memory (ROM) (1310) and a random access memory (RAM) (1320).

[0096] In addition, the network interface (1700) may include a communication module such as a modem that supports wired Internet communication, wireless Internet communication such as WiFi or WiBro, mobile communication such as WCDMA, LTE, or 5G in user terminals such as smartphones, laptop PCs, or desktop PCs, or a communication module such as a modem that supports short-range wireless communication (e.g., Bluetooth, Zigbee, or WiFi).

[0097] Accordingly, the steps of a method or algorithm described in connection with the embodiments disclosed herein may be implemented directly in hardware, a software module, or a combination of the two executed by the processor (1100). The software module may reside in a non-transitory computer-readable storage medium storing computer executable instructions (i.e., memory (1300) and / or storage (1600)), such as RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM. An exemplary storage medium is coupled to the processor (1100), such that the processor (1100) can read information (code) from, and write information (code) to, the storage medium. Alternatively, the storage medium may be integral to the processor (1100). The processor and the storage medium may reside in an application-specific integrated circuit (ASIC). The ASIC may reside within the user terminal. Alternatively, the processor and storage medium may reside as separate components within the user terminal.

[0098] As described above, the solution device (100) according to the present invention can reduce the size of the operation by fixing highly reliable qubit values ​​by applying various fixation methods in consideration of bias and system energy by statistical analysis for the sampling qubits of the quantum annealer (110) for the input of the problem, and can effectively obtain a solution of optimized qubits that is faster and more reliable than before.

[0099] In the present invention, in order to maintain the expected computational complexity gains from the quantum annealer (110), the size of a given problem can be effectively reduced based on statistical analysis between post-processing tasks in repetitive routines. Without the need to identify the root causes of errors inherent in the quantum annealer (110), the optimal value of each qubit is estimated through annealing result analysis, while simultaneously reducing the size of the problem, thereby enabling qualitative improvement in the results of subsequent repetitive annealing operations. Furthermore, the present invention, when calculating a solution to a problem based on quantum annealing, utilizes statistical analysis to derive a solution to the problem using a quantum computer or the like, enabling faster and more reliable solutions to be derived than before.

[0100] As described above, the present invention has been described with specific details such as specific components and limited examples and drawings, but these are provided only to help a more general understanding of the present invention, and the present invention is not limited to the above-described examples, and those with ordinary skill in the art to which the present invention pertains may make various modifications and variations without departing from the essential characteristics of the present invention. Therefore, the spirit of the present invention should not be limited to the described examples, and all technical ideas that are equivalent or equivalent to the claims described below as well as the claims should be interpreted as being included in the scope of the rights of the present invention.

Claims

1. In a solution method performed in a processor of a device, A step of generating multiple sample sets in a quantum annealer for an input; A step of setting a threshold to be referenced for qubit fixation; A step of calculating a bias for a target qubit in each sample set of the plurality of sample sets and determining candidates whose bias is greater than or equal to the threshold; A step of recalculating the bias of the target qubits for a sample set of the corresponding subset consisting of qubits having the same fixed expected qubit value as the above candidates; A step of calculating local system energy based on the sum of the energy of each of the candidates and the coupling energy of each of the candidates, according to the fixed expected qubit value of each of the candidates and the recalculated bias; and A step of classifying the candidates into fixed qubits and qubits to be re-operated based on whether the local system energy is reduced. A solution method including:

2. In paragraph 1, A solution method for finding solutions for fixed qubits while reducing the size of the operation by repeatedly generating the qubits to be re-operated as the input to the quantum annealer and excluding the fixed qubits again.

3. In paragraph 1, When excluding the fixed qubits and repeating the process by inputting the qubits to be re-operated into the quantum annealer, A solution method in which, in the step of setting the above threshold, the set value of the threshold is set to gradually increase or decrease compared to the threshold of the previous routine.

4. In paragraph 1, A solution method for setting the threshold in the step of setting the threshold so that the threshold is proportional to the number of qubits in each sample set generated by the annealer, and the bias is proportional to the number of qubits exceeding the threshold.

5. In paragraph 1, A solution method in which the number of fixed qubits in the above classification step is predetermined as a predetermined number, and among the qubits whose bias is greater than or equal to the threshold, the predetermined number of qubits are determined as the fixed qubits in order of their difference.

6. In paragraph 1, A solution method for determining the fixed qubit in the above classification step, among the qubits having the bias greater than or equal to the threshold, a qubit having a number of coupling qubits greater than or equal to a predetermined number is determined as the fixed qubit.

7. In paragraph 1, A solution method for determining the fixed qubit in the above classification step, wherein among the qubits having the bias greater than or equal to the threshold, one or more qubits having the local system energy greater than or equal to a predetermined value are determined as the fixed qubit.

8. In paragraph 1, When excluding the fixed qubits and repeating the process by inputting the qubits to be re-operated into the quantum annealer, A solution method for determining the fixed qubit in the above classification step, wherein one or more qubits having the bias greater than the threshold and the local system energy reduced compared to the previous routine are not determined as the fixed qubit.

9. In paragraph 1, When excluding the fixed qubits and repeating the process by inputting the qubits to be re-operated into the quantum annealer, A solution method for determining the fixed qubit in the above classification step, wherein one or more qubits having the bias greater than the threshold and the local system energy increased compared to the previous routine are determined as the fixed qubit.

10. In paragraph 1, When excluding the fixed qubits and repeating the process by inputting the qubits to be re-operated into the quantum annealer, A solution method for determining the fixed qubit in the above classification step, wherein one or more qubits having the bias less than the threshold and the local system energy increased compared to the previous routine are determined as the fixed qubit.

11. In paragraph 1, Before the step of calculating the bias, a step of determining a combination of qubits on a qubit sequence as the target qubit for the plurality of sample sets A solution method including more.

12. In paragraph 11, In the step of determining the above qubit combination as the target qubit, The combination is determined based on the number of coupled qubits that the two coupled qubits have in common, or A solution method for determining the combination of two coupled qubits, wherein the qubits have values ​​greater than the standard deviation of the distribution of the bias coefficient or coupling strength coefficient of all qubits.

13. In paragraph 11, In the step of determining the above qubit combination as the target qubit, The combination is determined based on the number of coupled qubits that two uncoupled qubits have in common, or A solution method for determining the combination of two uncoupled qubits, wherein the qubits have a value greater than or equal to the standard deviation of the distribution of the bias coefficient or coupling strength coefficient of all qubits.

14. In paragraph 11, In the step of determining the above qubit combination as the target qubit, A solution method for determining the combination of three or more qubits belonging to a cluster classified by a given clustering method regardless of coupling.

15. A quantum annealer that generates multiple sets of samples for the input; A threshold setting unit that sets a threshold to be referenced for qubit fixation; A bias calculation and combination setting unit for calculating a bias for a target qubit in each sample set of the plurality of sample sets and determining candidates whose bias is greater than or equal to the threshold; An adjacent qubit bias recalculation unit that recalculates the bias of target qubits for a sample set of the corresponding subset consisting of qubits having the same fixed expected qubit value as the above candidates; A system energy calculation unit that calculates local system energy based on the sum of the energy of each candidate and the coupling energy of each candidate, according to the fixed expected qubit value of each of the candidates and the recalculated bias; and A fixed part that classifies the candidates into fixed qubits and qubits to be re-operated depending on whether the local system energy is reduced. Solution device including.

16. In a recording medium having recorded thereon a computer-readable code for performing a solution function performed by the processor of the device, Ability to generate multiple sample sets from a quantum annealer for an input; Ability to set a threshold to be referenced for qubit pinning; A step of calculating a bias for a target qubit in each sample set of the plurality of sample sets and determining candidates whose bias is greater than or equal to the threshold; A function to recalculate the bias of the target qubits for a sample set of the corresponding subset consisting of qubits having the same fixed expected qubit value as the above candidates; A function for calculating local system energy based on the sum of the energy of each of the candidates and the coupling energy of each of the candidates, based on the fixed expected qubit value of each of the candidates and the recalculated bias; and A function to classify the candidates into fixed qubits and qubits to be re-operated based on whether the local system energy is reduced. A recording medium for performing .

Images