WO2016157333A1 - 計算機、及び演算プログラム - Google Patents
計算機、及び演算プログラム Download PDFInfo
- Publication number
- WO2016157333A1 WO2016157333A1 PCT/JP2015/059765 JP2015059765W WO2016157333A1 WO 2016157333 A1 WO2016157333 A1 WO 2016157333A1 JP 2015059765 W JP2015059765 W JP 2015059765W WO 2016157333 A1 WO2016157333 A1 WO 2016157333A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- time
- spin
- variable
- calculation
- pina
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N10/00—Quantum computing, i.e. information processing based on quantum-mechanical phenomena
Definitions
- the present invention relates to a computer that enables high-speed computation for inverse problems and combinatorial optimization problems that require exhaustive search.
- Big data has many problems that require complex analysis. For example, when a certain result is obtained, there is a case where it is desired to find the cause. This is called an inverse problem. The more complicated the phenomenon is, the more difficult it is to find the cause, and there is generally no efficient algorithm for obtaining the initial value from the result. In the worst case, an exhaustive search must be performed for the initial value. This is one of the difficult problems in big data. Alternatively, there are many problems such as selecting an optimal solution from many options based on big data. In this case, if all possibilities are taken into account, an exhaustive search becomes necessary. against this background, there is a need for a computer that can efficiently solve problems that require exhaustive search.
- a quantum computer consists of basic elements called qubits, and realizes "0" and "1" simultaneously. Therefore, all solution candidates can be calculated as initial values at the same time, and there is a possibility that an exhaustive search can be realized.
- quantum computers need to maintain quantum coherence over the entire computation time, and there is no prospect of realizing this.
- Non-Patent Document 1 a technique called adiabatic quantum computation has been attracting attention.
- H ⁇ p be the Hamiltonian of the physical system that sets the problem.
- the Hamiltonian is not H ⁇ p at the start of the calculation, but another Hamiltonian H ⁇ 0 whose ground state is clear and easy to prepare.
- H ⁇ p the Hamiltonian of the physical system that sets the problem.
- adiabatic quantum computation can be applied to problems that require exhaustive search, and reaches a solution in a one-way process.
- the calculation process needs to follow the Schrödinger equation (2), it is necessary to maintain quantum coherence as in the case of the quantum computer.
- quantum computers repeat the gate operation between 1 qubits or 2 qubits
- adiabatic quantum computations interact all over the entire qubit system, and the concept of coherence is Different. For example, consider a gate operation to a certain qubit. At this time, if there is interaction between the qubit and other qubits, it causes decoherence, but in adiabatic quantum computation, all qubits interact simultaneously, so in this case Does not become decoherence. Reflecting this difference, adiabatic quantum computation is considered to be more robust against decoherence than quantum computers.
- adiabatic quantum computation is effective for difficult problems that require exhaustive search.
- it still requires quantum coherence, and when using a superconducting flux qubit, a cryogenic cooling device is also required. It is a problem to be solved to eliminate these two requirements and to provide a practical computer.
- An object of the present invention is to solve the above-described problems and provide a computer and a calculation program that do not require quantum coherence or a cryogenic cooling device.
- the spin is a variable in the operation, and the problem to be solved is set by the interaction between spins and the local field acting for each spin.
- Each spin develops in time as the direction is determined according to the effective magnetic field determined by the external magnetic field at each site and all the interactions between the spins at time t.
- the spin direction is not completely aligned with the effective magnetic field, but the direction is corrected in a quantum mechanical manner so that the system substantially maintains the ground state.
- the vibration related to the direction of the spin in the time evolution is suppressed, and the convergence of the solution is improved.
- FIG. 20 is a block diagram illustrating a configuration example of a computer according to an eighth embodiment.
- FIG. 20 is a block diagram illustrating another configuration example of the computer according to the eighth embodiment. It is a block diagram which shows the example of 1 structure of the local field response calculating apparatus contained in the computer based on Example 9.
- FIG. It is a block diagram which shows one structural example of the optical part of FIG.
- FIG. It is a block diagram which shows the example of 1 structure of the local field response calculating apparatus contained in the computer based on Example 10.
- FIG. It is a block diagram which shows the example of 1 structure of the local field response calculating apparatus contained in the computer based on Example 11.
- FIG. It is a block diagram which shows the example of 1 structure of the local field response calculating apparatus contained in the computer based on Example 12.
- FIG. 12 shows the example of 1 structure of the local field response calculating apparatus contained in the computer based on Example 12.
- notations such as “first”, “second”, and “third” are attached to identify the constituent elements, and do not necessarily limit the number or order.
- a number for identifying a component is used for each context, and a number used in one context does not necessarily indicate the same configuration in another context. Further, it does not preclude that a component identified by a certain number also functions as a component identified by another number.
- Adiabatic quantum computation also known as quantum annealing, is an evolution of the classic annealing concept to quantum mechanics.
- the adiabatic quantum computation is inherently capable of classical operation, and can be interpreted as a quantum mechanical effect added to improve performance in terms of high speed and correct solution rate. Therefore, in the present invention, the arithmetic unit itself is classical, and by introducing a parameter determined in a quantum mechanical manner into the arithmetic process, a classical arithmetic method and apparatus including a quantum mechanical effect is realized.
- the following embodiment describes a classical algorithm for obtaining a ground state as a solution while explaining the relevance to adiabatic quantum computation, and a device for realizing it.
- a typical form described in the following embodiments is a computer that includes a calculation unit, a storage unit, and a control unit, and performs calculations while exchanging data between the storage unit and the calculation unit under the control of the control unit.
- the coefficient g pinb is, for example, a value from 50% to 200% of the average value of
- the correction term ⁇ g j ′ is added to g j ′ only for a certain site j ′, and the size of g j ′ is increased only for the site j ′. it can.
- the correction term ⁇ g j ′ is, for example, a value of 10% to 100% of the average value of
- Example 1 the principle of the present invention is described through a transition from the quantum mechanical description to the classical form.
- J ij and g j are task setting parameters, and ⁇ ⁇ j z is the z component of Pauli's spin matrix and takes an eigenvalue of ⁇ 1.
- i and j represent spin sites.
- Ising spin is a variable that can take only ⁇ 1 as a value.
- the eigenvalue of ⁇ ⁇ j z is ⁇ 1, so it is an Ising spin system.
- the Ising spin in equation (3) does not have to be a literal spin, and can be physically anything as long as the Hamiltonian is described by equation (3).
- the logic circuit high and low can be associated with ⁇ 1
- the longitudinal and transverse polarization of light can be associated with ⁇ 1
- the phase of 0, ⁇ can be associated with ⁇ 1. is there.
- Equation (4) corresponds to the application of a transverse magnetic field, and the ground state is when all spins are in the x direction ( ⁇ > 0).
- the problem setting Hamiltonian was defined as an Ising spin system with only the z component, but the x component of the spin appears in Equation (4). Therefore, the spin in the calculation process is not Ising but vector-like (Bloch vector).
- t 0, the start is made with the Hamiltonian in the equation (4), but the Hamiltonian is gradually changed with the progress of the time t, and finally the Hamiltonian described in the equation (3) is used to obtain the ground state as a solution.
- ⁇ ⁇ represents the three components of Pauli's spin matrix as a vector.
- the spin direction can be defined by ⁇ ⁇ j z > / ⁇ ⁇ j x >, if the spin direction follows the effective magnetic field, the spin direction is determined by equation (7).
- Equation (7) is a quantum mechanical description but has an expected value, so it is a relational equation for classical quantities, unlike equations (1)-(6).
- equations (1)-(6) there is no non-local correlation (quantum drift) of quantum mechanics, so the spin direction should be completely determined by the local field at each site, and Equation (7) determines the behavior of the classical spin system. Since there is a nonlocal correlation in the quantum system, Equation (7) is modified, but this will be described in Example 2 and below, and in this example, Equation (7) is used to describe the basic form of the invention. Describes the classical system determined by).
- FIG. 1 shows a timing chart (procedure 100) for obtaining the ground state of the spin system.
- the spin of site j is represented by s j instead of ⁇ ⁇ j .
- the effective magnetic field B j in FIG. 1 is a classical quantity.
- a right effective magnetic field B j is applied at all sites, and all spins s j are initialized to the right.
- the magnetic field in the z-axis direction and the spin-to-spin interaction are gradually applied.
- the time t is continuous, but it can be made discrete in the actual calculation process to improve convenience. The discrete case is described below.
- the spin according to the present invention is a vector spin because not only the z component but also the x component is added.
- the behavior as a vector can also be understood from FIG.
- a three-dimensional vector having a size of 1 this is called a Bloch vector and a state can be described by a point on the sphere
- only two dimensions are used. (The state can be described by a point on the circle).
- the two-dimensional spin vector can be described only by a semicircle. If s j z is specified by [ ⁇ 1, 1], the two-dimensional spin vector is determined by one variable of s j z . Therefore, although the spin of the present invention is a two-dimensional vector, it can also be expressed as a one-dimensional continuous variable with a range of [ ⁇ 1, 1].
- Equation (8) is a rewrite of Equation (7) into a notation related to classical quantities, it does not have a ⁇ •> symbol.
- the magnitude of the spin vector is 1.
- FIG. 2 shows a summary of the above algorithm in a flowchart.
- t m ⁇ .
- variable interaction J ij The strength of the correlation between the temperature and the activities of public facilities and private houses is expressed through the variable interaction J ij .
- ⁇ ⁇ j z is a variable representing power distribution, but correlates with the movement of people and the opening of public facilities. Therefore, depending on the solution obtained, it can be interpreted as “a certain public facility should be closed”.
- the above is a simple example of expressing a specific problem with Expression (3).
- the specific problems to which this embodiment can be applied are not limited to the power supply management problems exemplified above, but include travel route optimization, vehicle guidance for avoiding traffic congestion, circuit design, product supply management, scheduling, financial asset selection, etc. It can be applied to solve many problems.
- Example 1 the expected value was taken based on the quantum mechanical equation to shift to the classical quantity, and the algorithm based on the classical quantity was explained using FIG. 1 and FIG. Since Example 1 was mainly intended to explain the algorithm of the present invention, it was described without including quantum mechanical effects. However, if the quantum mechanical effect is added, it can be expected to improve the accuracy rate and the calculation speed. Therefore, in the second embodiment, although the algorithm itself is classic, a method of adding a correction parameter based on quantum mechanics to improve performance will be described.
- the features of quantum mechanics include linear superposition and quantum distortion (nonlocal correlation). For example, consider a qubit that can take two states,
- the linear superposition state is a sum of two states such as
- ⁇ > ⁇
- 0>, and if s j z (t k ) ⁇ 1, the state is
- 0> i
- 1> ⁇ i
- ⁇ > i ⁇
- the magnitude of the spin vector is not preserved to 1 when there is quantum distortion although it is an example.
- the magnitude of the spin vector should be a constant 1, but if there is a quantum droop, the spin vector will not have a magnitude of 1.
- s j z (t with ⁇ defined as tan ⁇ ⁇ B j z (t)> / ⁇ B j x (t)> as a parameter.
- this method does not reflect the nature of the quantum drowning inherent in this system. Therefore, let us consider reflecting the nature of quantum drowning.
- the correction parameters r s and r B originate from quantum distortion and are preferably finely controlled depending on t k , s j z (t k ), and s j x (t k ).
- r B is an amount that can change the sign and reflects the quantum blur best.
- r s and r B are not given site dependency, but if the characteristics of each site are known in advance, r s and r B are set accordingly. If you make it site-dependent, you can expect an improvement in the correct answer rate.
- FIG. 3 shows a flowchart when r s and r B are introduced. The difference from the flowchart of FIG. 2 is that steps 103, 105, and 107 are changed to steps 103a, 105a, and 107a including correction parameters r s and r B , respectively.
- Equation (12A) the relaxation term (pinning term) is added to Equation (10) to make the effective magnetic field into Equation (12A).
- the third term is a relaxation term.
- the relaxation term is an additional term for improving the convergence of the solution and needs to be sufficiently smaller than
- Relaxation term (third term) of the formula (12A) does not depend on the size of depends only on the sign of s j z s j z.
- the first term depends on the size of s i z . Therefore, there is also a method of making the third term depend on the size of s j z . In that case, Expression (12B) is obtained.
- the coefficient g pinb is typically about the average value of
- s j z ⁇ 0.
- the relaxation term is also useful for this induction.
- one site referred to as j site
- s j z (t 0 ) 1
- the orientation of other sites is determined at the initial stage based on the sign of J ij with reference to j site. To do. In this way, an appropriate zeroth approximation is set, and then converges to one of the correct answers through subsequent local field response calculations.
- the relaxation term contributes as follows.
- FIG. 4 shows a flowchart when a relaxation term is added.
- 4A corresponds to the equation (12A)
- FIG. 4B corresponds to the equation (12B).
- steps 102, 104, and 106 are changed to steps 102b, 104b, and 106b including a relaxation term.
- ⁇ g j ′ is added to the local field term g j ′ in equations (12A) and (12B) to obtain g j ′ ⁇ g j ′ + ⁇ g j ′ .
- the relaxation term improves the convergence of the solution and converges to one of the solutions even when the number of degeneracy is large.
- the relaxation term improves the convergence of the solution and converges to one of the solutions even when the number of degeneracy is large.
- formulas (12A) and (12B) only the local field term ⁇ g j ′ is added to only one site j ′ to strongly induce only the spin at the j ′ site in a certain direction. This leads to one of the degenerate solutions based on the j 'site.
- the convergence of the solution is improved by introducing a relaxation term.
- FIG. 5 shows a flowchart of an example of the algorithm in this case. 5A-5D correspond to FIGS. 4A-4D, respectively.
- FIG. 6 is a flowchart illustrating an example of an algorithm related to a final solution determination method.
- ⁇ ⁇ j z in Equation (3) has an eigenvalue of ⁇ 1, it is determined at each stage of the calculation process whether the eigenvalue of ⁇ ⁇ j z is 1 or -1 according to the sign of s j z .
- the fourth method terminates the operation when all s j z converge as in the second method.
- the final solution is not determined from s j z at the time of termination, but the energy at each time is calculated in the same way as the third method, and s j z at the time when the lowest energy is given ( The final solution is determined from t k ′ ) (125). The user decides which method to use.
- the second method evaluates the possibility of spin reversal at each time and sets a time interval based on the evaluation. For example: If the magnitude of
- a specific example of the time interval determination method is as follows. Let the minimum time interval be ⁇ t min .
- the minimum value of the time interval is ⁇ t min and the maximum value is 100 ⁇ ⁇ t min .
- the user decides which method to use.
- Example 1-7 the calculation principle and calculation algorithm have been described.
- a configuration example of a computer that operates this algorithm as a program will be described first.
- FIG. 7 shows an example of the computer configuration of this embodiment.
- FIG. 7 has basically the same configuration as that of an ordinary computer. Data is transferred from the main storage device 201 serving as a storage unit to the arithmetic device 202 serving as a calculation unit, and the data is returned to the main storage device 201 after the calculation. The operation is repeated by repeating this.
- the control tower at that time is a control device 203 as a control unit.
- the calculation in the configuration of the present embodiment is executed by the calculation device 202 for each time and for each site according to the flow of FIGS.
- the program executed by the arithmetic device 202 is stored in the main storage device 201 which is a storage unit. If the main storage device 201 has insufficient storage capacity, the auxiliary storage device 204, which is also a storage unit, is used.
- An input device 205 is used to input data and programs, and an output device 206 is used to output the results.
- the input device 205 includes an interface for network connection in addition to a manual input device such as a keyboard. This interface also serves as an output device. 7 applies the algorithm described in Embodiment 1-7 as a program, but the apparatus itself uses a normal computer.
- Fig. 8 shows a configuration diagram that achieves this.
- a local field response calculation device 1000 is included as a calculation unit. This is a difference in comparison with the configuration of FIG.
- the local field response calculation device 1000 is a dedicated calculation unit for executing the above-described algorithm, and performs only the local field response calculation described in the embodiment 1-7, and general processing is performed by the calculation device 202.
- Information used in the local field response calculation device 1000 is transferred from the main storage device 201.
- Necessary information includes task setting parameters such as interaction J ij between variables and local field g j , time parameter t k , correction parameters r s and r B related to quantum distortion , and coefficients g pina and g pinb of relaxation terms Etc. Processing such as synchronization is the same as that of the computer having the configuration shown in FIG.
- the transfer value from the local field response arithmetic unit 1000 in the intermediate stage of the calculation is used for, for example, calculating the Ising spin Hamiltonian energy in the intermediate stage of the calculation and evaluating s min (t k ) 2 and s ave (t k ) 2 . Use.
- the final solution is determined using the energy value of the Ising spin Hamiltonian at the intermediate stage. If it is determined whether s j zd is +1 or ⁇ 1, the calculation of the energy value of the Ising spin Hamiltonian is simple, and a normal arithmetic unit 202 is used.
- the local field response calculation device 1000 is a dedicated calculation unit for the local field response calculation described in the embodiment 1-4, and a normal calculation device 202 is used for other processing.
- the local field response calculation apparatus 1000 described in the eighth embodiment can be realized by various methods. In this embodiment, first, a method for effectively using the parallelism of light will be described.
- FIG. 9 shows an overall view.
- Information such as ⁇ , g j , and g pina required for calculation is sent from the main storage device 201 to the control unit 1100, and information about J ij is sent to the variable mask 1120.
- the output intensity of the LED array 1110 reflects the value of the variable s j z .
- phase information is not used. Therefore, an incoherent light source such as an LED array is used as the light source 1110.
- the measurement time in the detector array 1130 is sufficiently long so that the interference effect between the output lights of the LD array does not appear.
- the output light from the LED array 1110 spreads only in the horizontal direction, attenuates the amount of light according to J ij in the variable mask 1120, and this time converges only in the vertical direction and converts it into an electrical signal by the detector array 1130.
- FIG. 10 is a block diagram showing a configuration example of the optical unit in FIG.
- the optical system from the LED array 1110 to the detector array 1130 uses, for example, a lens system as shown in FIG.
- , and the difference between the two outputs s j z s j z + ⁇
- s j z + + s
- Two detector arrays 1130 are also paired in conjunction with the light source side. As a result, it is possible to cope with the variable mask 1120 that cannot take a negative value.
- the signal to be obtained by the detector array 1130 is b j z ⁇ i J ij s i z .
- Each of the detector array 1130 and two pair b j z + ⁇ i ( J ij + s j z + + J ij - s j z-) and
- ⁇ i (J ij +
- s j z + ), and the difference between the detectors is detected b j z b j z + ⁇
- b j z + + b j z -Becomes a signal.
- B j z can be obtained based on equations (10), (12A), and (12B) by adding the terms g j and g pina .
- This calculation is performed by the control unit 1100.
- the calculation to obtain the s j z from B j z also performed by the control unit 1100, sends the value of s j z to the LED array 1110.
- the control unit 1100 repeats the same processing, and is a dedicated circuit for that purpose. Further, s j z at each time is transferred to the main storage device 201 and used for analysis.
- the optical system from the LED array 1110 to the detector array 1130 was free space. This part can also be realized by an optical system using a waveguide.
- FIG. 11 shows such a case.
- the output light from the LED array (LD array) 1110 is divided by a demultiplexer 1115, passes through a variable attenuator 1120 (variable mask in FIG. 9) whose transmittance is set based on J ij, and is collected by a multiplexer 1125.
- the light is received and received by the detector array 1130. Only the spatial optical system in FIG. 9 is changed to a waveguide optical system, and the operation principle is the same. Accordingly, the LED array (LD array) 1110 represents s j z in pairs, and the detector array 1130 also operates in pairs.
- Example 9 s j z was expressed by combining two light sources. If s j z + and s j z ⁇ are expressed using polarized light, one light source can be used for s j z .
- FIG. 12 shows such a case.
- Each LD in the light source 1111 is for representing s j z .
- Acos ( ⁇ / 4 + ⁇ ).
- the polarization-modulated light is separated by the polarization separator 1113 and guided to the branching filter 1115.
- the operation from the duplexer 1115 to the detector 1130 is the same as that in FIG. Since the detector 1130 operates in pairs as in FIG.
- the local field response calculation device 1000 can be realized not only by a method using light as in the ninth to eleventh embodiments but also by an electric circuit.
- FIG. 13 shows a configuration example in that case.
- Information on each spin s i z is temporarily held in the buffer array 1210.
- Information on J ij is stored in the memory 1221.
- g pinb is held at 1221
- J ii g pinb .
- the cells for each s i z in the buffer arrays 1210 and 1230 are composed of multiple bits, and are assumed to be a pseudo continuous quantity.
- the influence of temperature in the present invention is estimated as follows. Bit manipulation is performed by the LED (LD) array 1110, the polarization modulator 1112, and the buffer arrays 1210 and 1230.
- the present invention is not limited to the above-described embodiment, and includes various modifications.
- a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Operations Research (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
以下、図1の手順100で模式的に示した手順に従い、スピンと有効磁場を交互に求めていく。
201 主記憶装置
202 演算装置
203 制御装置
204 補助記憶装置
205 入力装置
206 出力装置
1000 局所場応答演算装置
1100 制御部
1110 LED (LD)アレイ
1111 LD
1112 偏波変調器
1113 偏光分離器
1115 分波器
1120 可変マスク(可変減衰器)
1125 合波器
1130 検出器アレイ
1131 差動増幅器
1200 制御部
1210 バッファアレイ
1220 演算部
1221 メモリ
1230 バッファアレイ
Claims (15)
- 演算部、記憶部、制御部を具備し、前記制御部の制御により、前記記憶部と前記演算部との間でデータをやり取りしながら演算を行う計算機であって、
N個の変数sj z(j = 1, 2, …, N)が-1 ≦ sj z ≦ 1の値域を取り、局所場gjと変数間相互作用Jij (i, j = 1, 2, …, N)によって課題の設定がなされ、
前記演算部では、時刻をm分割して離散的にt = t0 (t0 = 0)からtm (tm = τ)まで演算するものとし、
各時刻tkにおける前記変数sj z(tk)を求めるに当たり、前時刻tk-1の変数si z(tk-1) (i = 1, 2, .., N)の値と緩和項の係数gpinaあるいはgpinbを用いてBj z(tk) = {ΣiJijsi z(tk-1) + gj + sgn(sj z(tk-1))・gpina}・tk/τあるいはBj z(tk) = {ΣiJijsi z(tk-1) + gj + gpinb・sj z(tk-1)}・tk/τを求め、 前記変数sj z(tk)の値域が-1 ≦ sj z(tk) ≦ 1になるように関数fを定めてsj z(tk) = f(Bj z(tk) , tk)とし、
時刻ステップをt = t0からt = tmに進めるにつれて前記変数sj zを-1あるいは1に近づけ、
最終的にsj z < 0ならばsj zd = -1、sj z > 0ならばsj zd = 1として解を定める、
ことを特徴とする計算機。 - 前記関数fに関して、
ある定数γを用いてBj x(tk) = γ(1 - tk/τ)とし、tanθ = Bj z(tk)/Bj x(tk)によりθを定義し、前記sj z(tk)をsj z(tk) = sinθによって定めることとし、従って前記関数fがf(Bj z(tk) , tk) = sin(arctan(Bj z(tk)/Bj x(tk)))となることを特徴とする請求項1記載の計算機。 - 前記関数fに関して補正パラメタrs及びrBを追加し、
tanθ = rB・Bj z(tk)/Bj x(tk)によりθを定義し、sj z(tk) = rs・sinθによって前記sj z(tk)を定めることとし、従って前記関数fがf(Bj z(tk) , tk) = rs・sin(arctan(rB・Bj z(tk)/Bj x(tk)))となることを特徴とする請求項1記載の計算機。 - 前記補正パラメタrs及びrBを、前記tkと前記Bj z(tk)に依存させることを特徴とする請求項3記載の計算機。
- 前記時刻tk各々においてsj z(tk) < 0ならばsj zd(tk) = -1、sj z(tk) > 0ならばsj zd(tk) = 1としてHp(tk) = - Σi>jJijsi zd(tk)sj zd(tk) - Σjgjsj zd(tk)を各時刻tkにおいて計算し、Hp(tk)が最小値となった時刻tk’におけるsj zd(tk’)を最終解とする、
ことを特徴とする請求項1記載の計算機。 - 前記係数gpinaは|Jij|の平均値の1%から50%の値であることを特徴とする請求項1記載の計算機。
- 前記課題設定の局所場gj (j = 1, 2, …, N)に関して、
gj ≠ 0の成分があれば前記記載のBj z(tk)をBj z(tk) = {ΣiJijsi z(tk-1) + gj}・tk/τにより定め、N個すべてでgj = 0ならば前記記載のBj z(tk)をBj z(tk) = {ΣiJijsi z(tk-1) + sgn(sj z(tk-1))・gpina}・tk/τあるいはBj z(tk) = {ΣiJijsi z(tk-1) + gj + gpinb・sj z(tk-1)}・tk/τにより定めることを特徴とする請求項1記載の計算機。 - 演算部で実行される演算プログラムであって、
N個の変数sj z(j = 1, 2, …, N)が-1 ≦ sj z ≦ 1の値域を取り、局所場gjと変数間相互作用Jij (i, j = 1, 2, …, N)によって課題の設定がなされ、
時刻をm分割して離散的にt = t0 (t0 = 0)からtm (tm = τ)まで演算するものとし、
各時刻tkにおける変数sj z(tk)を求めるに当たり、前時刻tk-1の変数si z(tk-1) (i = 1, 2, .., N)の値と緩和項の係数gpinaあるいはgpinbを用いてBj z(tk) = {ΣiJijsi z(tk-1) + gj + sgn(sj z(tk-1))・gpina}・tk/τあるいはBj z(tk) = {ΣiJijsi z(tk-1) + gj + gpinb・sj z(tk-1)}・tk/τを求め、 sj z(tk)の値域が-1 ≦ sj z(tk) ≦ 1になるように関数fを定めてsj z(tk) = f(Bj z(tk) , tk)とし、
時刻ステップをt = t0からt = tmに進めるにつれてsj zを-1あるいは1に近づけ、
最終的にsj z < 0ならばsj zd = -1、sj z > 0ならばsj zd = 1として解を定める、
ことを特徴とする演算プログラム。 - 前記関数fに関して、
ある定数γを用いてBj x(tk) = γ(1 - tk/τ)とし、tanθ = Bj z(tk)/Bj x(tk)によりθを定義し、前記sj z(tk)をsj z(tk) = sinθによって定めることとし、従って前記関数fがf(Bj z(tk) , tk) = sin(arctan(Bj z(tk)/Bj x(tk)))となることを特徴とする請求8記載の演算プログラム。 - 前記関数fに関して補正パラメタrs及びrBを追加し、
tanθ = rB・Bj z(tk)/Bj x(tk)によりθを定義し、sj z(tk) = rs・sinθによって前記sj z(tk)を定めることとし、従って前記関数fがf(Bj z(tk) , tk) = rs・sin(arctan(rB・Bj z(tk)/Bj x(tk)))となることを特徴とする請求項8記載の演算プログラム。 - 前記補正パラメタrs及びrBを、前記tkと前記Bj z(tk)に依存させることを特徴とする請求項10記載の演算プログラム。
- 前記時刻tk各々おいてsj z(tk) < 0ならばsj zd(tk) = -1、sj z(tk) > 0ならばsj zd(tk) = 1としてHp(tk) = - Σi>jJijsi zd(tk)sj zd(tk) - Σjgjsj zd(tk)を各時刻tkにおいて計算し、Hp(tk)が最小値となった時刻tk’におけるsj zd(tk’)を最終解とする、
ことを特徴とする請求項8記載の演算プログラム。 - 前記係数gpinaは|Jij|の平均値の1%から50%の値であることを特徴とする請求項8記載の演算プログラム。
- 前記課題設定の局所場gj (i, j = 1, 2, …, N)に関して、
gj ≠ 0の成分があれば前記記載のBj z(tk)をBj z(tk) = {ΣiJijsi z(tk-1) + gj}・tk/τにより定め、N個すべてでgj = 0ならば前記記載のBj z(tk)をBj z(tk) = {ΣiJijsi z(tk-1) + sgn(sj z(tk-1))・gpina}・tk/τあるいはBj z(tk) = {ΣiJijsi z(tk-1) + gj + gpinb・sj z(tk-1)}・tk/τにより定めることを特徴とする請求項8記載の演算プログラム。 - 演算部、記憶部、制御部を具備し、前記制御部の制御により、前記記憶部と前記演算部との間でデータをやり取りしながら演算を行う計算機であって、
当該計算機は、局所場応答演算装置を備え、
前記局所場応答演算装置では、
N個の変数sj z(j = 1, 2, …, N)が-1 ≦ sj z ≦ 1の値域を取り、局所場gjと変数間相互作用Jij (i, j = 1, 2, …, N)によって課題の設定がなされ、
前記演算部では、時刻をm分割して離散的にt = t0 (t0 = 0)からtm (tm = τ)まで演算するものとし、
各時刻tkにおける前記変数sj z(tk)を求めるに当たり、前時刻tk-1の変数si z(tk-1) (i = 1, 2, .., N)の値と緩和項の係数gpinaあるいはgpinbを用いてBj z(tk) = {ΣiJijsi z(tk-1) + gj + sgn(sj z(tk-1))・gpina}・tk/τあるいはBj z(tk) = {ΣiJijsi z(tk-1) + gj + gpinb・sj z(tk-1)}・tk/τを求め、 sj z(tk)の値域が-1 ≦ sj z(tk) ≦ 1になるように関数fを定めてsj z(tk) = f(Bj z(tk) , tk)とし、
時刻ステップをt = t0からt = tmに進めるにつれてsj zを-1あるいは1に近づけ、
最終的にsj z < 0ならばsj zd = -1、sj z > 0ならばsj zd = 1として解を定める、
処理を行い、
当該局所場応答演算装置における処理は、光学系で並列的に処理される複数の光信号の変調を利用して行われるか、もしくは、バッファアレイに蓄積されたデータを並列的に処理することにより行われる、
ことを特徴とする計算機。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US15/561,566 US10318607B2 (en) | 2015-03-27 | 2015-03-27 | Computer and computing program |
PCT/JP2015/059765 WO2016157333A1 (ja) | 2015-03-27 | 2015-03-27 | 計算機、及び演算プログラム |
JP2017508862A JP6291129B2 (ja) | 2015-03-27 | 2015-03-27 | 計算機、及び演算プログラム |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2015/059765 WO2016157333A1 (ja) | 2015-03-27 | 2015-03-27 | 計算機、及び演算プログラム |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016157333A1 true WO2016157333A1 (ja) | 2016-10-06 |
Family
ID=57004892
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/059765 WO2016157333A1 (ja) | 2015-03-27 | 2015-03-27 | 計算機、及び演算プログラム |
Country Status (3)
Country | Link |
---|---|
US (1) | US10318607B2 (ja) |
JP (1) | JP6291129B2 (ja) |
WO (1) | WO2016157333A1 (ja) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019079225A (ja) * | 2017-10-24 | 2019-05-23 | 株式会社日立製作所 | 計算機及び計算方法 |
WO2020090674A1 (ja) | 2018-10-31 | 2020-05-07 | 株式会社日立製作所 | 情報提供装置および情報提供方法 |
WO2020235300A1 (ja) | 2019-05-20 | 2020-11-26 | 株式会社日立製作所 | ポートフォリオ作成支援装置およびポートフォリオ作成支援方法 |
WO2021171746A1 (ja) | 2020-02-28 | 2021-09-02 | 株式会社日立製作所 | ポートフォリオ作成支援装置、ポートフォリオ作成支援方法、及びポートフォリオ作成支援システム |
US11868967B2 (en) | 2019-04-05 | 2024-01-09 | Hitachi, Ltd. | Schedule creation assisting device and schedule creation assisting method |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10332023B2 (en) | 2017-09-22 | 2019-06-25 | International Business Machines Corporation | Hardware-efficient variational quantum eigenvalue solver for quantum computing machines |
US10452990B2 (en) * | 2017-11-28 | 2019-10-22 | International Business Machines Corporation | Cost function deformation in quantum approximate optimization |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009524857A (ja) * | 2006-01-27 | 2009-07-02 | ディー−ウェイブ システムズ,インコーポレイテッド | 断熱量子計算の方法 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6445246B2 (ja) * | 2014-03-27 | 2018-12-26 | 株式会社日立製作所 | 情報処理装置及び情報処理方法 |
WO2017037331A1 (en) * | 2015-09-04 | 2017-03-09 | Nokia Technologies Oy | Method and apparatus for optimization |
-
2015
- 2015-03-27 US US15/561,566 patent/US10318607B2/en active Active
- 2015-03-27 WO PCT/JP2015/059765 patent/WO2016157333A1/ja active Application Filing
- 2015-03-27 JP JP2017508862A patent/JP6291129B2/ja active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009524857A (ja) * | 2006-01-27 | 2009-07-02 | ディー−ウェイブ システムズ,インコーポレイテッド | 断熱量子計算の方法 |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019079225A (ja) * | 2017-10-24 | 2019-05-23 | 株式会社日立製作所 | 計算機及び計算方法 |
WO2020090674A1 (ja) | 2018-10-31 | 2020-05-07 | 株式会社日立製作所 | 情報提供装置および情報提供方法 |
US11868967B2 (en) | 2019-04-05 | 2024-01-09 | Hitachi, Ltd. | Schedule creation assisting device and schedule creation assisting method |
WO2020235300A1 (ja) | 2019-05-20 | 2020-11-26 | 株式会社日立製作所 | ポートフォリオ作成支援装置およびポートフォリオ作成支援方法 |
WO2021171746A1 (ja) | 2020-02-28 | 2021-09-02 | 株式会社日立製作所 | ポートフォリオ作成支援装置、ポートフォリオ作成支援方法、及びポートフォリオ作成支援システム |
Also Published As
Publication number | Publication date |
---|---|
US20180089144A1 (en) | 2018-03-29 |
JPWO2016157333A1 (ja) | 2017-07-13 |
US10318607B2 (en) | 2019-06-11 |
JP6291129B2 (ja) | 2018-03-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6291129B2 (ja) | 計算機、及び演算プログラム | |
JP6483256B2 (ja) | 計算機 | |
Wiersema et al. | Exploring entanglement and optimization within the hamiltonian variational ansatz | |
US20200242501A1 (en) | Embedding electronic structure in controllable quantum systems | |
US10789329B2 (en) | Systems and methods for removing unwanted interactions in quantum devices | |
US10846366B1 (en) | Selecting parameters for a quantum approximate optimization algorithm (QAOA) | |
JP7149504B2 (ja) | 量子コンピューティング・マシンのための高ハードウェア効率変分量子固有値ソルバを実現するためのシステム、方法、量子コンピューティング・デバイスおよびコンピュータ・プログラム | |
Salehi et al. | Unconstrained binary models of the travelling salesman problem variants for quantum optimization | |
US20200342345A1 (en) | Classification using quantum neural networks | |
EP3991104A1 (en) | Method of computing a solution to a computational problem using a quantum system and apparatus for computing solutions to computational problems | |
CN116249992A (zh) | 囚禁离子量子计算机的优化电路编译器 | |
Fang et al. | Minimizing minor embedding energy: an application in quantum annealing | |
Wang et al. | Quantum computing in a statistical context | |
Lu et al. | Optimization of GPU parallel scheme for simulating ultrafast magnetization dynamics model | |
Faílde et al. | Using differential evolution to avoid local minima in variational quantum algorithms | |
JP6167188B2 (ja) | 計算機、及び演算プログラム | |
CN111527503B (zh) | 不均匀量子退火调度 | |
Yu et al. | Solving Quantum Many-Particle Models with Graph Attention Network | |
Mostame et al. | Decoherence in a dynamical quantum phase transition | |
JP6524340B2 (ja) | 計算機及び計算方法 | |
Asoudeh et al. | Perfect state transfer on spin-1 chains | |
Robson et al. | Advanced quantum Poisson solver in the NISQ era | |
Wu et al. | Finding quantum many-body ground states with artificial neural network | |
Al‐Rabadi | New dimensions in non‐classical neural computing, part II: quantum, nano, and optical | |
Skotiniotis et al. | Quantum resource theory for charge-parity-time inversion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15887486 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2017508862 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15561566 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 15887486 Country of ref document: EP Kind code of ref document: A1 |