JP7212891B2 - Quantum arithmetic device and method - Google Patents

Description

The present invention relates to quantum computing technology.

A quantum computer is a computer that consists of basic elements called quantum bits based on quantum theory. Since qubits have a finite error rate, if you increase the number of qubits in a computational task or lengthen the quantum computation time, eventually the results of quantum computation will become meaningless due to the effects of errors. turn into. Therefore, in order to perform useful computations on a quantum computer, it is necessary to design computational tasks appropriately so that errors do not render the computational results meaningless.

The quantum-classical hybrid algorithm is an idea to realize a useful application even with a quantum computer that has an error rate that cannot withstand long-term computation by using the quantum computer only for repeating small computational tasks. Within the framework of quantum-classical hybrid algorithms, quantum variational algorithms, which can be used to minimize functions that compute real numbers called losses from quantum states, have attracted particular attention. The mechanism of the quantum variational algorithm is explained below.

The task of quantum computation is a quantum circuit that describes a sequence of three types of operations called quantum operations, (i) initialization of the quantum state, (ii) quantum gate, and (iii) measurement of the quantum state, like a logic circuit. expressed in a way called In particular, a quantum gate that operates only on a single qubit is called a single-qubit rotation gate, and its operation can be characterized by a real parameter, the rotation angle. The computation time required for a quantum computing task corresponds to the length of the quantum circuit.

In the quantum variational algorithm, first, a quantum circuit with an allowable computation time is defined as a trial circuit. We regard the rotation angle of the single-qubit rotation gate in the trial circuit as a parameter set that characterizes the trial circuit. Quantum computation using the trial circuit results in a quantum state that depends on the parameter set. The rotation angle parameters of the trial circuit are optimized by variational methods so as to minimize the loss calculated from this quantum state. If the optimization is done sufficiently, we can expect to obtain a set of parameters for the input/output trial circuit that minimizes the loss and an optimized loss value.

Quantum variational algorithms such as those described above are expected to exhibit better performance than variational algorithms using classical computers, especially when input/output targets are naturally described using quantum mechanics. For example, chemicals are examples of objects that are naturally described using quantum mechanics. This is because the lowest energy state of complex chemical substances can only be expressed quantum-mechanically. Based on this, a specific application of the quantum variational algorithm is expected: the variational quantum eigenvalue method, in which loss is defined as the energy of a chemical substance of interest, and the lowest energy and corresponding quantum state of the substance are obtained. etc. (see, for example, Non-Patent Document 1).

Alberto Peruzzo, Jarrod McClean, Peter Shadbolt, Man-Hong Yung, Xiao-Qi Zhou, Peter J. Love, Alan Aspuru-Guzik, Jeremy L. O'Brien, "A variational eigenvalue solver on a photonic quantum processor", Nature communications 5, 4213 (2014)

In order to apply quantum variational algorithms to problems of practical scale, the error rate of individual qubits must be small enough to match the size of the problem. However, due to technical limitations in physical error rates, quantum variational algorithms cannot be applied to large-scale problems using naive methods.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a quantum arithmetic device and method capable of executing a quantum variational algorithm even if the error rate of the quantum arithmetic device is not sufficiently small.

In the quantum operation device according to one aspect of the present invention, k is a predetermined integer equal to or greater than 1, and a controllable k that characterizes the quantum operation that is a component of a trial quantum circuit that is a quantum circuit for executing a variational algorithm. An initialization unit for initializing parameters, a loss quantum circuit for calculating losses corresponding to trial quantum circuits corresponding to the initialized k parameters, and the initialized k parameters a quantum circuit generator for generating k update parameter quantum circuits for calculating k update parameters required for updating; and encoding each of the lossy quantum circuit and the k update parameter quantum circuits. an encoding unit, a quantum computation unit that executes each of the encoded lossy quantum circuit and the k updated parameter quantum circuits and obtains a measurement value of the execution result, and a decoding unit that decodes the measurement value of the execution result; , a loss update parameter calculator that calculates a loss and k update parameters using the decoded calculation results; and an update that updates the initialized k parameters based on the k update parameters. and a quantum circuit generation unit, an encoding unit, a quantum calculation unit, a decoding unit, a loss update parameter calculation unit, and an update unit using updated k parameters instead of the initialized k parameters. and a control unit that controls to repeat the process until a predetermined end condition is satisfied.

A quantum variational algorithm can be executed even if the error rate of the quantum arithmetic unit is not sufficiently small.

FIG. 1 is a diagram illustrating an example of the functional configuration of a quantum arithmetic device. FIG. 2 is a diagram showing an example of the processing procedure of the quantum arithmetic method. FIG. 3 is a diagram for explaining an example of processing by the updating unit 7. As shown in FIG. FIG. 4 is a diagram illustrating a functional configuration example of a computer.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail. In the drawings, constituent parts having the same function are denoted by the same numbers, and redundant explanations are omitted.

[Quantum arithmetic device and method]
In the present invention, a technique called quantum error correction is used in order to execute a quantum variational algorithm even if the error rate of a quantum arithmetic unit is not sufficiently small (see Reference 1, for example). In quantum error correction, the effective error rate can be reduced by redundantly representing a single quantum bit using multiple quantum bits. At this time, the redundantly expressed qubit is called a logical qubit, and each qubit constituting the logical qubit is called a physical qubit. Also, embedding a single quantum bit in a redundantly expressed space is called encoding, and the reverse operation is called decoding.

[Reference 1] Michael A. Nielsen, Isaac L. Chuang, “Quantum computation and quantum information”, pp.425-499 (2002)

Quantum operations that can lower the effective error rate by quantum error correction are called "fault-tolerant quantum operations."

As shown in FIG. 1, the quantum arithmetic device includes an initialization unit 1, a quantum circuit generation unit 2, an encoding unit 3, a quantum calculation unit 4, a decoding unit 5, a loss update parameter calculation unit 6, an update unit 7, and a control unit. 8 for example.

The quantum arithmetic method is realized, for example, by each component of the quantum arithmetic device performing the processing from step S1 to step S8 described below and shown in FIG.

Each component of the quantum arithmetic device will be described below.

It is assumed that the parameters characterizing the entire optimization process are the learning rate λ, the number of iterations L′, the allowable number of search steps P, and the allowable number of error-tolerant quantum operations m. It is also assumed that the search step number p is initialized to 0.

<Initialization part 1>
The initialization unit 1 initializes k controllable parameters that characterize quantum operations that are constituent elements of a predetermined trial quantum circuit (step S1). Information about the initialized k parameters is output to the quantum circuit generator 2 . k is a predetermined integer greater than or equal to 1;

For example, the initialization unit 1 initializes all rotation angle parameters in the trial quantum circuit to zero.

In the following, an example will be described in which the controllable parameter that characterizes the quantum operation that constitutes the trial quantum circuit is the rotation angle of a single-qubit rotation gate that constitutes the trial quantum circuit.

A controllable parameter that characterizes the quantum operations that constitute the trial quantum circuit is not limited to the rotation angle of the single-qubit rotation gate, and may be a parameter that does not correspond to the rotation angle, such as decoherence strength. Quantum operations that constitute the trial quantum circuit are not limited to one-quantum gate operations, and may include two-quantum gate operations, state initialization, and measurement.

<Quantum circuit generator 2>
Information about the k parameters initialized by the initialization unit 1 is input to the quantum circuit generation unit 2 . In the second and subsequent processes of the quantum circuit generation unit 2, information about k parameters updated by the update unit 7 is input to the quantum circuit generation unit 2. FIG.

Quantum circuit generation unit 2 generates a loss quantum circuit for calculating a loss for a trial quantum circuit characterized by information on input k parameters. The quantum circuit generation unit 2 also generates an update parameter quantum circuit for calculating an update parameter (for example, a gradient), which is a value necessary for updating the loss for each parameter.

That is, the quantum circuit generation unit 2 includes a loss quantum circuit for calculating losses corresponding to the trial quantum circuits corresponding to the k parameters initialized by the initialization unit 1, and the initialized k parameters k update parameter quantum circuits for calculating k update parameters required to update the parameters of (step S2).

A typical parameter update method is a method using a gradient for a parameter loss (see Reference 2, for example). Therefore, a case where the update parameter is a gradient will be described below as an example. Of course, the update parameter may be an update parameter other than the gradient (see Reference 3, for example).

[Reference 2] K. Mitarai, M. Negoro, M. Kitagawa, K. Fujii, "Quantum circuit learning", Phys. Rev. A 98, 032309 (2018)
[Reference 3] Ken M. Nakanishi, Keisuke Fujii, Synge Todo, "Sequential minimal optimization for quantum-classical hybrid algorithms", arXiv:1903.12166 (2019)

A loss quantum circuit is a trial quantum circuit appended with multiple measurement circuits to calculate the loss.

More specifically, the lossy quantum circuit is composed of, for example, M quantum circuits, where M is a predetermined number of 1 or more.

Each update parameter quantum circuit is also composed of a plurality of quantum circuits. The number of quantum circuits forming each update parameter quantum circuit depends on how the information required for updating is generated. For example, each update parameter quantum circuit is composed of 2M quantum circuits when the method of Reference 2 is used, and is composed of 3M quantum circuits when the method of Reference 3 is used.

A case where each of the k update parameter quantum circuits is composed of 2M quantum circuits will be described below as an example.

<Encoder 3>
The encoding unit 3 encodes each of the lossy quantum circuits and the k updated parameter quantum circuits (step S3).

At this time, the coding unit 3 selects the code length of the quantum code according to the configuration of the quantum circuit to be coded so as to sufficiently reduce the error rate of the entire quantum circuit finally realized.

In an example in which the lossy quantum circuit is composed of M quantum circuits and each of the k update parameter quantum circuits is composed of 2M quantum circuits, the encoding unit 3 performs the processing in step S3 to obtain a total of M Encode each of the +2kM quantum circuits.

Note that if the number of qubits of the trial quantum circuit is q, the number of qubits of each encoded quantum circuit is p. where q and p are positive integers and p>q.

Coding of the quantum circuit can be performed by the technique described in Reference 1, for example.

It should be noted that in the first-stage encoding in the encoding unit 3, there are some quantum operations that cannot be given error resilience. This fault-tolerant quantum operation can be implemented indirectly and approximately on logical qubits by so-called magic state injection and magic state distillation as described, for example, in ref. In other words, the encoding unit 3 performs the second-stage encoding different from the certain encoding on the code A obtained by the first-stage encoding to construct the code B. Quantum operations that could not be error-tolerant in the first-stage encoding can be implemented in code A in an error-tolerant manner. In this way, the encoding unit 3 uses the second-stage code to make error-tolerant quantum operations, which cannot be made error-tolerant in the first-stage encoding, as necessary. may be modified.

<Quantum computing unit 4>
The quantum calculator 4 executes each of the encoded lossy quantum circuits and the k updated parameter quantum circuits, and obtains measurement values of the execution results (step S4).

Note that for each quantum circuit, N executions are performed, and measurements of N execution results are obtained. N is a predetermined integer of 2 or more. Since the number of quantum circuits is k+1, we obtain measurements of (k+1)×N execution results.

The obtained measurement value of the execution result is output to the decoding unit 5 .

Note that when a quantum circuit with p quantum bits is executed, a value of p′ bits is obtained. where p' is an integer greater than p. This is because the same qubit may be measured several times. The value of this p' bit stochastically changes each time quantum computation is executed. To obtain statistics of this change, in an example where the loss quantum circuit consists of M quantum circuits and each of the k updated parameter quantum circuits consists of 2M quantum circuits, M+2kM quantum circuits Each of the circuits is executed N times. This yields an N×p′-bit value for each quantum circuit. In other words, all the values obtained are p′-bit strings obtained by running N times for each of the (M+2kM) types of quantum circuits. The total number of bits obtained by the quantum calculator 4 is (M+2kM)*N*p'.

Only the processing of the quantum computing unit 4 is performed by a quantum computer, and the processing of other units is performed by a classical computer. An example of a classical computer is computer 2000 illustrated in FIG. The computer 2000 has a control section 2010 , a recording section 2020 , an input section 2030 and an output section 2040 . The computer 2000 may include a display section 2050 as required.

A program to be executed is read into the recording unit 2020 of the computer, and the control unit 2010, the input unit 2030, the output unit 2040, etc. are operated to realize the processing by the classical computer.

<Decryption unit 5>
A measurement value of the execution result obtained by the quantum computing unit 4 is input to the decoding unit 5 .

The decoding unit 5 decodes the measurement value of the execution result (step S5). This makes it possible to obtain the calculation result in the quantum circuit before encoding.

Decoding is performed for each of the (k+1)×N run result measurements. Therefore, the decoding unit 5 obtains (k+1)×N decoded calculation results.

More specifically, in an example in which the lossy quantum circuit is composed of M quantum circuits and each of the k update parameter quantum circuits is composed of 2M quantum circuits, (M+2kM)*N Decoding is performed for each of the 'p' bit strings. As a result of this decoding, q bit strings are obtained. This q bit string is expressed as follows: "If a quantum computer with a small error exists virtually, the quantum circuit of q qubits input to the encoding unit 3 can be q bit strings obtained when running on a small quantum computer with Therefore, in this example, N q-bit strings are obtained for each of (M+2kM) kinds of quantum circuits by the processing in step S5.

The encoding of the measurement value of each execution result can be performed by the technique described in Reference 1, for example.

The calculation result decoded by the decoding unit 5 is output to the loss update parameter calculation unit 6 .

<Loss update parameter calculator 6>
The loss update parameter calculator 6 uses the decoded calculation results to calculate the loss and k update parameters (step S6). The calculated loss is output to the control section 8 . The calculated k update parameters are output to the updating unit 7 .

First, the loss update parameter calculator 6 performs statistical processing on the decoded calculation results, and obtains post-statistical calculation results for each quantum circuit.

In an example in which the lossy quantum circuit is composed of M quantum circuits and each of the k update parameter quantum circuits is composed of 2M quantum circuits, the decoding unit 5 processes (M+2kM) kinds of For each quantum circuit, N "q-bit strings" are obtained. The loss update parameter calculator 6 obtains one real number by performing statistical processing on the N*q bits assigned to each quantum circuit.

Since the lossy quantum circuit consists of M quantum circuits, we get M real numbers for the loss. By further processing the M real numbers, one real number called "loss" is obtained.

In this example, since each update parameter quantum circuit is composed of 2M quantum circuits, 2M real numbers are obtained for each update parameter. Further processing on the 2M reals yields one real for each update parameter. Since the number of update parameters is k, k update parameters are obtained.

For example, the calculation of the loss and the update parameter, eg the gradient, can be performed by the technique described in Non-Patent Document 1.

If the number of search steps is not p=0 and the loss has not been improved from the previously calculated loss, the loss update parameter calculator 6 sets p to p←p+1. That is, increment p by one.

<Update part 7>
The updating unit 7 receives input of the k update parameters calculated by the loss update parameter calculator 6 .

The update unit 7 updates the k parameters initialized by the initialization unit 1 based on the k update parameters (step S7).

The updated k parameters are input to the quantum circuit generator 2 and the controller 8 .

For example, if the parameter is the slope of the loss, the parameter before update is θ ₀ and the slope is δ, then the parameter after update can be θ ₀ −λδ. The update unit 7 can update the k parameters by performing this process for each of the k update parameters.

<Control unit 8>
The control unit 8 controls the quantum circuit generation unit 2, the encoding unit 3, and the quantum calculation using the k parameters updated by the update unit 7 instead of the k parameters initialized by the initialization unit 1. The processing of the unit 4, the decoding unit 5, the loss update parameter calculation unit 6, and the update unit 7 is controlled to be repeated until a predetermined end condition is satisfied (step S8).

The predetermined termination condition is, for example, search step number p=permissible search step number P. If the number of search steps p=the allowable number of search steps P, then a predetermined termination condition is satisfied.

When the number of search steps p=the allowable number of search steps P, the control unit 8 determines the smallest loss among the losses calculated so far by the loss update parameter calculation unit 6 and the loss with the smallest value. Output at least one of the k parameters corresponding to , and terminate the process. The k parameters corresponding to the loss are the k controllable parameters that characterize the quantum operations that constitute the trial quantum circuit corresponding to the loss quantum circuit used to calculate the loss.

For example, if the loss is defined as the energy of a chemical substance, the smallest loss obtained by this quantum variational algorithm is the lowest energy of the chemical substance. The quantum state of the chemical can be approximated by a trial quantum circuit characterized by the k parameters corresponding to the smallest losses obtained by this quantum variational algorithm.

In this example, the processing of the control unit 8 is performed immediately after the processing of the updating unit 7, but the processing of the control unit 8 may be performed immediately before the processing of the updating unit 7.

In this way, by using quantum error correction, a quantum variational algorithm can be executed even if the error rate of the quantum arithmetic device is not sufficiently small.

Quantum error correction can reduce the effective error rate to an arbitrary value with appropriate redundancy. Therefore, given a sufficient number of qubits with a low error rate, quantum variational algorithms can be applied to problems of arbitrary size.

[Variation]
When the rotation angle is updated variationally in a quantum circuit that has been coded by quantum error correction, as mentioned earlier, the first-stage coding generates quantum operations that cannot be made error-tolerant. can be lost. This error-tolerant quantum operation can be implemented error-tolerantly by second-stage encoding using magic state injection and magic state distillation techniques. It can take time.

Therefore, when updating the parameters during the execution of the quantum variational algorithm, the update unit 7 may perform the update while suppressing the number of quantum operations for which error resilience cannot be provided to an allowable number m or less. . This allows computation of the quantum variational algorithm to be performed in a shorter time.

In the following, we assume that the quantum operation that constitutes the trial quantum circuit is a rotation operation on a single qubit, the controllable parameter that characterizes the quantum operation is the rotation angle, and the update parameter is the gradient to the loss. An example will be described.

For example, when updating a parameter, the updating unit 7 updates the quantum operation in the trial quantum circuit corresponding to the parameter by the method described in FIG. It is decided to provisionally replace with an approximate quantum circuit that suppresses . The update unit 7 performs this process for each parameter.

First, symbols appearing in the following description and FIG. 3 will be described.

U is a single-qubit rotation gate. An expression with a rotation angle θ as an argument, such as U(θ), means a quantum gate that operates a quantum bit according to the rotation angle θ.

U ₀ is a description of the single-qubit rotation gate before the update, using a small number of "failure-tolerant quantum operations". In other words, when the processing by the updating unit 7 is the second time or later, U ₀ is U _out finally obtained by the previous processing by the updating unit 7 .

θ ₀ is the rotation angle of the gate that was attempted to be decomposed with the single-qubit rotation gate before updating.

δ is the slope, the real derivative of the single-qubit rotation gate with respect to the rotation angle loss.

λ is a learning rate, which is a real coefficient representing how much the rotation angle is updated according to the gradient δ.

U' is a single-qubit rotation gate after gradient-aware updates.

U _SK is a description of a quantum gate composed of at most m _SK “non-error-tolerant quantum operations” resulting from Gate decomposition.

m _SK is the number of "non-error-tolerant quantum operations" required to describe the U _SK obtained as a result of Gate decomposition.

L' is the number of iterations, and is an integer representing how many times the loop of the updating unit 7 is used in total.

L is a repetition counter, and is a variable that stores how many times the loop of the updating unit 7 has been used.

U _out is the variable that holds the best performing single-qubit rotation gate decomposition in the loop to date.

ε _min is the lower bound on the feasible gate operation error when searching for the best single-qubit rotation gate represented by m or fewer “failure-tolerant operations” using the bisection method. be.

ε _max is the upper bound on the feasible gate operation error when searching for the best single-qubit rotation gate represented by m or fewer “failure-tolerant operations” using the bisection method. be.

The updating unit 7 stores information about the quantum operation to be processed, the gradient δ, the learning rate λ, the number of iterations L′, the quantum circuit U ₀ searched before the update, the rotation angle θ ₀ before the update, the error resilience Information is input about the number m of quantum operations that cannot have

First, the updating unit 7 sets U'←U(θ ₀ -λδ). That is, the update unit 7 sets U′ as a rotation gate of the rotation angle θ ₀ −λδ obtained by updating the rotation angle θ ₀ before update using the gradient δ and the learning rate λ (step S71).

Further, the update unit 7 sets (ε _min ,ε _max )←(0,1),L←0 (step S71). That is, the updating unit 7 initializes ε _min to 0, ε _max to 1, and L to 0.

Also, the updating unit 7 sets U _out ←U ₀ . That is, the updating unit 7 sets U ₀ to U _out (step S71).

The update unit 7 sets (U _sk ,M _sk )←GateDecompose(U′,(ε _min +ε _max )/2) (step S72).

GateDecompose is a quantum gate construct that constructs a given single-qubit rotation gate from less than a given number of quantum operations that are not error-tolerant given the accuracy of the allowable error. It is a function that decomposes into series.

For example, the Solovay-Kitaev method (see, for example, reference 4) or a method improved from the Solovay-Kitaev method (see, for example, reference 5) can be used as GateDecompose.

[Reference 4] Christopher M. Dawson, Michael A. Nielsen, "The Solovay-Kitaev algorithm", arXiv:quant-ph/0505030 (2005)
[Reference 5] Neil J. Ross, Peter Selinger, "Optimal ancilla-free Clifford+T approximation of z-rotations", Quantum Information and Computation 16 (11-12):901-953, (2016)

That is, the updating unit 7 converts a given single-qubit rotation gate U′ to a number of errors less than a predetermined number under the accuracy of a given allowable error (ε _min +ε _max )/2. Decompose into a series U _SK of quantum gates composed of quantum operations that cannot be made tolerant. This decomposition also yields the number m _SK of "non-error-tolerant quantum operations" needed to describe U _SK .

The updating unit 7 determines whether m _SK >m (step S73).

If m _SK >m, the updating unit 7 sets ε _min ←(ε _min +ε _max )/2 (step S74). That is, in this case, the updating unit 7 calculates (ε _min +ε _max )/2 and sets the calculation result as ε _min .

If not m _SK >m, the updating unit 7 sets ε _max ←(ε _min +ε _max )/2 and U _out ←US _SK (step S75). That is, in this case, the updating unit 7 calculates (ε _min +ε _max )/2, sets the calculation result as ε _max , and sets U _SK as U _out .

The update unit 7 determines whether L=L' (step S76).

If not L=L', the update unit 7 sets L←L+1 (step S77). That is, the update unit 7 increments L by one. After that, the update unit 7 returns to step S72.

If L=L', the updating unit 7 outputs information about U _out corresponding to each parameter to the quantum circuit generating unit 2 and the control unit 8 (step S78). This U _out is an approximate quantum circuit corresponding to updated parameters (for example, rotation angle θ ₀ -λδ) composed of at most m error-tolerant quantum operations.

In this way, when updating the parameters, the updating unit 7 converts a predetermined quantum operation into a series of quantum gates composed of quantum operations that cannot be made less error-tolerant than a predetermined number. A non-error-tolerant quantum operation that approximates each quantum operation characterized by k parameters initialized or updated by binary search using a decomposing function. A reduced number of approximate quantum circuits may be searched.

Note that the quantum circuit generation unit 2 that has received the information about U _out performs quantum operations corresponding to each parameter in the trial quantum circuit characterized by the k parameters after updating to U _out corresponding to each parameter. Generate a loss quantum circuit and an update parameter quantum circuit for the trial quantum circuit replaced with .

Subsequent processing is the same as above. That is, the trial quantum circuit replaces the quantum operation corresponding to each parameter in the trial quantum circuit characterized by the updated k parameters with the quantum circuit searched by the updating unit 7 corresponding to each parameter. Assuming that it is a trial quantum circuit, the processes of the quantum circuit generator 2, the encoder 3, the quantum calculator 4, the decoder 5, and the loss update parameter calculator 6 are performed for the second and subsequent times.

This replaced trial quantum circuit and this generated lossy quantum circuit have m or less quantum operations for which error resilience cannot be provided. processing time is reduced.

In this way, when updating is performed while keeping the number of quantum operations for which error resilience cannot be provided to be less than the allowable number m, the control unit 8, when a predetermined termination condition is satisfied, outputs (i ) the loss with the smallest value among the losses calculated by the loss update parameter calculator 6 so far, and (ii) k parameters corresponding to the loss with the smallest value and U _out corresponding to each parameter Output at least one of them and terminate the process.

For example, if the loss is defined as the energy of a chemical substance, the smallest loss obtained by this quantum variational algorithm is the lowest energy of the chemical substance. Also, the quantum operation corresponding to each parameter in the trial quantum circuit characterized by k parameters corresponding to the smallest loss obtained by this quantum variational algorithm is replaced by U _out corresponding to each parameter. A trial quantum circuit can approximate the quantum state of a certain chemical substance.

Although the embodiments and modifications of the present invention have been described above, the specific configuration is not limited to these embodiments, and design changes may be made as appropriate without departing from the gist of the present invention. are also included in the present invention.

The various processes described in the embodiments are not only executed in chronological order according to the order of description, but may also be executed in parallel or individually according to the processing capacity of the device that executes the processes or as necessary. .

For example, data exchange between components of the quantum arithmetic device may be performed directly or may be performed via a storage unit (not shown).

1 Initialization unit 2 Quantum circuit generation unit 3 Encoding unit 4 Quantum calculation unit 5 Decoding unit 6 Loss update parameter calculation unit 7 Update unit 8 Control unit

Claims (3)

k is a predetermined integer of 1 or more, and an initialization unit that initializes k controllable parameters that characterize quantum operations that are constituent elements of a trial quantum circuit that is a quantum circuit for executing a variational algorithm ; ,
Loss quantum circuits for calculating losses corresponding to the trial quantum circuits corresponding to the initialized k parameters, and k parameters required to update the initialized k parameters respectively a quantum circuit generator for generating k update parameter quantum circuits for calculating update parameters;
an encoding unit that encodes each of the lossy quantum circuit and the k update parameter quantum circuits;
a quantum computing unit that executes each of the encoded lossy quantum circuit and the k updated parameter quantum circuits and obtains a measurement of execution results;
a decoding unit that decodes the measurement value of the execution result;
a loss update parameter calculation unit that calculates each of the loss and the k update parameters using the decoded calculation result;
an update unit that updates the initialized k parameters based on the k update parameters;
The quantum circuit generation unit, the encoding unit, the quantum calculation unit, the decoding unit, and the loss update parameter calculation unit using the updated k parameters instead of the initialized k parameters. and a control unit that controls to repeatedly perform the processing of the update unit until a predetermined end condition is satisfied;
Quantum computing device including The quantum arithmetic device of claim 1,
The update unit further uses a function that decomposes a predetermined quantum operation into a series of quantum gates composed of quantum operations that are less than a predetermined number and cannot be made error-tolerant, and performs a binary search. A quantum circuit that approximates each quantum operation characterized by the k initialized parameters or the k updated parameters by a method that reduces the number of quantum operations that cannot be made error-tolerant. explore and
wherein the trial quantum circuit replaces a quantum operation corresponding to each parameter in the trial quantum circuit characterized by the updated k parameters with the searched quantum circuit corresponding to each parameter. Assuming that it is a circuit, the processing of the quantum circuit generation unit, the encoding unit, the quantum calculation unit, the decoding unit, and the loss update parameter calculation unit after the second time is performed.
Quantum computing device. An initialization unit initializes k controllable parameters that characterize quantum operations, where k is a predetermined integer equal to or greater than 1, and that are constituents of a trial quantum circuit, which is a quantum circuit for executing a variational algorithm . an initialization step that
A quantum circuit generation unit for updating a loss quantum circuit for calculating a loss corresponding to the trial quantum circuit corresponding to the initialized k parameters and the initialized k parameters, respectively. a quantum circuit generation step for generating k update parameter quantum circuits for calculating k update parameters required for
an encoding step in which an encoding unit encodes each of the lossy quantum circuit and the k update parameter quantum circuits;
a quantum computing step in which a quantum computing unit performs respective executions of the encoded lossy quantum circuit and the k updated parameter quantum circuits and obtains a measurement of the execution results;
a decoding step in which the decoding unit decodes the measured value of the execution result;
a loss update parameter calculation step in which a loss update parameter calculation unit calculates each of the loss and the k update parameters using the decoded calculation result;
an update step in which the update unit updates the initialized k parameters based on the k update parameters;
The quantum circuit generation unit, the encoding unit, the quantum calculation unit, the decoding unit, and the loss, wherein the control unit uses the updated k parameters instead of the initialized k parameters. a control step of controlling to repeatedly perform the processing of the update parameter calculation unit and the update unit until a predetermined termination condition is satisfied;
Quantum arithmetic methods, including

Images