Method and apparatus for generating a quantum circuit autoencoder

Abstract

The application provides a quantum circuit autoencoder generation method and device. The method comprises the following steps: generating a quantum circuit encoder and a first quantum channel according to an input quantum channel; generating a second quantum channel according to the initial quantum circuit encoder and the first quantum channel; generating a target quantum state according to the second quantum channel and an initial quantum state; iteratively adjusting a parameter value of the initial quantum circuit encoder to determine a single-bit noise and a target parameter value of a parameter-containing quantum circuit according to a target loss value of the target quantum state corresponding to the initial quantum circuit encoder; generating a third quantum channel according to the target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed according to the single-bit noise and the target parameter value of the parameter-containing quantum circuit; generating a latent channel and a noise channel according to the third quantum channel; and generating a reconstructed quantum channel according to the latent channel, the noise channel and a target quantum circuit decoder.

Description

Technical Field

[0001] The present application relates to the field of quantum computing technology, and more specifically, to a method for generating a quantum circuit autoencoder, a generating device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] As a popular neural network tool, the autoencoder's network structure maps the input raw data from high-dimensional to low-dimensional data, and then maps the low-dimensional data back to the reconstructed data of the original dimension. The neural network parameters are then trained with the goal of reducing the reconstruction error of the reconstructed data relative to the original data.

[0003] In the process of implementing the concept of this application, it was found that the fidelity of the quantum channel reconstructed by the quantum circuit autoencoder in the related art was poor. Summary of the Invention

[0004] In view of this, the present application provides a method for generating a quantum circuit autoencoder, a generating device, an electronic device, a computer-readable storage medium, and a computer program product.

[0005] One aspect of the present application provides a method for generating a quantum circuit autoencoder, comprising:

[0006] In response to the compression instruction, an initial quantum circuit encoder and a first quantum channel are generated according to the input quantum channel, wherein each layer of the initial quantum circuit encoder is generated according to a plurality of single-bit noises and parameter-containing quantum circuits;

[0007] generating a second quantum channel according to the initial quantum circuit encoder and the first quantum channel;

[0008] generating a target quantum state according to the second quantum channel and the initial quantum state, wherein the initial quantum state is obtained according to the maximum mixed state and the maximum entangled state;

[0009] Iteratively adjusting parameter values ​​of the initial quantum circuit encoder to determine target parameter values ​​of the single-bit noise and the parameter-containing quantum circuit according to a target loss value of a target quantum state corresponding to the initial quantum circuit encoder;

[0010] generating a third quantum channel according to a target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed according to the single-bit noise and the target parameter value of the parameter-containing quantum circuit;

[0011] According to the third quantum channel, a potential channel and a noise channel are generated;

[0012] In response to a reconstruction instruction, the reconstructed quantum channel is generated according to the potential channel, the noise channel, and a target quantum circuit decoder, wherein the target quantum circuit decoder is the conjugate transpose of the target quantum circuit encoder, and the quantum circuit autoencoder includes the target quantum circuit encoder and the target quantum circuit decoder.

[0013] According to an embodiment of the present application, generating a potential channel and a noise channel according to the third quantum channel includes:

[0014] performing a detour operation on the third quantum channel while applying the third quantum channel to a target quantum space to obtain the potential channel, wherein the original quantum space includes a potential space and a garbage space, and the target quantum space includes the original quantum space and an isomorphic space corresponding to the garbage space;

[0015] Extracting the system state of the quantum system corresponding to the garbage space in the target quantum space, and measuring the system state in a standard basis to obtain state information;

[0016] The noise channel is generated according to the state information.

[0017] According to an embodiment of the present application, the target quantum circuit decoder includes a first decoder and a second decoder;

[0018] The reconstructed quantum channel is generated according to the potential channel, the noise channel and the target quantum circuit decoder, including:

[0019] generating a target channel according to the potential channel and the noise channel;

[0020] The reconstructed quantum channel is generated according to the target channel, the first decoder and the second decoder.

[0021] According to an embodiment of the present application, generating a second quantum channel according to an initial quantum circuit encoder and a first quantum channel includes:

[0022] Acquire a quantum circuit set, wherein the quantum circuit set includes a plurality of first quantum circuits;

[0023] Processing the plurality of the first quantum circuits based on a preprocessing formula to obtain the first quantum channel;

[0024] The second quantum channel is generated according to the initial quantum circuit encoder and the first quantum channel.

[0025] According to an embodiment of the present application, the original quantum space includes potential space and garbage space;

[0026] According to an embodiment of the present application, generating a target quantum state according to the second quantum channel and the initial quantum state includes:

[0027] The maximum mixing state acting on the latent space;

[0028] The maximum entangled state acting on junk space and its reconstruction space;

[0029] generating the initial quantum state according to the maximum mixed state and the maximum entangled state;

[0030] The second quantum channel is applied to the initial quantum state to obtain the target quantum state.

[0031] According to an embodiment of the present application, iteratively adjusting the parameter values ​​of the initial quantum circuit encoder to determine the single-bit noise and the target parameter values ​​of the parameter-containing quantum circuit according to the target loss value of the target quantum state corresponding to the initial quantum circuit encoder includes:

[0032] Determining initial parameter values ​​of the single-bit noise and the parameter-containing quantum circuit, so as to input the target quantum state corresponding to the initial parameter values ​​into the target loss function to obtain an initial loss value;

[0033] The initial parameter value is iteratively adjusted according to the initial loss value to obtain an adjusted parameter value, so that the adjusted parameter value is determined as the target parameter value when the iteration condition is met.

[0034] According to an embodiment of the present application, iteratively adjusting the initial parameter value according to the initial loss value to obtain an adjusted parameter value, and determining the adjusted parameter value as the target parameter value when an iteration condition is satisfied, includes:

[0035] During the i-th iteration, the i-th gradient information of the i-th initial loss value corresponding to the i-th iteration is calculated using the finite difference algorithm;

[0036] Generate an i+1th initial parameter value for an i+1th iteration based on the i-th initial parameter value and the i-th gradient information, and generate an i+1th gradient information based on the i+1th initial parameter value for the i+1th iteration;

[0037] When the difference between the i-th gradient information and the (i+1)-th gradient information is smaller than a preset gradient threshold, the i-th initial parameter value corresponding to the i-th gradient information is determined as the target parameter value.

[0038] According to an embodiment of the present application, the initial parameter value is iteratively adjusted according to the initial loss value to obtain an adjusted parameter value, and the adjusted parameter value is determined as the target parameter value when an iteration condition is met, including:

[0039] In the i-th iteration process, the i-th gradient information of the i-th initial loss value corresponding to the i-th iteration is calculated by using a finite difference algorithm;

[0040] According to the i-th initial parameter value in the i-th iteration process and the i-th gradient information, an i+1-th initial parameter value of an i+1-th iteration is generated;

[0041] When the iteration number i meets a preset iteration threshold, the i+1-th initial parameter value is determined as the target parameter value.

[0042] According to an embodiment of the present application, the target quantum circuit encoder includes a first encoder and a second encoder;

[0043] According to the target quantum circuit encoder and the first quantum channel, a third quantum channel is generated, including:

[0044] According to the target parameter value and the initial quantum circuit encoder, the first encoder and the second encoder are respectively generated;

[0045] According to the first encoder, the second encoder and the first quantum channel, the third quantum channel is generated.

[0046] Another aspect of the present application provides a quantum circuit autoencoder generation device, including:

[0047] A first generation module is configured to generate an initial quantum circuit encoder and a first quantum channel according to an input quantum channel in response to a compression instruction, wherein each layer in the initial quantum circuit encoder is generated according to a plurality of single-bit noises and a parameter-containing quantum circuit;

[0048] A second generation module is configured to generate a second quantum channel according to the initial quantum circuit encoder and the first quantum channel in response to a reconstruction instruction;

[0049] A third generation module is configured to generate a target quantum state according to the second quantum channel and an initial quantum state, wherein the initial quantum state is obtained according to a maximum mixed state and a maximum entangled state;

[0050] A determination module is configured to iteratively adjust a parameter value of the initial quantum circuit encoder to determine a target parameter value of the single-bit noise and the parameter-containing quantum circuit according to a target loss value of a target quantum state corresponding to the initial quantum circuit encoder.

[0051] a fourth generating module, configured to generate a third quantum channel according to a target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed according to the single-bit noise and a target parameter value of the quantum circuit with parameters;

[0052] a fifth generating module, configured to generate a latent channel and a noise channel according to the third quantum channel;

[0053] a sixth generating module, configured to generate the reconstructed quantum channel according to the latent channel, the noise channel and a target quantum circuit decoder, wherein the target quantum circuit decoder is a conjugate transpose of the target quantum circuit encoder, and the quantum circuit self-encoder comprises the target quantum circuit encoder and the target quantum circuit decoder.

[0054] Another aspect of the present application provides an electronic device, comprising:

[0055] one or more processors;

[0056] a memory configured to store one or more programs,

[0057] wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method as described above.

[0058] Another aspect of the present application provides a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the method as described above.

[0059] Another aspect of the present application provides a computer program product comprising computer-executable instructions, which, when executed, implement the method as described above.

[0060] According to an embodiment of the present application, a second quantum channel is generated based on an initial quantum circuit encoder and a first quantum channel. A target quantum state is generated by combining the initial quantum state. The single-bit noise and target parameter values ​​of the parameter-containing quantum circuit used for the target quantum circuit encoder are determined based on the target loss value of the target quantum state corresponding to the initial quantum circuit encoder. A third quantum channel is generated based on the first quantum channel. A potential channel and a noise channel are generated based on the third quantum channel. Thus, a reconstructed quantum channel is generated based on the potential channel, the noise channel, and the target quantum circuit decoder. By constructing a parameter-containing quantum channel with the aid of controllable single-bit noise and the target parameter values ​​of the parameter-containing quantum circuit, and using it as the basic structure of the encoder and decoder, it is more suitable for processing information compression tasks in a quantum channel composed of a group of quantum circuits. At the same time, during the reconstruction phase, noise is used to assist in reconstructing the original quantum channel, allowing the information contained in the quantum circuit to complete the compression-recovery process with higher reconstruction fidelity. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The above and other objects, features and advantages of the present application will become more apparent through the following description of the embodiments of the present application with reference to the accompanying drawings, in which:

[0062] Figure 1 A flowchart of a method for generating a quantum circuit autoencoder according to an embodiment of the present application is shown;

[0063] Figure 2 A schematic structural diagram of an initial quantum circuit encoder according to an embodiment of the present application is shown;

[0064] Figure 3 A schematic diagram of training target parameter values ​​according to an embodiment of the present application is shown;

[0065] Figure 4 A schematic diagram of a quantum circuit autoencoder model according to an embodiment of the present application is shown;

[0066] Figure 5 A block diagram of a generating device of a quantum circuit autoencoder according to an embodiment of the present application is shown;

[0067] Figure 6 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0068] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present application. In the detailed description below, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present application. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present application.

[0069] The terms used herein are only for describing specific embodiments and are not intended to limit this application. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0070] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0071] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0072] Quantum autoencoders process the information contained in quantum states, mapping the input high-dimensional quantum state to a low-dimensional quantum state via a parameterized quantum circuit, and then mapping the low-dimensional quantum state to a reconstructed quantum state of the original dimension. These two mapping functions are controlled by two parameterized quantum circuit structures controlled by a set of parameters. The control parameters of the parameterized quantum circuits are also trained to reduce the reconstruction error of the reconstructed data relative to the original data.

[0073] However, in practice, many tasks store information in quantum circuits. For example, many quantum machine learning tasks use quantum circuits to map classical data to quantum data. In this case, the information is contained in both the quantum circuit and the quantum state obtained after the quantum circuit is executed. However, the autoencoder scheme for quantum states does not seem to be directly applicable to quantum circuit information processing.

[0074] Related technologies for quantum gate compression have all proposed prototypes that process information in quantum circuits in a manner similar to an autoencoder. However, these methods only consider the compression and recovery of quantum gates composed of multiple single-qubit gates of the same form, arranged in an independent and identically distributed manner. This limits the fidelity of their reconstruction.

[0075] In view of this, embodiments of the present application provide a method and apparatus for generating a quantum circuit autoencoder. The method includes generating a quantum circuit encoder and a first quantum channel based on an input quantum channel in response to a compression instruction; generating a second quantum channel based on the initial quantum circuit encoder and the first quantum channel in response to a reconstruction instruction, wherein each layer in the initial quantum circuit encoder is generated based on multiple single-bit noises and parameter-containing quantum circuits; generating a target quantum state based on the second quantum channel and the initial quantum state; iteratively adjusting a parameter value of the single-bit noise to determine target parameter values ​​of the single-bit noise and the parameter-containing quantum circuit based on a target loss value of the target quantum state corresponding to the single-bit noise; generating a third quantum channel based on the target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed based on the single-bit noise and the target parameter values ​​of the parameter-containing quantum circuit; generating a potential channel and a noise channel based on the third quantum channel; and generating a reconstructed quantum channel based on the potential channel, the noise channel, and the target quantum circuit decoder.

[0076] Figure 1 A flowchart of a method for generating a quantum circuit autoencoder according to an embodiment of the present application is shown. Figure 2 A schematic structural diagram of an initial quantum circuit encoder according to an embodiment of the present application is shown.

[0077] like Figure 1 As shown, the method for generating a quantum circuit autoencoder includes operations S101 to S107.

[0078] In operation S101, in response to a compression instruction, an initial quantum circuit encoder and a first quantum channel are generated according to an input quantum channel, wherein each layer in the initial quantum circuit encoder is generated according to a plurality of single-bit noises and parameter-containing quantum circuits;

[0079] In operation S102, a second quantum channel is generated according to the initial quantum circuit encoder and the first quantum channel;

[0080] In operation S103, a target quantum state is generated according to the second quantum channel and the initial quantum state, wherein the initial quantum state is obtained according to the maximum mixed state and the maximum entangled state;

[0081] In operation S104, iteratively adjusting parameter values ​​of the encoder to determine target parameter values ​​of the single-bit noise and the parameter-containing quantum circuit according to a target loss value of the target quantum state;

[0082] In operation S105, a third quantum channel is generated according to the target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed according to single-bit noise and target parameter values ​​of the parameter-containing quantum circuit;

[0083] In operation S106, a potential channel and a noise channel are generated according to the third quantum channel;

[0084] In operation S107, in response to the reconstruction instruction, a reconstructed quantum channel is generated according to the potential channel, the noise channel, and the target quantum circuit decoder, wherein the target quantum circuit decoder is the conjugate transpose of the target quantum circuit encoder, and the quantum circuit autoencoder includes the target quantum circuit encoder and the target quantum circuit decoder.

[0085] According to an embodiment of the present application, when the method for generating a quantum circuit autoencoder is executed, a generation instruction required to execute the quantum circuit autoencoder is first obtained according to a generation request of the quantum circuit autoencoder, and the generation instruction includes a compression instruction and a reconstruction instruction.

[0086] According to an embodiment of the present application, the initial quantum state is the state of the original quantum space, specifically generated by the maximum mixed state and the maximum entangled state.

[0087] According to an embodiment of the present application, the initial quantum circuit encoder is a parameter-containing quantum channel assisted by controlled noise, and its specific structure is as follows: Figure 2 As shown, it consists of two parts. The first part is a number of single-bit noises controlled by parameters, namely Figure 2 in The second part of the column is the traditional parameter-containing quantum circuit, that is, The structure on the right. Figure 2 What is shown is a layer of parameterized quantum channel. In specific tasks, multiple layers of this structure can be used according to the problem, i.e. Figure 2 Only one layer is shown, which can be Figure 2 Add multiple layers of the same two-part structure to the right of , where q0, q1, q2, and q3 are all quantum bits, and RX is a controlled Pauli X rotation gate.

[0088] Specifically, the structures of the initial quantum circuit encoder and the target quantum circuit decoder can be expressed by Formula 1:

[0089] (1)

[0090] in, For the control parameters, is the parameter control noise of the jth layer, that is, single-bit noise, is the parameter-containing quantum circuit of the jth layer. Parameter-controlled noise includes common noises such as depolarization noise, bit flip noise, phase damping noise, etc. The specific form of the parameter-containing quantum circuit is shown in formula (2):

[0091] (2)

[0092] in, and are Hermitian operators and unitary operators respectively.

[0093] According to an embodiment of the present application, the first quantum channel acts on the same Hilbert space A quantum channel is obtained by data preprocessing of multiple quantum circuits on . A second quantum channel is generated based on the above initial quantum circuit encoder and the first quantum channel. According to the second quantum channel and the initial quantum state shown in formula (3) , generating the target quantum state .

[0094] (3)

[0095] By partially measuring the target quantum state, the target loss value corresponding to this set of parameters can be obtained. The target loss value is calculated as follows:

[0096]

[0097] According to an embodiment of the present application, the parameter value of the single-bit noise is iteratively adjusted , to determine the target parameter value of the single-bit noise according to the target loss value of the target quantum state corresponding to the single-bit noise According to the target parameter values ​​of single-bit noise and parametric subcircuits The target quantum circuit encoder and the first quantum channel are constructed to generate the third quantum channel. Based on the third quantum channel and the partial measurement results of the target quantum state, the potential channel is generated. and noisy channels According to the potential channel, noise channel and target quantum circuit decoder, the reconstructed quantum channel is generated. .

[0098] According to an embodiment of the present application, a second quantum channel is generated based on an initial quantum circuit encoder and a first quantum channel. A target quantum state is generated by combining the initial quantum state. The single-bit noise and target parameter values ​​of the parametric quantum circuit used for the target quantum circuit encoder are determined based on the target loss value of the target quantum state corresponding to the target quantum circuit encoder. A third quantum channel is generated based on the first quantum channel. A potential channel and a noise channel are generated based on the third quantum channel. Thus, a reconstructed quantum channel is generated based on the potential channel, the noise channel, and the target quantum circuit decoder. By constructing a parametric quantum channel with the aid of controllable single-bit noise and the target parameter values ​​of the parametric quantum circuit, and using it as the basic structure of the encoder and decoder, the parametric quantum channel is more suitable for processing information compression tasks in a quantum channel composed of a group of quantum circuits. At the same time, during the reconstruction phase, noise is used to assist in reconstructing the original quantum channel, allowing the information contained in the quantum circuit to complete the compression-recovery process with higher reconstruction fidelity.

[0099] According to an embodiment of the present application, generating a potential channel and a noise channel according to the third quantum channel includes:

[0100] performing a detrace operation on the third quantum channel to obtain a potential channel, wherein the original quantum space includes the potential space and the garbage space, and the target quantum space includes the original quantum space and an isomorphic space corresponding to the garbage space;

[0101] Extract the system state of the quantum system corresponding to the garbage space in the target quantum space, and measure the subspace of the system state under the standard basis to obtain the state information;

[0102] A noise channel is generated based on the state information.

[0103] According to the embodiment of the present application, the channel acts on the spatial target quantum space On, among them, is a potential channel, For garbage channels, for The isomorphic space of , performs the deviation operation on the channel and obtains the potential channel, as shown in formula (4):

[0104] (4)

[0105] According to an embodiment of the present application, the system state of the quantum system corresponding to the garbage space and its reconstruction space in the target quantum space is extracted, and the state of this system is recorded as , perform measurement operations on it under the standard basis and obtain status information , Among them, the state information is shown in formula (5):

[0106] (5)

[0107] According to an embodiment of the present application, a noise channel as shown in formula (6) is constructed based on the state information .

[0108] (6)

[0109] According to an embodiment of the present application, the target quantum circuit decoder includes a first decoder and the second decoder .

[0110] The reconstructed quantum channel is generated according to the potential channel, the noise channel and the target quantum circuit decoder, including:

[0111] Generate a target channel based on the potential channel and the noise channel;

[0112] A reconstructed quantum channel is generated according to the target channel, the first decoder and the second decoder.

[0113] According to an embodiment of the present application, the potential channel and noisy channels Synthesize into a new channel, namely the target channel, and construct the reconstructed quantum channel shown in formula (7) together with the first decoder and the second decoder :

[0114] (7)

[0115] According to an embodiment of the present application, generating a second quantum channel according to an initial quantum circuit encoder and a first quantum channel includes:

[0116] Acquire a quantum circuit set, wherein the quantum circuit set includes a plurality of first quantum circuits acting on the same Hilbert space;

[0117] Processing the plurality of first quantum circuits based on a preprocessing formula to obtain a first quantum channel;

[0118] A second quantum channel is generated according to the initial quantum circuit encoder and the first quantum channel.

[0119] According to an embodiment of the present application, a quantum circuit set is obtained , the quantum circuit set consists of multiple actions in the same Hilbert space The first quantum circuit on . Process multiple first quantum circuits to obtain the first quantum channel , satisfying formula (8):

[0120] (8)

[0121] in, For any quantum state, is the conjugate transpose.

[0122] According to an embodiment of the present application, according to the initial quantum circuit encoder 、 and the first quantum channel , generating the second quantum channel as shown in formula (3).

[0123] According to an embodiment of the present application, the original quantum space includes the potential space and Junkspace and the isomorphic space of junkspace .

[0124] According to an embodiment of the present application, generating a target quantum state according to the second quantum channel and the initial quantum state includes:

[0125] The maximum mixing state acting on the latent space;

[0126] The maximum entangled state acting on junk space and its reconstruction space;

[0127] Generate an initial quantum state according to the maximum mixed state and the maximum entangled state;

[0128] The second quantum channel is applied to the initial quantum state to obtain the target quantum state.

[0129] According to an embodiment of the present application, the original channel (i.e., the first quantum channel ) After compression, the space where the channel is located is obtained, which is the mixed state above In the space where the entangled state is a maximum entangled state acting on the garbage space (i.e. the space that is redundant with the potential space) and its isomorphic space. In which, assuming that the original space is , then the mixed state and entangled states As shown in formula (9) and formula (10) respectively:

[0130] (9)

[0131] (10)

[0132] in, and The latent space and Junkspace The dimensions of , i and j are both standard bases.

[0133] According to the embodiment of the present application, according to the mixed state and entangled states , generating the initial quantum state ; The second quantum channel and the initial quantum state shown in formula (3) Composite into target quantum state .

[0134] Figure 3 A schematic diagram of training target parameter values ​​according to an embodiment of the present application is shown.

[0135] According to an embodiment of the present application, iteratively adjusting the parameter value of the initial quantum circuit encoder to determine the single-bit noise and the target parameter value of the parameter-containing quantum circuit according to the target loss value of the target quantum state corresponding to the initial quantum circuit encoder includes:

[0136] Determining initial parameter values ​​of the single-bit noise and the parameter-containing quantum circuit (generally by random initialization) so as to input the target quantum state corresponding to the initial parameter values ​​into the target loss function to obtain an initial loss value;

[0137] The initial parameter value is iteratively adjusted according to the initial loss value to obtain an adjusted parameter value, so as to determine the adjusted parameter value as the target parameter value when an iteration condition is met.

[0138] According to an embodiment of the present application, iteratively adjusting an initial parameter value according to an initial loss value to obtain an adjusted parameter value, and determining the adjusted parameter value as a target parameter value when an iteration condition is satisfied, includes:

[0139] During the i-th iteration, the i-th gradient information of the i-th initial loss value corresponding to the i-th iteration is calculated using the finite difference algorithm;

[0140] Generate an i+1th initial parameter value for an i+1th iteration based on the i-th initial parameter value and the i-th gradient information in the i-th iteration, and generate an i+1th gradient information based on the i+1th initial parameter value for the i+1th iteration;

[0141] When the difference between the i-th gradient information and the (i+1)-th gradient information is less than a preset gradient threshold, the i-th initial parameter value corresponding to the i-th gradient information is determined as the target parameter value.

[0142] According to an embodiment of the present application, iteratively adjusting an initial parameter value according to an initial loss value to obtain an adjusted parameter value, and determining the adjusted parameter value as a target parameter value when an iteration condition is satisfied, includes:

[0143] During the i-th iteration, the i-th gradient information of the i-th initial loss value corresponding to the i-th iteration is calculated using the finite difference algorithm;

[0144] Generate the i+1th initial parameter value of the i+1th iteration according to the i-th initial parameter value and the i-th gradient information in the i-th iteration;

[0145] When the number of iterations i meets the preset iteration threshold, the i+1th initial parameter value is determined as the target parameter value.

[0146] According to an embodiment of the present application, a classical optimizer is used to train the target quantum state:

[0147] Reference Figure 3 , first randomly initialize the initial parameter values ;

[0148] Then repeat the following steps wheel:

[0149] (1) The initial parameter value of each round is , for the initial parameter value The corresponding target quantum state is input into the target loss function shown in formula (11) to obtain the initial loss value;

[0150] (11)

[0151] The observation operator is , and They are and The corresponding number of systematic bits. is the maximum entangled state of the two bits, generated according to formula (12):

[0152] (12)

[0153] Among them, I, Z, and X are all Pauli operators.

[0154] (2) Use the finite difference method to estimate the gradient information of the target loss function at this point, as shown in formula (12):

[0155] (12)

[0156] in, and and The difference is that parameters, specifically expressed as , .

[0157] Update the initial parameter value according to the gradient information to obtain the initial parameter value for the next iteration , as shown in formula (13)

[0158] (13)

[0159] in It is the learning rate set according to actual needs, for example 0.01, is the gradient information of the current point.

[0160] Determine whether the current gradient information has converged or whether the maximum number of training rounds has been reached. If either condition is met, terminate the loop; otherwise, repeat the above steps. This will output approximately optimal parameters.

[0161] In one embodiment, when the difference between the i-th gradient information and the (i+1)-th gradient information is less than a preset gradient threshold, the i-th initial parameter value corresponding to the i-th gradient information is determined as the target parameter value.

[0162] According to the embodiment of the present application, the preset gradient threshold can be set according to actual needs, for example, 10 -5 , when the difference is less than 10 -5 , it can be considered converged, so the i-th initial parameter value corresponding to the i-th gradient information is determined as the target parameter value

[0163] In another embodiment, when the number of iterations i meets a preset iteration threshold, the i+1th initial parameter value is determined as the target parameter value. It can be set according to actual needs, for example, T can be 50, 100, etc.

[0164] It should be noted that Figure 3 middle For the "potential" system, For "junk" systems, is an isomorphic system of the "junk" system, is the input system of the first quantum channel, is the input system of the first quantum channel, For the "potential" system, For "junk" systems, Output system for the "junk" system.

[0165] Figure 4 A schematic diagram of a model of a quantum circuit autoencoder according to an embodiment of the present application is shown.

[0166] According to an embodiment of the present application, the target quantum circuit encoder includes a first encoder and the second encoder

[0167] The step of generating a third quantum channel according to the target quantum circuit encoder and the first quantum channel includes:

[0168] Generate a first encoder and a second encoder respectively according to the target parameter value and the initial quantum circuit encoder;

[0169] A second quantum channel is generated according to the first encoder, the second encoder and the first quantum channel.

[0170] According to the embodiment of the present application, according to the target parameter value and the initial quantum circuit encoder, the following are generated respectively: Figure 4 The first encoder shown and the second encoder .

[0171] According to an embodiment of the present application, according to the first encoder and the second encoder and the first quantum channel , generate the second quantum channel, as shown in formula (14):

[0172] (14).

[0173] According to an embodiment of the present application, the quantum channel is reconstructed by calculation With the original channel The fidelity can verify the effect of training. The calculation method of fidelity is shown in formula (15):

[0174]

[0175] (15)

[0176] in, Indicates the original channel The corresponding Choi state, and is a set of homogeneous quantum systems.

[0177] According to an embodiment of the present application, if the fidelity does not meet the preset fidelity requirement, the above-mentioned iterative update can be performed again, wherein the preset fidelity can be specifically set according to actual needs.

[0178] Figure 5 A block diagram of a generating device of a quantum circuit autoencoder according to an embodiment of the present application is shown.

[0179] like Figure 5As shown, the generation device 600 of the quantum circuit autoencoder includes a first generation module 510, a second generation module 520, a third generation module 530, a determination module 540, a fourth generation module 550, a fifth generation module 560, and a sixth generation module 570.

[0180] A first generating module 510 is configured to generate an initial quantum circuit encoder and a first quantum channel according to the input quantum channel in response to the compression instruction, wherein each layer of the initial quantum circuit encoder is generated according to multiple single-bit noises and parameter-containing quantum circuits;

[0181] A second generating module 520 is configured to generate a second quantum channel according to the initial quantum circuit encoder and the first quantum channel in response to the reconstruction instruction;

[0182] A third generating module 530 is configured to generate a target quantum state based on the second quantum channel and an initial quantum state, wherein the initial quantum state is obtained by the quantum bit acting on the original quantum space;

[0183] a determination module 540, configured to iteratively adjust parameter values ​​of the initial quantum circuit encoder to determine target parameter values ​​of the single-bit noise and the parameter-containing quantum circuit according to a target loss value of a target quantum state corresponding to the initial quantum circuit encoder;

[0184] a fourth generating module 550, configured to generate a third quantum channel based on a target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed based on the single-bit noise and a target parameter value of the parameter-containing quantum circuit;

[0185] a fifth generating module 560, configured to generate a potential channel and a noise channel based on the third quantum channel;

[0186] A sixth generation module 570 is configured to generate the reconstructed quantum channel based on the potential channel, the noise channel, and a target quantum circuit decoder, wherein the target quantum circuit decoder is the conjugate transpose of the target quantum circuit encoder, and the quantum circuit autoencoder includes the target quantum circuit encoder and the target quantum circuit decoder.

[0187] According to an embodiment of the present application, a second quantum channel is generated based on an initial quantum circuit encoder and a first quantum channel. A target quantum state is generated by combining the initial quantum state. The single-bit noise and target parameter values ​​of the parameterized quantum circuit used for the target quantum circuit encoder are determined based on the target loss value measured on the target quantum state. A third quantum channel is generated based on the first quantum channel. A potential channel and a noise channel are generated based on the third quantum channel. Thus, a reconstructed quantum channel is generated based on the potential channel, the noise channel, and the target quantum circuit decoder. By constructing a parameterized quantum channel with the aid of the target parameter value of the single-bit noise with controllable parameters, and using it as the basic structure of the encoder and decoder, this method is more suitable for processing information compression tasks in a quantum channel composed of a group of quantum circuits. Furthermore, during the reconstruction phase, noise is used to assist in reconstructing the original quantum channel, allowing the information contained in the quantum circuit to complete the compression-recovery process with higher reconstruction fidelity.

[0188] According to the embodiments of the present application, any multiple modules, or at least part of the functions of any multiple modules, can be implemented in one module. According to the embodiments of the present application, any one or more modules can be split into multiple modules for implementation. According to the embodiments of the present application, any one or more modules can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), system-on-chips, systems on substrates, systems on packages, application-specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging the circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in an appropriate combination of any of them. Alternatively, according to the embodiments of the present application, one or more modules can be at least partially implemented as computer program modules, which can perform corresponding functions when the computer program modules are run.

[0189] For example, any multiple of the first generation module 510, the second generation module 520, the third generation module 530, the determination module 540, the fourth generation module 550, the fifth generation module 560, and the sixth generation module 570 can be combined into one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functionality of one or more of these modules / units / sub-units can be combined with at least part of the functionality of other modules / units / sub-units and implemented in one module / unit / sub-unit.

[0190] According to an embodiment of the present application, at least one of the first generation module 510, the second generation module 520, the third generation module 530, the determination module 540, the fourth generation module 550, the fifth generation module 560, and the sixth generation module 570 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, at least one of the first generation module 510, the second generation module 520, the third generation module 530, the determination module 540, the fourth generation module 550, the fifth generation module 560, and the sixth generation module 570 may be at least partially implemented as a computer program module, which, when executed, may perform the corresponding function.

[0191] It should be noted that the generating device part of the quantum circuit autoencoder in the embodiment of the present application corresponds to the generating method part of the quantum circuit autoencoder in the embodiment of the present application. The description of the generating device part of the quantum circuit autoencoder specifically refers to the generating method part of the quantum circuit autoencoder, which will not be repeated here.

[0192] Figure 6 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present application is shown. Figure 6 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0193] like Figure 6 As shown, an electronic device 600 according to an embodiment of the present application includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage unit 608 into a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present application.

[0194] Various programs and data required for the operation of the electronic device 600 are stored in the RAM 603. The processor 601, ROM 602, and RAM 603 are connected to each other via a bus 604. The processor 601 performs various operations of the method flow according to the embodiment of the present application by executing the programs in the ROM 602 and / or RAM 603. It should be noted that the programs may also be stored in one or more memories other than the ROM 602 and RAM 603. The processor 601 may also perform various operations of the method flow according to the embodiment of the present application by executing the programs stored in the one or more memories.

[0195] According to an embodiment of the present application, electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to bus 604. Electronic device 600 may also include one or more of the following components connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including devices such as a cathode ray tube (CRT), liquid crystal display (LCD), and speakers; a storage section 608 including a hard disk; and a communication section 609 including a network interface card such as a LAN card or modem. Communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. Removable media 611, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed in drive 610 as needed, so that computer programs read from the removable media can be installed into storage section 608 as needed.

[0196] According to an embodiment of the present application, the method flow according to the embodiment of the present application can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, the above-mentioned functions defined in the system of the embodiment of the present application are executed. According to an embodiment of the present application, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0197] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiments of this application is implemented.

[0198] According to embodiments of the present application, a computer-readable storage medium may be a non-volatile computer-readable storage medium. Examples include, but are not limited to, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0199] For example, according to an embodiment of the present application, the computer-readable storage medium may include the ROM 602 and / or the RAM 603 described above and / or one or more memories other than the ROM 602 and the RAM 603 .

[0200] An embodiment of the present application also includes a computer program product, which includes a computer program, which contains program code for executing the method provided by the embodiment of the present application. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the method provided by the embodiment of the present application.

[0201] When the computer program is executed by the processor 601, the above functions defined in the system / device of the embodiment of the present application are performed. According to the embodiment of the present application, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0202] In one embodiment, the computer program may be stored on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may be transmitted and distributed in the form of a signal on a network medium, downloaded and installed via the communication portion 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0203] According to an embodiment of the present application, the program code for executing the computer program provided by the embodiment of the present application can be written in any combination of one or more programming languages. Specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect via the Internet).

[0204] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It will be understood by those skilled in the art that the features described in the various embodiments of the present application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments of the present application may be combined and / or coupled in various ways, and all of these combinations and / or couplings fall within the scope of the present application.

[0205] The embodiments of the present application have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present application. Although each embodiment has been described separately above, this does not mean that the measures in each embodiment cannot be advantageously used in combination. The present application does not depart from the scope of the present application, and those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present application.

Claims

1. A method for generating a quantum circuit autoencoder, characterized in that: include: In response to the compression instruction, an initial quantum circuit encoder and a first quantum channel are generated according to the input quantum channel, wherein each layer of the initial quantum circuit encoder is generated according to a plurality of single-bit noises and parameter-containing quantum circuits; In response to the reconstruction instruction, generating a second quantum channel according to the initial quantum circuit encoder and the first quantum channel; generating a target quantum state according to the second quantum channel and an initial quantum state, wherein the initial quantum state is a state of the original quantum space; Iteratively adjusting parameter values ​​of the initial quantum circuit encoder to determine target parameter values ​​of the single-bit noise and the parameter-containing quantum circuit according to a target loss value of a target quantum state corresponding to the initial quantum circuit encoder; generating a third quantum channel according to a target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed according to the single-bit noise and a target parameter value of the parameter-containing quantum circuit; generating a potential channel and a noise channel according to the third quantum channel; generating a reconstructed quantum channel according to the potential channel, the noise channel, and a target quantum circuit decoder, wherein the target quantum circuit decoder is the conjugate transpose of the target quantum circuit encoder, and the quantum circuit autoencoder includes the target quantum circuit encoder and the target quantum circuit decoder; The step of generating a potential channel and a noise channel according to the third quantum channel includes: performing a detour operation on the third quantum channel to obtain the potential channel when the third quantum channel acts on a target quantum space, wherein the original quantum space includes a potential space and a garbage space, and the target quantum space includes the original quantum space and an isomorphic space corresponding to the garbage space; Extracting a system state of a quantum system corresponding to an isomorphic space in the target quantum space, and measuring the system state in a standard basis to obtain state information; generating the noise channel according to the state information; Iteratively adjusting the parameter value of the initial quantum circuit encoder to determine the single-bit noise and the target parameter value of the parameter-containing quantum circuit according to the target loss value of the target quantum state corresponding to the initial quantum circuit encoder includes: Determining initial parameter values ​​of the single-bit noise and the parameter-containing quantum circuit, so as to input a target quantum state corresponding to the initial parameter values ​​into a target loss function to obtain an initial loss value; The initial parameter value is iteratively adjusted according to the initial loss value to obtain an adjusted parameter value, so as to determine the adjusted parameter value as the target parameter value when an iteration condition is met.

2. The method according to claim 1, characterized in that The target quantum circuit decoder includes a first decoder and a second decoder; The step of generating the reconstructed quantum channel according to the potential channel, the noise channel and the target quantum circuit decoder comprises: generating a target channel according to the potential channel and the noise channel; The reconstructed quantum channel is generated according to the target channel, the first decoder and the second decoder.

3. The method according to claim 1, characterized in that Generating a second quantum channel according to the initial quantum circuit encoder and the first quantum channel includes: Acquire a quantum circuit set, wherein the quantum circuit set includes a plurality of first quantum circuits; Processing the plurality of first quantum circuits based on a preprocessing formula to obtain the first quantum channel; The second quantum channel is generated according to the initial quantum circuit encoder and the first quantum channel.

4. The method according to claim 1 or 3, characterized in that The original quantum space includes potential space and garbage space; Generating a target quantum state according to the second quantum channel and the initial quantum state includes: The maximum mixing state acting on the latent space; The maximum entangled state acting on junk space and its isomorphic space; generating the initial quantum state according to the maximum mixed state and the maximum entangled state; The second quantum channel is applied to the initial quantum state to obtain the target quantum state.

5. The method according to claim 1, wherein Iteratively adjusting the initial parameter value according to the initial loss value to obtain an adjusted parameter value, and determining the adjusted parameter value as the target parameter value when an iteration condition is satisfied, including: During the i-th iteration, the finite difference algorithm is used to calculate the i-th gradient information of the i-th initial loss value corresponding to the i-th iteration; Generating an i+1th initial parameter value for an i+1th iteration according to the i-th initial parameter value and the i-th gradient information in the i-th iteration, and generating an i+1th gradient information according to the i+1th initial parameter value for the i+1th iteration; When a difference between the i-th gradient information and the (i+1)-th gradient information is smaller than a preset gradient threshold, an i-th initial parameter value corresponding to the i-th gradient information is determined as the target parameter value.

6. The method according to claim 1, characterized in that Iteratively adjusting the initial parameter value according to the initial loss value to obtain an adjusted parameter value, and determining the adjusted parameter value as the target parameter value when an iteration condition is satisfied, including: During the i-th iteration, the finite difference algorithm is used to calculate the i-th gradient information of the i-th initial loss value corresponding to the i-th iteration; Generate an i+1th initial parameter value for an i+1th iteration according to the i-th initial parameter value and the i-th gradient information in the i-th iteration process; When the number of iterations i meets a preset iteration threshold, the (i+1)th initial parameter value is determined as the target parameter value.

7. The method according to claim 1, characterized in that The target quantum circuit encoder includes a first encoder and a second encoder; Generating a third quantum channel according to the target quantum circuit encoder and the first quantum channel includes: generating the first encoder and the second encoder respectively according to the target parameter value and the initial quantum circuit encoder; The third quantum channel is generated according to the first encoder, the second encoder and the first quantum channel.

8. A device for generating a quantum circuit autoencoder, characterized in that: include: a first generation module, configured to generate an initial quantum circuit encoder and a first quantum channel according to an input quantum channel in response to a compression instruction, wherein each layer of the initial quantum circuit encoder is generated according to a plurality of single-bit noises and parameter-containing quantum circuits; A second generating module is configured to generate a second quantum channel according to the initial quantum circuit encoder and the first quantum channel in response to a reconstruction instruction; a third generating module, configured to generate a target quantum state according to the second quantum channel and an initial quantum state, wherein the initial quantum state is a state of the original quantum space; a determination module, configured to iteratively adjust parameter values ​​of the initial quantum circuit encoder to determine target parameter values ​​of the single-bit noise and the parameter-containing quantum circuit according to a target loss value of a target quantum state corresponding to the initial quantum circuit encoder; a fourth generating module, configured to generate a third quantum channel based on a target quantum circuit encoder and the first quantum channel, wherein the target quantum circuit encoder is constructed based on the single-bit noise and a target parameter value of the parameter-containing quantum circuit; a fifth generating module, configured to generate a potential channel and a noise channel based on the third quantum channel; a sixth generation module, configured to generate a reconstructed quantum channel based on the potential channel, the noise channel, and a target quantum circuit decoder, wherein the target quantum circuit decoder is the conjugate transpose of the target quantum circuit encoder, and the quantum circuit autoencoder includes the target quantum circuit encoder and the target quantum circuit decoder; The step of generating a potential channel and a noise channel according to the third quantum channel includes: performing a detour operation on the third quantum channel to obtain the potential channel when the third quantum channel acts on a target quantum space, wherein the original quantum space includes a potential space and a garbage space, and the target quantum space includes the original quantum space and an isomorphic space corresponding to the garbage space; Extracting a system state of a quantum system corresponding to an isomorphic space in the target quantum space, and measuring the system state in a standard basis to obtain state information; generating the noise channel according to the state information; Iteratively adjusting the parameter value of the initial quantum circuit encoder to determine the single-bit noise and the target parameter value of the parameter-containing quantum circuit according to the target loss value of the target quantum state corresponding to the initial quantum circuit encoder includes: Determining initial parameter values ​​of the single-bit noise and the parameter-containing quantum circuit, so as to input a target quantum state corresponding to the initial parameter values ​​into a target loss function to obtain an initial loss value; The initial parameter value is iteratively adjusted according to the initial loss value to obtain an adjusted parameter value, so as to determine the adjusted parameter value as the target parameter value when an iteration condition is met.

Images