Gradient estimation method, device, equipment and product of variational quantum feature solver

Abstract

The application discloses a gradient estimation method and device of a variational quantum feature solver, equipment and products, and relates to the technical field of quantum computing. The method comprises the following steps: acquiring molecular interaction information of a target molecular system, and constructing a parameterized quantum circuit according to the molecular interaction information; acquiring current parameter values corresponding to each parameter in the parameterized quantum circuit; performing synchronous measurement on the parameterized quantum circuit based on the current parameter values on each independent measurement region of the reconfigurable atomic array to obtain measurement results corresponding to each independent measurement region; and obtaining gradient values of each parameter according to the measurement results. The application completes synchronous gradient estimation of multiple parameters in one circuit execution cycle by utilizing the multi-region parallel measurement capability of the reconfigurable atomic array, thereby improving the calculation efficiency of quantum chemical simulation.

Description

Technical Field

[0001] This application relates to the field of quantum computing technology, and in particular to a gradient estimation method, apparatus, device and product for a variable quantum characteristic solver. Background Technology

[0002] In the variational quantum feature solver, gradient estimation of the parameterized quantum circuit is the core of parameter optimization, and the efficiency of gradient estimation directly affects the overall convergence speed of the algorithm.

[0003] Existing gradient estimation methods for variational quantum characteristic solvers typically employ parameter shifting rules. This method requires performing two quantum circuit measurements—one forward and one backward—for each parameter in the parameterized quantum circuit. The total number of measurements is proportional to the number of parameters, resulting in a computational complexity of O(P), where P is the number of parameters. As the simulated molecular system increases, the number of parameters grows linearly, leading to a sharp increase in the number of gradient measurements. This results in significant computational time consumption and hardware resource consumption, ultimately making gradient estimation inefficient in quantum chemical simulations. Summary of the Invention

[0004] The main objective of this application is to provide a gradient estimation method, apparatus, device, and product for a variational quantum characteristic solver, aiming to solve the technical problem of how to improve the efficiency of gradient estimation in molecular quantum simulation.

[0005] To achieve the above objectives, this application proposes a gradient estimation method for a variational quantum feature solver. This method is applied to a neutral atom quantum computing platform, which includes a reconfigurable atom array containing several independent measurement regions. The method comprises: Obtain molecular interaction information of the target molecular system, and construct a parameterized quantum circuit based on the molecular interaction information; Obtain the current parameter value corresponding to each parameter in the parameterized quantum circuit; On each of the independent measurement regions of the reconfigurable atom array, the parameterized quantum circuit is synchronously measured based on each of the current parameter values ​​to obtain the measurement results corresponding to each of the independent measurement regions; The gradient values ​​of each parameter are obtained based on the measurement results.

[0006] In one embodiment, the step of synchronously measuring the parameterized quantum circuit based on the current parameter values ​​in each of the independent measurement regions of the reconfigurable atom array to obtain the measurement results corresponding to each of the independent measurement regions includes: In each of the independent measurement regions, the quantum state corresponding to the parameterized quantum circuit is prepared based on each of the current parameter values; Ground state measurements are performed on each of the quantum states to obtain the measurement results corresponding to each independent measurement region.

[0007] In one embodiment, the step of obtaining the gradient value of each parameter based on each of the measurement results includes: Based on the measurement results, the current gradient value of each parameter is obtained through a preset gradient calculation method; Quantum state fidelity information of each independent measurement region is obtained by quantum state tomography. Based on the quantum state fidelity information, variance reduction processing is performed on the current gradient estimates of each parameter to obtain the gradient values ​​of each parameter.

[0008] In one embodiment, the step of obtaining the gradient value of each parameter based on each of the measurement results includes: Based on the measurement results, the current gradient value of each parameter is calculated using a random perturbation approximation algorithm, which includes an adaptive step size and a momentum term. In each iteration of the random perturbation approximation algorithm, the current gradient value is updated according to the adaptive step size and the momentum term to obtain the gradient value of each parameter.

[0009] In one embodiment, before acquiring the molecular interaction information of the target molecular system, the method further includes: Obtain the spatial distribution position of each atomic qubit in the reconfigurable atomic array; Based on the spatial distribution of each atomic qubit, the reconfigurable atomic array is divided into several independent measurement regions.

[0010] In one embodiment, the step of dividing the reconfigurable atom array into several independent measurement regions according to the spatial distribution of each of the atom qubits includes: Based on the spatial distribution of each atomic qubit in the reconfigurable atomic array, each atomic qubit is divided into several sets of atomic qubits; The reconfigurable atomic array is divided into several independent measurement regions corresponding to each set of atomic qubits, wherein the parameter set corresponding to each set of atomic qubits is measured in the corresponding independent measurement region.

[0011] In one embodiment, prior to performing synchronous measurements of the parameterized quantum circuit based on each of the current parameter values ​​in each of the independent measurement regions of the reconfigurable atom array, the method further includes: Based on the spatial distribution of each atomic qubit in the reconfigurable atomic array, the parameters of the parameterized quantum circuit are divided into multiple parameter sets. Each set of parameters is assigned to each independent measurement region, wherein each set of parameters corresponds one-to-one with each independent measurement region.

[0012] Furthermore, to achieve the above objectives, this application also proposes a gradient estimation device for a variational quantum feature solver, the device comprising: A circuit construction module is used to acquire molecular interaction information of the target molecular system and construct a parameterized quantum circuit based on the molecular interaction information. The parameter acquisition module is used to acquire the current parameter value corresponding to each parameter in the parameterized quantum circuit; The synchronous measurement module is used to perform synchronous measurements on the parameterized quantum circuit based on the current parameter values ​​in each independent measurement region of the reconfigurable atomic array, so as to obtain the measurement results corresponding to each independent measurement region. The gradient calculation module is used to obtain the gradient value of each parameter based on the measurement results.

[0013] Furthermore, to achieve the above objectives, this application also proposes a gradient estimation method for a variable quantum feature solver, wherein the device is applied to a neutral atom quantum computing platform, the platform comprising a reconfigurable atom array, the reconfigurable atom array containing several independent measurement regions; The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the parameterized mimetic circuit construction method as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a computer program product comprising a computer program that, when executed by a processor, implements the steps of the gradient estimation method of the variational quantum feature solver as described above.

[0015] One or more technical solutions proposed in this application have at least the following technical effects: This application acquires molecular interaction information of the target molecular system and constructs a parameterized quantum circuit accordingly. It obtains the current parameter values ​​for each parameter in the circuit and performs synchronous measurements on each independent measurement region of a reconfigurable atomic array based on these current parameter values. The measurement results for each region are then obtained, and the gradient values ​​for each parameter are derived from these measurement results. Furthermore, in the actual estimation process, the system utilizes the independent measurement region partitioning capability of the reconfigurable atomic array to allocate measurement tasks for different parameters to different regions for parallel execution. Compared to existing gradient estimation methods that measure each parameter sequentially on a single quantum processor, resulting in a linear increase in the number of measurements with the number of parameters and enormous sampling overhead, this application, by utilizing the multi-region parallel measurement capability of the reconfigurable atomic array, reduces the number of measurement samplings. This allows users to obtain a more efficient gradient estimation process when running a variable quantum characteristic solver, thus improving the computational efficiency of quantum chemical simulations. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the first embodiment of the gradient estimation method for the variable quantum feature solver of this application. Figure 2 This is a flowchart illustrating the second embodiment of the gradient estimation method for the variable quantum feature solver of this application. Figure 3 This is a flowchart illustrating the third embodiment of the gradient estimation method for the variable quantum feature solver of this application. Figure 4 This is a schematic diagram of the module structure of the gradient estimation device of the variable quantum feature solver in an embodiment of this application; Figure 5 This is a schematic diagram of the hardware operating environment involved in the gradient estimation method of the variable quantum feature solver in the embodiments of this application.

[0019] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0021] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0022] In the variational quantum feature solver, gradient estimation of the parameterized quantum circuit is the core of parameter optimization, and the efficiency of gradient estimation directly affects the overall convergence speed of the algorithm.

[0023] Existing gradient estimation methods for variational quantum characteristic solvers typically employ parameter shifting rules. This method requires performing two quantum circuit measurements—one forward and one backward—for each parameter in the parameterized quantum circuit. The total number of measurements is proportional to the number of parameters, resulting in a computational complexity of O(P), where P is the number of parameters. As the simulated molecular system increases, the number of parameters grows linearly, leading to a sharp increase in the number of gradient measurements. This results in significant computational time consumption and hardware resource consumption, ultimately making gradient estimation inefficient in quantum chemical simulations.

[0024] The main solution of this embodiment is as follows: Obtain molecular interaction information of the target molecular system and construct a parameterized quantum circuit accordingly; obtain the current parameter values ​​corresponding to each parameter in the circuit; perform synchronous measurements on the circuit based on the current parameter values ​​in each independent measurement region of the reconfigurable atomic array to obtain the measurement results corresponding to each region; and then obtain the gradient values ​​of each parameter based on the measurement results. Furthermore, in the actual estimation process, the system utilizes the independent measurement region partitioning capability of the reconfigurable atomic array to allocate measurement tasks corresponding to different parameters to different regions for parallel execution. Compared to existing gradient estimation methods that measure each parameter sequentially on a single quantum processor, resulting in a linear increase in the number of measurements with the number of parameters and huge sampling overhead, this application, by utilizing the multi-region parallel measurement capability of the reconfigurable atomic array, can reduce the number of measurement samplings. Therefore, when running a variable quantum characteristic solver, users can obtain a more efficient gradient estimation process, improving the computational efficiency of quantum chemical simulations.

[0025] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of implementing the above functions, a gradient estimation device for a variable quantum feature solver, etc. The following description uses a gradient estimation device for a variable quantum feature solver (hereinafter referred to as the device) as an example to illustrate this embodiment and the following embodiments.

[0026] Based on this, embodiments of this application provide a gradient estimation method based on a variable quantum feature solver, referring to... Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the gradient estimation method for the variable quantum feature solver of this application.

[0027] In this embodiment, the method is applied to the gradient estimation device of the variable quantum feature solver in a neutral atom quantum computing platform. The platform includes a reconfigurable atom array, which contains several independent measurement regions.

[0028] It is understood that the aforementioned neutral atom quantum computing platform can be a hardware system that uses neutral atoms as qubit carriers for quantum computing, such as a quantum computing device containing optical tweezers arrays, Rydberg excitation laser systems, and single-atom addressing systems. The aforementioned variable quantum characteristic solver is a quantum algorithm that prepares test states through parameterized quantum circuits and iteratively solves for the molecular ground state energy using a classical optimizer. The aforementioned gradient estimation device can be a functional module or physical device integrated into the neutral atom quantum computing platform for calculating the gradient of parameterized quantum circuit parameters. The aforementioned reconfigurable atom array can be a two-dimensional or one-dimensional array composed of neutral atoms captured and arranged by optical tweezers, where the spatial positions of the atomic qubits can be dynamically adjusted according to computational needs. The aforementioned independent measurement region can be a spatially separated local region within the reconfigurable atom array capable of independently performing quantum state projection measurements and data readout, with each region containing at least one atomic qubit, and the measurement processes in different regions can be performed in parallel over time.

[0029] The method includes steps S10 to S40: Step S10: Obtain molecular interaction information of the target molecular system, and construct a parameterized quantum circuit based on the molecular interaction information.

[0030] It should be noted that the target molecular system mentioned above can be a chemical molecular system for which the ground state energy is to be determined, such as hydrogen molecules, lithium hydride molecules, or water molecules. The molecular interaction information mentioned above can be physical quantities used to describe the types of interactions between particles within the target molecular system, including single-body rotational interactions, two-body Ising-type interactions, and two-body XY-type interactions. The parameterized quantum circuit mentioned above can be a quantum circuit containing tunable parameters for preparing quantum trial states, whose circuit structure is constructed based on the physical interactions between atomic qubits.

[0031] In its specific implementation, the aforementioned device acquires molecular interaction information of the target molecular system. This acquisition can be achieved through various methods. For example, the device can directly receive pre-processed molecular interaction information input from a host computer, where the types and corresponding coefficients of each interaction term are organized according to a preset data format. Alternatively, based on a preset quantum chemical basis set, the device can generate the second-order quantized Hamiltonian of the target molecular system using classical quantum chemical calculation methods. This Hamiltonian can then be decomposed to extract monomer rotation terms, two-body Ising interaction terms, and two-body XY interaction terms. These extracted interaction terms are then used as the molecular interaction information of the target molecular system.

[0032] After acquiring molecular interaction information, the aforementioned device first determines the basic structure of the parameterized quantum circuit based on the interaction types contained in the molecular interaction information. This basic structure includes a monomer operation layer and a two-body operation layer. The monomer operation layer is used to realize monomer rotational interactions in the molecular interaction information, and the two-body operation layer is used to realize two-body interactions in the molecular interaction information. The device arranges each operation layer according to a preset timing sequence and sets adjustable parameters in the corresponding operation layers, thereby forming a complete parameterized quantum circuit.

[0033] To facilitate understanding, the following explanation uses examples, but does not impose specific limitations on this embodiment. For instance, the molecular interaction information received by the above device corresponds to a hydrogen molecule system. This information includes a monomer Z-rotation term, a two-body Ising interaction term, and a two-body XY interaction term, where the two-body interaction term involves two atomic qubits. Based on this information, the above device constructs a parameterized quantum circuit. The circuit structure first performs a monomer rotation operation acting on the two atomic qubits, then performs a two-body Ising interaction operation, and finally performs a two-body XY interaction operation. Each operation is set with corresponding adjustable parameters, forming a complete parameterized quantum circuit.

[0034] Step S20: Obtain the current parameter values ​​corresponding to each parameter in the parameterized quantum circuit.

[0035] It should be noted that the above parameters can be adjustable variables in the parameterized quantum circuit, used to control the interaction strength or rotation angle of the quantum gate. Their values ​​are continuously updated during the optimization process to approximate the ground state energy of the target molecular system. The above current parameter values ​​can be the specific values ​​taken by each parameter in the parameterized quantum circuit in the current iteration round, which are used to determine the specific form of the current quantum trial state.

[0036] In its implementation, the aforementioned device acquires the current parameter values ​​corresponding to each parameter in the parameterized quantum circuit. During the first optimization iteration, the device reads pre-stored initial parameter values ​​from memory and uses these initial values ​​as the current parameter values ​​for each parameter. These initial parameter values ​​can be determined by generating a set of values ​​within a preset parameter range through random sampling, or by pre-calculating a set of reference values ​​based on prior physical information. During subsequent optimization iterations, the device reads the parameter values ​​stored after the previous iteration update from memory and uses them as the current parameter values. After acquiring the current parameter values ​​for each parameter, the device establishes a one-to-one correspondence between these values ​​and the parameters in the parameterized quantum circuit.

[0037] To facilitate understanding, the following explanation uses examples, but does not limit the scope of this embodiment. For example, consider a parameterized quantum circuit constructed from a hydrogen molecule system, which includes two parameters. In the first iteration, the device reads pre-stored initial parameter values ​​from memory, namely 0.5 and 0.3, and uses these two values ​​as the current parameter values ​​for the two parameters. In subsequent iterations, the device reads the parameter values ​​stored after the optimizer update in the previous round, such as 0.52 and 0.28, and uses them as the current parameter values ​​for the current iteration.

[0038] Step S30: On each of the independent measurement regions of the reconfigurable atom array, the parameterized quantum circuit is synchronously measured based on each of the current parameter values ​​to obtain the measurement results corresponding to each of the independent measurement regions.

[0039] It should be noted that the aforementioned synchronous measurement can be performed within the same time period, with each independent measurement region performing quantum state projection measurements on its corresponding atomic qubits. The measurement operations in each region are performed in parallel in time, without any waiting. The measurement results can be physical quantity data obtained after performing measurements on a parameterized quantum circuit, such as the projection probability distribution of each atomic qubit in different ground states, the state occupancy number of each qubit, or the regional contribution of the expected energy value obtained through statistical analysis of multiple measurements.

[0040] In its implementation, after acquiring the current parameter values ​​for each parameter, the device loads these values ​​into the constructed parameterized quantum circuit, causing the adjustable parameters in the circuit to take their current values. Under its control, the device simultaneously executes the parameterized quantum circuit in each independent measurement region of the reconfigurable atomic array. Since each independent measurement region is spatially separated and has its own independent measurement channel, the device can synchronously acquire quantum state data after circuit operation through the measurement channels of each region within the same time period, without waiting for one region to complete measurement before starting the next. For each independent measurement region, the device performs projection measurement on the atomic qubits contained in that region, acquiring the original measurement signal corresponding to that region. This original measurement signal reflects the state distribution of the atomic qubits in that region under the measurement basis. Subsequently, the device processes the original measurement signals acquired from each region, for example, converting the photon counting signal into the occupancy probability of each quantum state, to obtain the measurement results corresponding to each independent measurement region. The device stores the measurement results of each independent measurement region according to region identifiers to distinguish the contribution of different regions to the global energy expectation value.

[0041] To facilitate understanding, the following explanation uses examples, but does not limit the scope of this embodiment. For example, in a simulation of lithium hydride molecules, the device has divided a 4×4 reconfigurable atomic array into two independent measurement regions, each containing 8 atomic qubits. After acquiring the current parameter values ​​(e.g., 0.5 and 0.3), the device loads these two values ​​into the corresponding parameters of the parameterized quantum circuit. Subsequently, the device simultaneously sends measurement control signals to the 8 atomic qubits in the left and right regions. The detector in the left region acquires the projected distribution of atoms in that region under the Z-basis and calculates the sampling probability of each quantum state in that region. The detector in the right region simultaneously acquires the projected distribution of atoms in that region under the X-basis. The device marks the probability distribution data of the left region as the first measurement result and the probability distribution data of the right region as the second measurement result, and stores them in the memory at the storage locations corresponding to their respective region identifiers.

[0042] Step S40: Obtain the gradient value of each parameter based on the measurement results.

[0043] It should be noted that the gradient value mentioned above can be the rate of change of the objective function (e.g., the expected energy value) relative to a certain parameter in a parameterized quantum circuit. It measures the direction and extent of the influence of a small change in that parameter on the objective function value. This gradient value is the core basis for updating parameters in classical optimizers; its magnitude and sign determine the direction and step size of parameter adjustment in the next iteration. Since there is usually no analytical expression between the objective function and parameters in a parameterized quantum circuit, the gradient value needs to be obtained indirectly through measurement data on quantum hardware combined with specific estimation algorithms.

[0044] In its implementation, after obtaining the measurement results from each independent measurement region, the device calculates the gradient values ​​corresponding to each parameter based on these results. Different gradient estimation algorithms utilize the measurement results differently. When using the finite difference method or parameter shifting method, the device performs multiple circuit measurements on each parameter to obtain the objective function value under different parameter offsets. The measurement results from each independent measurement region correspond to the contribution of the objective function in different subspaces. The device summarizes these contributions to obtain the complete objective function value and then estimates the gradient based on the difference between the objective function values ​​obtained from multiple measurements. When using the random perturbation approximation algorithm, the device simultaneously applies random perturbations to all parameters to obtain the change in the objective function. The measurement results from each independent measurement region synchronously provide the change in the objective function value in each subspace before and after the perturbation. The device combines these changes and uses them to estimate the gradient vector. Regardless of the estimation algorithm used, the device uses the measurement results output from each independent measurement region as the basic data. After summarizing and mathematical operations, it obtains the gradient values ​​corresponding to each parameter and stores these gradient values ​​as gradient vectors in parameter number order.

[0045] To facilitate understanding, the following example illustrates the concept, but does not limit the scope of this embodiment. For instance, consider a parameterized quantum circuit for a lithium hydride molecule, containing 10 parameters. The device divides a 4×4 atom array into two independent measurement regions. In a certain iteration, the device simultaneously applies perturbation to all 10 parameters using a random perturbation approximation algorithm. The measurement results in the left region provide the contribution data of the objective function to the atoms in the left half before perturbation, and the measurement results in the right region provide the contribution data of the objective function to the atoms in the right half before perturbation. The device merges the data from both sides to obtain the global objective function value before perturbation. Subsequently, the device executes the perturbation pulse again, and the left and right regions synchronously complete the measurement after perturbation to obtain the global objective function value after perturbation. Based on the difference between the global objective function values ​​before and after perturbation and the preset perturbation step size, the device calculates the gradient values ​​of all 10 parameters in the current iteration and arranges them into a gradient vector according to the parameter numbers 1 to 10.

[0046] This embodiment acquires molecular interaction information of the target molecular system and constructs a parameterized quantum circuit accordingly. It obtains the current parameter values ​​for each parameter in the circuit and performs synchronous measurements on each independent measurement region of the reconfigurable atomic array based on these current parameter values ​​to obtain measurement results for each region. Then, it obtains the gradient values ​​for each parameter based on these measurement results. Furthermore, in the actual estimation process, the system utilizes the independent measurement region partitioning capability of the reconfigurable atomic array to allocate measurement tasks corresponding to different parameters to different regions for parallel execution. Compared to existing gradient estimation methods that measure each parameter sequentially on a single quantum processor, resulting in a linear increase in the number of measurements with the number of parameters and huge sampling overhead, this application, by utilizing the multi-region parallel measurement capability of the reconfigurable atomic array, can reduce the number of measurement samplings. Therefore, when running a variable quantum characteristic solver, users can obtain a more efficient gradient estimation process, improving the computational efficiency of quantum chemical simulations.

[0047] Based on the first embodiment of this application, a second embodiment of this application is proposed. In the second embodiment, content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the gradient estimation method for the variable quantum feature solver of this application.

[0048] In this embodiment, the step of synchronously measuring the parameterized quantum circuit based on the current parameter values ​​in each of the independent measurement regions of the reconfigurable atom array to obtain the measurement results corresponding to each of the independent measurement regions includes: Step S31: In each of the independent measurement regions, prepare the quantum state corresponding to the parameterized quantum circuit based on each of the current parameter values; Step S32: Perform ground state measurement on each of the quantum states to obtain the measurement results corresponding to each independent measurement region.

[0049] It should be noted that the aforementioned quantum state can be the state of an atomic qubit after evolving from its initial state under the action of a parameterized quantum circuit. This state is jointly determined by the values ​​of various parameters in the circuit and reflects the approximate wave function of the target molecular system under the current parameter configuration. The aforementioned ground state measurement can be achieved by projecting the atomic qubit onto a preset measurement basis (e.g., Z-basis or X-basis), and obtaining the probability distribution information of the quantum state under that basis by collecting the projection results of each qubit under the measurement basis. This probability distribution information can be approximated by the statistical results after multiple repeated measurements.

[0050] In its implementation, the aforementioned device, within each independent measurement region, loads the current parameter values ​​into the parameterized quantum circuit and then sends circuit execution instructions to the atomic qubits in each region. Each region synchronously executes the complete parameterized quantum circuit operation. These circuit operations act on each atomic qubit within the region, causing them to evolve from an initial state to a quantum state determined by the current parameter values ​​after a series of quantum gate operations. After each independent measurement region executes the aforementioned circuit operations, the atomic qubits within each region are in their corresponding quantum states.

[0051] Subsequently, the aforementioned device performs ground-state measurements on the prepared quantum states in each independent measurement region. For each independent measurement region, the device applies a measurement pulse to the atomic qubits within that region, projecting the quantum state of the atomic qubits onto a preset measurement basis, and reads the projection results of each atomic qubit through a detector. The device independently performs the aforementioned ground-state measurements in each independent measurement region, with the measurement processes in each region occurring synchronously in time. The device records the projection results measured in each region according to the region identifier, with each region corresponding to a set of measurement results. These measurement results reflect the state distribution information of the atomic qubits in that region under the current parameter values.

[0052] Furthermore, it should be noted that the aforementioned ground-state measurements typically need to be repeated multiple times to obtain statistically valid data. When the aforementioned device performs the circuit fabrication and ground-state measurement, for each independent measurement region, the device repeats the process according to a preset number of sampling times, accumulating the counts of different projection results in each measurement to obtain the measurement results corresponding to each region.

[0053] To facilitate understanding, the following explanation uses examples, but does not limit the scope of this embodiment. For instance, in a simulation of lithium hydride molecules, the device divides a 4×4 atomic array into two independent measurement regions, left and right. The left region contains 8 atomic qubits, and the right region contains 8 atomic qubits. After acquiring the current parameter values, the device loads these two values ​​into the parameterized quantum circuit. Then, the device simultaneously sends circuit execution instructions to all 16 atomic qubits in both regions. The 8 atomic qubits in the left region and the 8 atomic qubits in the right region each execute the complete parameterized quantum circuit operation, evolving into the left quantum state and the right quantum state, respectively. Subsequently, the device applies Z-basis measurement pulses to the eight atomic qubits in the left region, reads the projection results of each qubit under the Z-basis using a detector, and accumulates 1024 measurement results to obtain the probability distribution data of the eight qubits in the left region under the Z-basis, which is taken as the measurement result for the left region. Simultaneously, the device applies X-basis measurement pulses to the eight atomic qubits in the right region, accumulates 1024 measurement results, and obtains the probability distribution data of the eight qubits in the right region under the X-basis, which is taken as the measurement result for the right region. The device stores the measurement results from both sides in memory according to region identifiers.

[0054] Furthermore, in this embodiment, the step of obtaining the gradient value of each parameter based on each measurement result includes: Step S41: Based on the measurement results, obtain the current gradient value of each parameter using a preset gradient calculation method; Step S42: Obtain quantum state fidelity information for each of the independent measurement regions by quantum state tomography; Step S43: Based on the quantum state fidelity information, perform variance reduction processing on the current gradient estimates of each parameter to obtain the gradient values ​​of each parameter.

[0055] It should be noted that the aforementioned preset gradient calculation method can be a pre-defined mathematical method used to calculate the parameter gradient based on the measurement results of the quantum circuit. Examples include numerical gradient calculation based on the finite difference method, analytical gradient calculation based on the parameter shift rule, or approximate gradient calculation based on random perturbation approximation. The aforementioned quantum state tomography can be a technique that involves performing multiple sets of projection measurements on the quantum state under different bases and using the measurement results to reconstruct the quantum state density matrix or extract key feature information of the quantum state. The aforementioned quantum state fidelity information can be used to describe the closeness between the actually prepared quantum state and the theoretical target quantum state; its value is between 0 and 1, with the value closer to 1 indicating a closer consistency between the actual quantum state and the theoretical state. The aforementioned variance reduction processing can be an operation that reduces the statistical fluctuation amplitude of the gradient estimate by introducing auxiliary information, resulting in a smaller dispersion of the gradient estimate between different measurements.

[0056] In its implementation, after obtaining the measurement results from each independent measurement area, the device calculates the current gradient estimate for each parameter using a preset gradient calculation method. This preset gradient calculation method defines the specific conversion relationship from the measurement results to the gradient estimate. The device substitutes the measurement results from each independent measurement area into this conversion relationship for calculation. When using the parameter shifting method, the device uses the energy values ​​measured for each parameter under both positive and negative shifts, dividing the difference by twice the shift amount to obtain the gradient estimate for that parameter. When using the random perturbation approximation algorithm, the device uses the global energy values ​​before and after applying the perturbation to all parameters, calculating the proportional relationship between the energy change and the perturbation vector to obtain the gradient estimate for each parameter. Regardless of the method used, after completing the calculation, the device obtains a set of values ​​equal to the number of parameters, serving as the current gradient estimate for each parameter under its current parameter value.

[0057] Next, the device performs multiple sets of projection measurements under different bases on each independent measurement region, covering multiple measurement directions to obtain sufficient state information. Based on these multi-base measurement data, the device uses a quantum state reconstruction algorithm to recover the density matrix of the actual quantum state in each region, or extract key indicators that characterize the quality of the quantum state. Subsequently, the device compares the reconstructed actual quantum state with the theoretical target quantum state to calculate the fidelity between the two. The theoretical target quantum state is a quantum state pre-calculated based on the current parameter values ​​and an ideal noise-free circuit model. Through the above comparison, the device obtains the quantum state fidelity information corresponding to each independent measurement region, which reflects the degree to which each region is affected by noise during actual hardware execution. Among them, regions with higher fidelity have higher reliability of measurement results and fewer noise components in the contributed gradient estimates; regions with lower fidelity have more severe hardware noise contamination of measurement results and more noise components in the contributed gradient estimates.

[0058] Subsequently, the aforementioned device performs variance reduction processing on the current gradient estimates of each parameter based on the quantum state fidelity information, obtaining the final gradient values ​​for each parameter. During the variance reduction process, the device uses the quantum state fidelity of each independent measurement region as a weighting factor to adjust the gradient components contributed by each region. Regions with higher fidelity have a higher weight in the final gradient value, while regions with lower fidelity have a lower weight. Through this weighted processing method, the device can suppress the adverse effects of heavily noise-contaminated regions on the gradient estimates and reduce the statistical variance of the gradient estimates across multiple iterations.

[0059] To facilitate understanding, the following explanation uses examples, but does not limit the scope of this embodiment. For instance, in a simulation of lithium hydride molecules, the aforementioned device divides a 4×4 atomic array into two independent measurement regions, left and right. In a certain iteration, the device calculates the current gradient estimate of each parameter using a random perturbation approximation algorithm, where the current gradient estimate of the first parameter is 0.12. Subsequently, the device performs quantum state tomography on both regions. On the left region, it reconstructs the density matrix of the actual quantum state through six sets of projection measurements in different directions and compares it with the ideal state, calculating a fidelity of 0.92 for the left region. On the right region, the same method is used to calculate a fidelity of 0.78. Based on the fidelity information of the left and right regions, the device weights the components contributed by the left and right regions in the current gradient estimates of each parameter, multiplying the component contributed by the left region by a weight of 0.92 and the component contributed by the right region by a weight of 0.78. After the above weighting process, the gradient value of the first parameter after variance reduction is adjusted from the original 0.12 to 0.11. After performing the same variance reduction process on all parameters, the device stores the processed gradient values ​​of each parameter as the final gradient value vector.

[0060] Furthermore, in this embodiment, the platform further includes a laser, and the step of obtaining the gradient values ​​of each parameter based on each measurement result includes: Step S45: Based on the measurement results, calculate the current gradient value of each parameter using a random perturbation approximation algorithm, which includes an adaptive step size and a momentum term; Step S46: In each iteration of the random perturbation approximation algorithm, the current gradient value is updated according to the adaptive step size and the momentum term to obtain the gradient value of each parameter.

[0061] It should be noted that the aforementioned stochastic perturbation approximation algorithm can be a numerical optimization method that estimates the gradient solely through changes in the objective function value without calculating analytical gradients. Its core idea is to simultaneously apply random perturbations to all parameters in each iteration and approximate the gradient direction of each parameter based on the change in the objective function value before and after the perturbation. The aforementioned adaptive step size can be a perturbation amount or update step size dynamically adjusted according to the current iteration state during algorithm iteration, used to control the magnitude of parameter updates to balance convergence speed and convergence stability. The aforementioned momentum term can be a weighted accumulation of historical gradients introduced during parameter updates to smooth random fluctuations in gradient estimation, making the parameter update direction more stable and avoiding optimization oscillations caused by excessive errors in a single gradient estimation.

[0062] In its implementation, the device first acquires the current parameter values ​​of all parameters in the current iteration. Then, it generates a set of random perturbation vectors, where each component corresponds one-to-one with a parameter. Each component is sampled from a preset probability distribution, such as a Bernoulli distribution with zero mean or a normal distribution. The device then superimposes the random perturbation vectors onto the current parameter values ​​in both positive and negative directions, obtaining positive and negative perturbation parameter combinations. The device controls the reconfigurable atom array to execute and measure parameterized quantum circuits under the positive perturbation parameter combination in each independent measurement region, obtaining measurement results for each region under the positive perturbation. Then, it executes and measures parameterized quantum circuits under the negative perturbation parameter combination, obtaining measurement results for each region under the negative perturbation. Based on the measurement results under both positive and negative perturbations, the device summarizes the global objective function values ​​under both positive and negative perturbations. The aforementioned device divides the difference between the global objective function value under positive perturbation and the global objective function value under negative perturbation by twice the perturbation step size, and then multiplies it by each component of the random perturbation vector to obtain the current gradient value of each parameter.

[0063] In each iteration of the stochastic perturbation approximation algorithm, the aforementioned device updates the current gradient value based on an adaptive step size and a momentum term. The adaptive step size in the current iteration is dynamically determined according to the iteration progress; for example, it can gradually decrease with increasing iterations to converge to a more accurate solution, or it can adaptively adjust based on the rate of change of the objective function to escape local optima. After calculating the current gradient value, the device first obtains the momentum term accumulated in the previous iteration, which is a weighted sum of historical gradient information. Then, the device combines the current gradient value with the momentum term from the previous iteration using a preset weighting coefficient to obtain the updated momentum term for this iteration. Next, based on the updated momentum term and the adaptive step size, the device calculates the update amount for each parameter in this iteration; this update amount is equal to the product of the adaptive step size and the updated momentum term. Finally, the device adds the corresponding update amount to the current parameter value for each parameter to obtain the updated parameter value. The device uses the updated parameter values ​​as the current parameter values ​​for the next iteration, and stores the updated momentum term for use in the next iteration. Through this method, the device processes the current gradient value using an adaptive step size and momentum term in each iteration to obtain the actual gradient values ​​used to update each parameter in the current iteration.

[0064] To facilitate understanding, the following explanation uses examples, but does not limit the scope of this embodiment. For example, in the simulation of lithium hydride molecules, the parameterized quantum circuit contains 10 parameters, and the device uses a random perturbation approximation algorithm in a certain iteration. The device first generates a random perturbation vector containing 10 components, each taken from a standard normal distribution, with a perturbation step size set to 0.01. The device superimposes the current parameter values ​​with the positive and negative perturbation vectors respectively, performs two circuit measurements, and obtains the objective function value under the positive perturbation (-8.060 Hartree) and the objective function value under the negative perturbation (-8.075 Hartree), respectively. The difference between the two is 0.015 Hartree. Dividing this by twice the perturbation step size of 0.02 and then multiplying by each component of the random perturbation vector, the current gradient value of each parameter is obtained. Subsequently, the device obtains the momentum term accumulated in the previous iteration, where the values ​​of each component are the weighted cumulative amounts of historical gradients in the previous iteration. The device weights the current gradient value with the momentum term from the previous iteration at a ratio of 0.5 to 0.5 to obtain the updated momentum term for this iteration. The device calculates the adaptive step size based on the current iteration number; the current iteration step size is 0.02. This adaptive step size is multiplied by the updated momentum term to obtain the update amount for each parameter in this iteration; for example, the update amount for the first parameter is 0.003. The device adds the current value of the first parameter (0.52) to the update amount (0.003) to obtain 0.523, which is used as the value for that parameter in the next iteration. After updating all 10 parameters one by one, the device stores the updated parameter values ​​and the updated momentum term for this iteration for use in the next iteration.

[0065] Based on the first and second embodiments of this application, a third embodiment of this application is proposed. In this third embodiment, content that is the same as or similar to the first and second embodiments described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 , Figure 3 This is a flowchart illustrating the third embodiment of the gradient estimation method for the variable quantum feature solver of this application.

[0066] In this embodiment, before obtaining the molecular interaction information of the target molecular system, the following steps are also included: Step S11: Obtain the spatial distribution position of each atomic qubit in the reconfigurable atomic array; Step S12: Divide the reconfigurable atom array into several independent measurement regions according to the spatial distribution of each atomic qubit.

[0067] It should be noted that the aforementioned spatial distribution positions can be the specific coordinate data of each atomic qubit in the two-dimensional plane of the reconfigurable atomic array, such as the horizontal and vertical coordinate values ​​in micrometers. This data reflects the relative distance between the atomic qubits and the overall geometric arrangement of the array. The aforementioned independent measurement regions can be several sub-regions divided in the reconfigurable atomic array according to the spatial distribution positions of the atomic qubits. Each sub-region is spatially separated and contains at least one atomic qubit. The measurement processes between different regions are physically independent and can independently perform quantum state projection measurements and data readout.

[0068] In its implementation, the aforementioned device obtains the real-time spatial coordinates of each atomic qubit by reading the configuration parameters of the optical tweezers array. These spatial coordinates reflect the specific arrangement of each atomic qubit in a two-dimensional plane, i.e., the relative distance between each atomic qubit and the overall geometric arrangement of the array. This coordinate data is typically recorded in micrometers. These spatial coordinates are determined by the preceding atomic layout optimization process; different molecular simulation tasks correspond to different atomic spatial arrangement methods. For example, for strongly coupled atomic pairs, they are arranged in close proximity within the Rydberg blocking radius; for weakly coupled atomic pairs, they are arranged in distant proximity beyond the blocking radius.

[0069] After acquiring the spatial distribution of each atomic qubit, the aforementioned device analyzes the spatial aggregation and distribution density of the atomic qubits to identify several sub-regions with relatively concentrated atomic distribution, or several groups of atoms with large interatomic spacing and natural separation. The device adheres to the principle of spatial separation between independent measurement regions during partitioning, ensuring that measurement signals between regions do not interfere with each other. Simultaneously, each independent measurement region contains at least one atomic qubit to ensure that the region can independently perform effective quantum state measurements. After completing the region partitioning, the device records the region affiliation identifier of each atomic qubit and stores the boundary information and the list of atomic qubits contained in each independent measurement region, so that circuit measurements can be performed separately for each region during subsequent gradient estimation.

[0070] For ease of understanding, the following explanation uses examples, but does not limit the specific implementation of this embodiment. For example, in the neutral atom quantum computing platform where the above-mentioned device is located, the reconfigurable atom array contains 16 atomic qubits arranged in a 4×4 two-dimensional array, with the spacing between each atom between 5 micrometers and 12 micrometers. The above-mentioned device obtains the current spatial distribution position of these 16 atomic qubits by reading the control parameters of the optical tweezers array, and obtains the horizontal and vertical coordinates of each atom in the two-dimensional plane. Based on these spatial coordinates, the above-mentioned device identifies that the 8 atoms on the left are clustered into a dense region in space, and the 8 atoms on the right are clustered into another dense region, with a large spacing separating the atoms on the left and right sides. Based on this, the above-mentioned device divides the 4×4 array horizontally into two independent measurement regions. The left region contains 8 atoms in columns 1 to 2, and the right region contains 8 atoms in columns 3 to 4. The spatial distance between the two regions is sufficient to ensure that the measurement signals do not interfere with each other. The above-mentioned device records the atom numbers of the left and right regions respectively, and uses this region division scheme as the basis for the region configuration of synchronous measurement in the subsequent gradient estimation process.

[0071] Furthermore, the step of dividing the reconfigurable atomic array into several independent measurement regions according to the spatial distribution of each atomic qubit includes: Step S121: Based on the spatial distribution position of each atomic qubit in the reconfigurable atomic array, divide each atomic qubit into several sets of atomic qubits; Step S122: Divide the reconfigurable atom array into several independent measurement regions corresponding to each set of atomic qubits, wherein the parameter sets corresponding to each set of atomic qubits are measured in their respective independent measurement regions.

[0072] It should be noted that the aforementioned set of atomic qubits can be several subsets obtained by grouping all atomic qubits in the reconfigurable atomic array according to their spatial distribution. Each subset contains at least one atomic qubit, and the subsets are spatially separated. The aforementioned parameter set can be the set of parameters associated with a certain set of atomic qubits in a parameterized quantum circuit, that is, the adjustable parameters corresponding to the quantum gate operations carried by that set of atomic qubits. The aforementioned measurement can be performed by executing the parameterized quantum circuit within the corresponding independent measurement region and performing quantum state projection measurements on the atomic qubits within that region to obtain the measurement data corresponding to the parameter set associated with that region.

[0073] In its implementation, after acquiring the spatial coordinates of each atomic qubit, the device analyzes the distribution characteristics of these coordinates in a two-dimensional plane, identifying several regions with a high degree of atomic aggregation. Based on this spatial aggregation, the device groups spatially adjacent atomic qubits into the same group and spatially distant atomic qubits into different groups, thus forming several sets of atomic qubits. Each set of atomic qubits is spatially separated, and each set contains at least one atomic qubit. After grouping, the device records the correspondence between each atomic qubit and its group in its memory.

[0074] Subsequently, the device determines the spatial boundary of each independent measurement region based on the atomic qubits contained in each set of atomic qubits and their spatial locations. For each set of atomic qubits, the device defines the spatial range occupied by each atomic qubit in the set as an independent measurement region. The boundary of this region is sufficient to contain all atomic qubits in the set and maintains sufficient spatial spacing with adjacent regions to ensure that the measurement signals do not interfere with each other.

[0075] After the region partitioning is completed, each parameter in the parameterized quantum circuit is assigned to a corresponding independent measurement region based on the spatial assignment of the atomic qubits it affects. The parameter set corresponding to a set of atomic qubits is executed and measured only within that corresponding independent measurement region. Each independent measurement region executes its own circuit operations and measurements corresponding to its parameter set simultaneously, without interfering with each other.

[0076] To facilitate understanding, the following explanation uses examples, but does not limit the scope of this embodiment. For instance, in a neutral atom quantum computing platform containing 16 atomic qubits arranged in a 4×4 two-dimensional array, after acquiring the spatial coordinates of each atom, the device identifies four atoms in the upper left, upper right, lower left, and lower right corners as four independent dense regions based on their coordinate distribution. The device then divides all atoms into four sets of atomic qubits: the first set contains the four atoms in the upper left corner, the second set contains the four atoms in the upper right corner, the third set contains the four atoms in the lower left corner, and the fourth set contains the four atoms in the lower right corner. Subsequently, the device divides the reconfigurable atom array into four independent measurement regions, with the boundary of each region surrounding the corresponding set of atoms. In the subsequent gradient estimation process, the device assigns each parameter in the parameterized quantum circuit to the corresponding region according to the atom it acts upon. For example, the parameter acting on the first set of atoms is measured in the first region, the parameter acting on the second set of atoms is measured in the second region, and the measurement operations in all four regions are performed synchronously.

[0077] Furthermore, before performing synchronous measurements of the parameterized quantum circuit based on the current parameter values ​​in each of the independent measurement regions of the reconfigurable atom array, the method further includes: Step S31: Based on the spatial distribution position of each atomic qubit in the reconfigurable atomic array, divide each parameter of the parameterized quantum circuit into multiple parameter sets.

[0078] Step S32: Assign each set of parameters to each independent measurement region, wherein each set of parameters corresponds one-to-one with each independent measurement region.

[0079] It should be noted that the spatial distribution mentioned above refers to the specific coordinate distribution of each atomic qubit in the two-dimensional plane of the reconfigurable atomic array. This distribution determines the physical distance between atomic qubits and the spatial region to which each atomic qubit belongs in the array.

[0080] The aforementioned parameterized quantum circuit is a quantum circuit containing adjustable parameters. The parameters in this circuit are used to control physical quantities such as the angle or phase of the quantum state evolution, and their values ​​are continuously updated during the optimization iteration process.

[0081] The parameter sets mentioned above are collections of some of the adjustable parameters in a parameterized quantum circuit, with each parameter set containing one or more adjustable parameters.

[0082] The aforementioned independent measurement regions refer to multiple non-overlapping sub-regions in space where the reconfigurable atom array is divided. Each sub-region contains at least one atomic qubit. The independent measurement regions are spatially isolated from each other and can perform quantum state measurement operations simultaneously and independently without mutual interference.

[0083] In its specific implementation, after constructing the parameterized quantum circuit and spatially arranging the atomic qubits, the device acquires the spatial distribution position of each atomic qubit in the reconfigurable atomic array. This spatial distribution position is determined by the spatial coordinates generated by the aforementioned graph embedding layout calculation. Based on the acquired spatial distribution positions, the device divides all tunable parameters contained in the parameterized quantum circuit into multiple parameter sets. This division process is based on the spatial distribution position of each atomic qubit. The device acquires the identification information of the atomic qubit associated with each tunable parameter, and then, based on the spatial coordinates of that atomic qubit in the array, assigns the tunable parameter to the parameter set corresponding to the corresponding spatial region. After completing the division of all tunable parameters, the device obtains multiple parameter sets, each corresponding to a spatial region in the reconfigurable atomic array, and all tunable parameters associated with atomic qubits within the same spatial region are assigned to the same parameter set. Subsequently, the device acquires the configuration information of the independent measurement regions of the reconfigurable atomic array. This configuration information records the spatial boundary range of each independent measurement region and the identification information of the atomic qubits contained in each independent measurement region. The device assigns parameter sets to each independent measurement region based on the identification information of the atomic qubits contained within that region. This ensures that each parameter set is assigned to the independent measurement region containing the corresponding atomic qubit, establishing a one-to-one mapping between parameter sets and independent measurement regions. After this assignment, each independent measurement region is only responsible for the measurement tasks related to its corresponding parameter set.

[0084] For ease of understanding, the following explanation uses examples, but does not impose specific limitations on this embodiment. For instance, assume the reconfigurable atomic array is a 2×2 two-dimensional array containing four atomic qubits, located at the top left, top right, bottom left, and bottom right spatial positions of the array. The aforementioned graph embedding layout calculation has already assigned corresponding spatial coordinates to these four atomic qubits. The parameterized quantum circuit contains eight adjustable parameters, where the first and second parameters are associated with the atomic qubit at the top left position, the third and fourth parameters are associated with the atomic qubit at the top right position, the fifth and sixth parameters are associated with the atomic qubit at the bottom left position, and the seventh and eighth parameters are associated with the atomic qubit at the bottom right position. After acquiring the spatial coordinates of these four atomic qubits in the array, the device divides the eight parameters into four parameter sets based on the spatial positions of the atomic qubits associated with each parameter: the first parameter set contains the first and second parameters, corresponding to the top left region; the second parameter set contains the third and fourth parameters, corresponding to the top right region; the third parameter set contains the fifth and sixth parameters, corresponding to the bottom left region; and the fourth parameter set contains the seventh and eighth parameters, corresponding to the bottom right region. The reconfigurable atom array is divided into four independent measurement regions, each containing one atom qubit. The device assigns the first set of parameters to the independent measurement region in the upper left, the second set of parameters to the independent measurement region in the upper right, the third set of parameters to the independent measurement region in the lower left, and the fourth set of parameters to the independent measurement region in the lower right.

[0085] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the gradient estimation method of the variable quantum feature solver of this application. Any simple transformations based on this technical concept are within the protection scope of this application.

[0086] This application also provides a gradient estimation device for a variational quantum feature solver; please refer to [reference needed]. Figure 4 The device includes: The circuit construction module 10 is used to acquire molecular interaction information of the target molecular system and construct a parameterized quantum circuit based on the molecular interaction information. Parameter acquisition module 20 is used to acquire the current parameter value corresponding to each parameter in the parameterized quantum circuit; The synchronous measurement module 30 is used to perform synchronous measurements on the parameterized quantum circuit based on the current parameter values ​​in each independent measurement region of the reconfigurable atom array, so as to obtain the measurement results corresponding to each independent measurement region. The gradient calculation module 40 is used to obtain the gradient value of each parameter based on the measurement results.

[0087] The gradient estimation device for the variable quantum feature solver provided in this application employs the gradient estimation method of the variable quantum feature solver in the above embodiments, and can solve the technical problem of how to improve the gradient estimation efficiency in molecular quantum simulation. Compared with the prior art, the beneficial effects of the gradient estimation device for the variable quantum feature solver provided in this application are the same as the beneficial effects of the gradient estimation method of the variable quantum feature solver provided in the above embodiments, and other technical features in the gradient estimation device of the variable quantum feature solver are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0088] This application provides a gradient estimation device for a variable quantum feature solver, which is applied to a neutral atom quantum computing platform. The platform includes a reconfigurable atom array, which contains several independent measurement regions. The gradient estimation device of the variable quantum feature solver includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the gradient estimation method of the variable quantum feature solver in Embodiment 1 above.

[0089] The following is for reference. Figure 5 This document illustrates a schematic diagram of a gradient estimation device suitable for implementing the variable quantum feature solver in the embodiments of this application. The gradient estimation device for the variable quantum feature solver in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The gradient estimation device of the variable quantum feature solver shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0090] like Figure 5As shown, the gradient estimation device of the variable quantum characteristic solver may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1002 or a program loaded from storage device 1003 into random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the gradient estimation device of the variable quantum characteristic solver. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the gradient estimation device of the variational quantum characteristic solver to communicate wirelessly or wiredly with other devices to exchange data. Although gradient estimation devices of the variational quantum characteristic solver with various systems are shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0091] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0092] The gradient estimation device for the variable quantum feature solver provided in this application employs the gradient estimation method of the variable quantum feature solver in the above embodiments, and can solve the technical problem of how to improve the gradient estimation efficiency in molecular quantum simulation. Compared with the prior art, the beneficial effects of the gradient estimation device for the variable quantum feature solver provided in this application are the same as the beneficial effects of the gradient estimation method of the variable quantum feature solver provided in the above embodiments, and other technical features in the gradient estimation device of the variable quantum feature solver are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0093] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0094] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0096] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0097] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A gradient estimation method for a variational quantum feature solver, applied to a neutral atom quantum computing platform, the platform comprising a reconfigurable atom array, the reconfigurable atom array containing several independent measurement regions, characterized in that... The method includes: Obtain molecular interaction information of the target molecular system, and construct a parameterized quantum circuit based on the molecular interaction information; Obtain the current parameter value corresponding to each parameter in the parameterized quantum circuit; On each of the independent measurement regions of the reconfigurable atom array, the parameterized quantum circuit is synchronously measured based on each of the current parameter values ​​to obtain the measurement results corresponding to each of the independent measurement regions; The gradient values ​​of each parameter are obtained based on the measurement results.

2. The method as described in claim 1, characterized in that, The step of synchronously measuring the parameterized quantum circuit based on the current parameter values ​​in each of the independent measurement regions of the reconfigurable atom array to obtain the measurement results corresponding to each of the independent measurement regions includes: In each of the independent measurement regions, the quantum state corresponding to the parameterized quantum circuit is prepared based on each of the current parameter values; Ground state measurements are performed on each of the quantum states to obtain the measurement results corresponding to each independent measurement region.

3. The method as described in claim 1, characterized in that, The step of obtaining the gradient value of each parameter based on each measurement result includes: Based on the measurement results, the current gradient value of each parameter is obtained through a preset gradient calculation method; Quantum state fidelity information of each independent measurement region is obtained by quantum state tomography. Based on the quantum state fidelity information, variance reduction processing is performed on the current gradient estimates of each parameter to obtain the gradient values ​​of each parameter.

4. The method as described in claim 1, characterized in that, The step of obtaining the gradient value of each parameter based on each measurement result includes: Based on the measurement results, the current gradient value of each parameter is calculated using a random perturbation approximation algorithm, which includes an adaptive step size and a momentum term. In each iteration of the random perturbation approximation algorithm, the current gradient value is updated according to the adaptive step size and the momentum term to obtain the gradient value of each parameter.

5. The method as described in claim 1, characterized in that, Before obtaining the molecular interaction information of the target molecular system, the method further includes: Obtain the spatial distribution position of each atomic qubit in the reconfigurable atomic array; Based on the spatial distribution of each atomic qubit, the reconfigurable atomic array is divided into several independent measurement regions.

6. The method as described in claim 5, characterized in that, The step of dividing the reconfigurable atomic array into several independent measurement regions according to the spatial distribution of each atomic qubit includes: Based on the spatial distribution of each atomic qubit in the reconfigurable atomic array, each atomic qubit is divided into several sets of atomic qubits; The reconfigurable atomic array is divided into several independent measurement regions corresponding to each set of atomic qubits, wherein the parameter set corresponding to each set of atomic qubits is measured in the corresponding independent measurement region.

7. The method as described in claim 1, characterized in that, Before performing synchronous measurements of the parameterized quantum circuit based on the current parameter values ​​in each of the independent measurement regions of the reconfigurable atom array, the method further includes: Based on the spatial distribution of each atomic qubit in the reconfigurable atomic array, the parameters of the parameterized quantum circuit are divided into multiple parameter sets. Each set of parameters is assigned to each independent measurement region, wherein each set of parameters corresponds one-to-one with each independent measurement region.

8. A gradient estimation device for a variational quantum feature solver, characterized in that, The device includes: A circuit construction module is used to acquire molecular interaction information of the target molecular system and construct a parameterized quantum circuit based on the molecular interaction information. The parameter acquisition module is used to acquire the current parameter value corresponding to each parameter in the parameterized quantum circuit; The synchronous measurement module is used to perform synchronous measurements on the parameterized quantum circuit based on the current parameter values ​​in each independent measurement region of the reconfigurable atomic array, so as to obtain the measurement results corresponding to each independent measurement region. The gradient calculation module is used to obtain the gradient value of each parameter based on the measurement results.

9. A gradient estimation device for a variational quantum feature solver, characterized in that, The device is applied to a neutral atom quantum computing platform, which includes a reconfigurable atom array containing several independent measurement regions. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the gradient estimation method for the variable quantum feature solver as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the gradient estimation method for the variable quantum feature solver as described in any one of claims 1 to 7.

Images