A multi-body quantum phase classification method, system, terminal and storage medium based on shallow dynamics interlacing

Abstract

The application relates to the technical field of quantum computing, and discloses a multi-body quantum phase classification method, system, terminal and storage medium based on shallow layer dynamics interlacing. The method comprises the following steps: obtaining the dynamic access authority of a target quantum system, wherein the dynamic access authority comprises the authority of applying a Hamiltonian time evolution operation and the authority of performing a local observable measurement, and the control parameter is unknown; taking a direct product state as an initial state, constructing a quantum comb unitary operator which is alternately cascaded by S evolution windows and S+1 trainable circuit blocks; continuously evolving each evolution window into a local term product through single-step Trotterization; independently configuring a trainable time parameter for each evolution window and mapping the trainable time parameter to a positive interval through a bounded mapping; applying the quantum comb unitary operator to the initial state, measuring single-bit Pauli Z and adjacent Pauli ZZ expectation values to obtain a readout vector, obtaining quantum phase prediction probability through a classical classifier; and outputting a classification result according to the prediction probability. According to the application, quantum phase classification can be completed without knowing the Hamiltonian parameter.

Description

Technical Field

[0001] This invention relates to the field of quantum computing technology, and in particular to a method, system, terminal, and computer-readable storage medium for classifying many-body quantum phases based on shallow dynamical interleaving. Background Technology

[0002] Identifying quantum phases is a fundamental challenge in condensed matter physics and many-body physics. Existing quantum phase identification methods typically use explicit Hamiltonian control parameters as input or rely on quantum states prepared offline from known Hamiltonians. However, in many experimental platforms, control knobs do not directly map to effective Hamiltonian coefficients, and calibrating Hamiltonian parameters from measurement data is itself a costly inverse problem. When the Hamiltonian control parameters are unknown to the classifier, existing parameter-input-dependent methods fail. For symmetry-protected topological phases lacking conventional local order parameters, local measurements alone cannot directly reveal phase information.

[0003] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0004] The main objective of this invention is to provide a multi-body quantum phase classification method, system, terminal, and computer-readable storage medium based on shallow dynamical interleaving, aiming to solve the problem in the prior art that quantum phase classification depends on explicit Hamiltonian control parameters and cannot be implemented solely by dynamical access under unknown parameter conditions.

[0005] To achieve the above objectives, this invention provides a multi-body quantum phase classification method based on shallow dynamical interleaving, which includes the following steps: Obtain dynamic access permissions to the target quantum system, the dynamic access permissions including the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown; The initial state is prepared as a direct product state, and the quantum comb unitary operator of the target quantum system is constructed. The quantum comb unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. The Hamiltonian time evolution window is a single instance of performing the Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows. The continuous evolution operation of each Hamiltonian time evolution window is approximated as a product of local terms by using single-step Trotterization; Independent trainable parameters are configured for the evolution time of each Hamiltonian time evolution window, and each trainable parameter is mapped to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window. After applying the quantum comb unitary operator to the initial state, the expected value of the single-qubit Pauli Z operator of each qubit and the expected value of the Pauli ZZ operator of the adjacent qubit are measured to form a readout vector. The readout vector is then input into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. The quantum phase classification result of the target quantum system is output based on the predicted probability.

[0006] Furthermore, the acquisition of dynamic access permissions for the target quantum system includes the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown, specifically including: Establish a communication connection with the target quantum system and obtain first authority to apply a Hamiltonian time evolution operation of a specified duration to the target quantum system; Obtain second permission to perform single-bit projection measurements on each qubit of the target quantum system; While keeping the Hamiltonian control parameters of the target quantum system unknown, the first permission and the second permission are combined to form the dynamic access permission.

[0007] Furthermore, the initial state is prepared as a direct product state, and a quantum comb-unitary operator of the target quantum system is constructed. This quantum comb-unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterizable quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. Specifically, it includes: The initial state of n qubits is prepared as follows The n-fold tensor product of the states, The initial state is the n-fold tensor product of the states, where n is the total number of qubits; Construct each of the trainable parameterized quantum circuit blocks, each of the trainable parameterized quantum circuit blocks consisting of a single-bit rotation layer acting on each qubit and a controlled NOT gate loop acting between adjacent qubits; The quantum comb unitary operator is constructed in the form that the S-th trainable parameterized quantum circuit block, the S-th Hamiltonian time evolution window, the (S-1)-th trainable parameterized quantum circuit block, the (S-1)-th Hamiltonian time evolution window, ..., the 1st trainable parameterized quantum circuit block, the 1st Hamiltonian time evolution window, and the 0th trainable parameterized quantum circuit block act sequentially.

[0008] Furthermore, the method of approximating the continuous evolution operation of each Hamiltonian time evolution window as a product of local terms using single-step Trottering specifically includes: The Hamiltonian of the target quantum system is decomposed into several local terms; The Hamiltonians in the continuous evolution operators of each Hamiltonian time evolution window are replaced with the sum of each local term, and the continuous evolution operators are approximated as the product of the evolution operators corresponding to each local term.

[0009] Furthermore, the step of configuring independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and mapping each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window, specifically includes: Each Hamiltonian time evolution window is configured with an independent real-number trainable parameter; Set the lower and upper limits of the evolution time; Each of the real-number trainable parameters is mapped to a value between 0 and 1 using the sigmoid function. The mapped values ​​are then linearly transformed into the bounded positive interval based on the lower and upper limits to obtain the evolution time values ​​of each evolution window.

[0010] Furthermore, after applying the quantum comb-unitary operator to the initial state, the expected value of the single-qubit Pauli Z operator and the expected value of the adjacent qubit Pauli ZZ operator are measured to form a readout vector. The readout vector is then input into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. Specifically, this includes: The quantum comb unitary operator is applied to the initial state to obtain the evolved quantum state of the target quantum system; The Pauli Z operator expectation value of each qubit is measured for the evolved quantum state to obtain n single-bit measurement values, where n is the total number of qubits; The Pauli ZZ operator expectation values ​​of adjacent qubits are measured for the evolved quantum state to obtain n-1 adjacent two-qubit measurement values; The n single-bit measurements are concatenated with the n-1 adjacent double-bit measurements to form a readout vector with a dimension of 2n-1; The readout vector is input into the classical classifier, and after being processed by the hidden layer mapping and activation function of the classical classifier, the predicted probability of the quantum phase category to which the target quantum system belongs is output.

[0011] Furthermore, the step of outputting the quantum phase classification result of the target quantum system based on the predicted probability specifically includes: Determine whether the predicted probability is greater than or equal to a preset threshold; When the predicted probability is greater than or equal to the preset threshold, the first quantum phase category label of the target quantum system is output; When the predicted probability is less than the preset threshold, the second quantum phase category label of the target quantum system is output.

[0012] Furthermore, to achieve the above objectives, the present invention also provides a multi-body quantum phase classification system based on shallow dynamical interleaving. This system is used to implement the multi-body quantum phase classification method based on shallow dynamical interleaving as described above. The system comprises: A dynamics access module is used to acquire dynamics access permissions to a target quantum system. The dynamics access permissions include the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown. A quantum comb construction module is used to prepare the initial state into a product state and construct the quantum comb unitary operator of the target quantum system. The quantum comb unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. The Hamiltonian time evolution window is a single instance of performing the Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows. The Trotter approximation module is used to approximate the continuous evolution operation of each Hamiltonian time evolution window as a product of local terms using single-step Trottering; The time parameterization module is used to configure independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and to map each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window. The readout classification module is used to apply the quantum comb unitary operator to the initial state, measure the expected value of the single-qubit Pauli Z operator of each qubit and the expected value of the Pauli ZZ operator of adjacent qubits, form a readout vector, and input the readout vector into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. The classification output module is used to output the quantum phase classification result of the target quantum system based on the predicted probability.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a multi-body quantum phase classification program based on shallow dynamics interleaving stored in the memory and executable on the processor, wherein when the multi-body quantum phase classification program based on shallow dynamics interleaving is executed by the processor, it implements the steps of the multi-body quantum phase classification method based on shallow dynamics interleaving as described above.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a multi-body quantum phase classification program based on shallow dynamical interleaving, and when the multi-body quantum phase classification program based on shallow dynamical interleaving is executed by a processor, it implements the steps of the multi-body quantum phase classification method based on shallow dynamical interleaving as described above.

[0015] The beneficial effects of this invention are as follows: By acquiring dynamic access capabilities that include only time evolution operations and local measurement permissions without exposing Hamiltonian control parameters, phase discrimination of quantum systems with unknown parameters can be implemented without relying on parameter calibration; a quantum comb-unitary operator is constructed using direct product states as input, which alternately cascades time evolution windows and trainable parameterized quantum circuit blocks, injecting the dynamic characteristics of the target system into the quantum information processing flow multiple times; single-step Trottering is used to decompose the continuous evolution of each evolution window into a product of local terms, enabling the evolution operations to be implemented on quantum hardware with an executable gate sequence; trainable time parameters are independently configured for each evolution window and constrained to the positive interval through bounded nonlinear mapping, allowing dynamic information at multiple time scales to be sampled; the local single-bit and nearest-neighbor two-bit measurements after the action of the quantum comb-unitary operator constitute a readout vector and are mapped to phase prediction probabilities by a classical classifier; finally, the quantum phase classification result is output based on the prediction probability, thereby completing the determination of the quantum phase category of the target quantum system without knowing the Hamiltonian parameters. Attached Figure Description

[0016] Figure 1 This is a flowchart of a preferred embodiment of the multibody quantum phase classification method based on shallow dynamics interleaving of the present invention; Figure 2 This is a schematic diagram of the overall architecture of the quantum comb classifier of the present invention; Figure 3 This is a phase diagram reconstruction effect of the quantum comb classifier of the present invention under three benchmark Hamiltonian families and four evolution slot numbers; Figure 4 This is a structural diagram of a preferred embodiment of the present invention based on a shallow dynamically interlaced many-body quantum phase classification system; Figure 5 This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0017] This application provides a method, system, terminal, and storage medium for classifying many-body quantum phases based on shallow dynamical interleaving. To make the objectives, technical solutions, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0018] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0019] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0020] The preferred embodiment of the present invention describes a many-body quantum phase classification method based on shallow dynamical interleaving, such as... Figure 1 As shown, the many-body quantum phase classification method based on shallow dynamical interleaving includes the following steps: S10. Obtain dynamic access permissions to the target quantum system, wherein the dynamic access permissions include the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown.

[0021] The purpose of this step is to provide an operational basis for all subsequent steps involving quantum system manipulation and readout in this invention, ensuring that the classifier can perform time evolution and measurement operations on the target quantum system without knowing the specific parameter values ​​of the Hamiltonian.

[0022] Furthermore, the acquisition of dynamic access permissions for the target quantum system includes the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown, specifically including: S11. Establish a communication connection with the target quantum system and obtain the first authority to apply a Hamiltonian time evolution operation of a specified duration to the target quantum system. S12. Obtain second permission to perform single-bit projection measurements on each qubit of the target quantum system; S13. While keeping the Hamiltonian control parameters of the target quantum system unknown, combine the first permission and the second permission to form the dynamic access permission.

[0023] In this embodiment, the target quantum system is an n-qubit quantum system controlled by an unknown Hamiltonian, such as a quantum simulator with an unknown composition. Its Hamiltonian belongs to a known family of Hamiltonians (e.g., the Cluster-Ising model family, the disordered transverse field Ising model family, or the XXZ model family), but the specific control parameter values ​​are not visible to the classifier. The system first establishes a communication connection with the target quantum system and acquires first and second permissions. The first permission allows the classifier to apply a Hamiltonian time evolution operation of a specified duration to the target quantum system, and the second permission allows the classifier to perform single-qubit projection measurements on each qubit. Throughout the acquisition of these permissions, the classifier remains inaccessible to the Hamiltonian control parameters, only gaining the ability to manipulate and measure the system, but unable to directly read any parameter values. Combining the first and second permissions constitutes a complete dynamic access permission, which can be invoked in subsequent steps.

[0024] It should be noted that the Hamiltonian control parameters of the target quantum system are unknown. This is a key difference between this invention and existing parameter-aware methods (such as quantum convolutional neural networks (QCNN), quantum kernel methods, and circuit-centric quantum classifiers). Existing methods typically use explicit Hamiltonian control parameters as input or rely on states prepared offline from known Hamiltonians for classification. In contrast, this invention relies solely on access to the system's dynamics and can complete the phase classification task without knowing any parameter values.

[0025] S20. Prepare the initial state as a direct product state and construct the quantum comb unitary operator of the target quantum system. The quantum comb unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. The Hamiltonian time evolution window is a single instance of performing the Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows.

[0026] The purpose of this step is to construct a unitary operator structure that can inject the dynamic characteristics of the target quantum system into the quantum information processing flow multiple times. By alternately cascading trainable parameterized quantum circuit blocks and Hamiltonian evolution windows, the characteristic information of the many-body quantum phase can be effectively extracted in the subsequent readout process.

[0027] Furthermore, the initial state is prepared as a direct product state to construct the quantum comb-unitary operator of the target quantum system. The quantum comb-unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterizable quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks, specifically including: S21. Prepare the initial state of n qubits as follows: The n-fold tensor product of the states, The initial state is the n-fold tensor product of the states, where n is the total number of qubits; S22. Construct each of the trainable parameterized quantum circuit blocks, each of the trainable parameterized quantum circuit blocks consisting of a single-bit rotation layer acting on each quantum bit and a controlled NOT gate loop acting between adjacent quantum bits; S23. The quantum comb unitary operator is constructed in the form that the Sth trainable parameterized quantum circuit block, the Sth Hamiltonian time evolution window, the (S-1)th trainable parameterized quantum circuit block, the (S-1)th Hamiltonian time evolution window, ..., the 1st trainable parameterized quantum circuit block, the 1st Hamiltonian time evolution window, and the 0th trainable parameterized quantum circuit block act in sequence.

[0028] In this embodiment, the system first prepares the initial state of the n qubits as follows: The n-fold tensor product of the states is used as the initial state for all subsequent operations. Subsequently, the system constructs a quantum comb unitary operator, which is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks (PQC blocks) cascaded in an alternating order of evolution windows and circuit blocks. Each Hamiltonian time evolution window is a single instance of performing a single Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows, and the s-th evolution window corresponds to the s-th evolution time.

[0029] Each trainable parameterized quantum circuit block consists of a single-bit rotation gate acting on all n working bits, followed by a controlled NOT gate loop. The control bit of the controlled NOT gate loop is the i-th qubit, and the target bit is the (i+1)-th qubit (modulo n). The parameter set of each trainable parameterized quantum circuit block contains n rotation angles, which are adjusted during training using an optimization algorithm. The entire quantum comb-unitary operator operates in the following order: the 0th trainable parameterized quantum circuit block acts on the initial state first, followed by the 1st Hamiltonian time evolution window, the 1st trainable parameterized quantum circuit block, ..., the S-th Hamiltonian time evolution window, and the S-th trainable parameterized quantum circuit block.

[0030] The expression for the quantum comb unitary operator is: ; in, This represents the quantum comb unitary operator. The Hamiltonian of the target quantum system is represented. This represents the set of trainable parameters for each trainable parameterized quantum circuit block. This represents the set of trainable durations for each Hamiltonian temporal evolution window. Indicates the first A trainable parameterizable quantum circuit block ( =0,1,…,S), Indicates the first Continuous evolution operators of a Hamiltonian time evolution window ( =1,…,S), It represents the imaginary unit.

[0031] In this embodiment, each trainable parameterized quantum circuit block contains only one layer of single-bit rotating gates and one layer of two-bit controlled NOT gates. The depth of the entire circuit is independent of the system size n, maintaining a shallow structure.

[0032] Figure 2 This is a schematic diagram of the overall architecture of the quantum comb classifier of this invention. Figure 2As shown, the architecture is divided into four parts from left to right: the first part is the black-box target access unit, which represents the target Hamiltonian that the physical system interface interfaces with. The control parameters of this Hamiltonian are not visible to the classifier. The second part is the interleaved mapping circuit unit, which represents the quantum comb structure formed by cascading variational gates and evolution windows in an alternating order. The evolution time of each evolution window is constrained by bounded time, that is, the evolution time is equal to the sum of the product of the lower limit value and the time span with the sigmoid function value. The third part is the local dimensionality reduction measurement unit, which represents the measurement of the single-qubit Pauli Z operator expectation value and the nearest neighbor Pauli ZZ operator expectation value of the quantum state after the quantum comb evolution. The fourth part is the classical decoding and decision unit, which represents the multilayer perceptron outputting the phase classification probability and the corresponding confidence level based on the readout vector obtained by the measurement. Figure 2 This invention demonstrates the complete information processing chain from black-box Hamiltonian access, through quantum comb alternating evolution and local measurement, to classical classification decision.

[0033] It should be noted that each Hamiltonian time evolution window in the quantum comb unitary operator is an evolution operation performed on the same unknown Hamiltonian H, but the duration of each evolution window is independent of each other. This structure is fundamentally different from existing data re-upload classifiers (each slot is filled with classical data) and parameterized quantum combs (each slot is filled with learned process tensors): in this invention, each slot is filled by the same unknown Hamiltonian channel in trainable, independent evolution times, and classification relies only on local readouts without any parameter access.

[0034] S30. The continuous evolution operation of each Hamiltonian time evolution window is approximated as a product of local terms by using single-step Trotterization.

[0035] The purpose of this step is to approximate the continuous evolution operators in each Hamiltonian time evolution window that are difficult to implement directly on quantum hardware by Trottering into the product of the evolution operators corresponding to the local terms, so that each evolution window can be implemented on quantum hardware as an executable sequence of quantum gates.

[0036] Furthermore, the step of using single-step Trottering to approximate the continuous evolution operation of each Hamiltonian time evolution window as a product of local terms specifically includes: S31. Decompose the Hamiltonian of the target quantum system into several local terms; S32. Replace the Hamiltonian in the continuous evolution operator of each Hamiltonian time evolution window with the sum of each local term, and approximate the continuous evolution operator as the product of the evolution operators corresponding to each local term.

[0037] In this embodiment, the system first converts the Hamiltonian of the target quantum system... Decomposed into several local terms The sum of Since Hamiltonians typically contain nearest-neighbor interaction terms (such as ZZ coupling, XX coupling, YY coupling, etc.), these terms are all local terms and can be directly obtained from the Pauli operator expansion of the Hamiltonian.

[0038] Subsequently, regarding the first The system uses a continuous evolution operator with a Hamiltonian time evolution window. The Hamiltonian in the continuous evolution operator is replaced by the sum of each local term, and the continuous evolution operator is approximated as the product of the evolution operators corresponding to each local term. ; in, Indicates the first The evolutionary time of each evolutionary window Indicates the first Local item, Indicates the first The evolution operators corresponding to each local term, This indicates that the evolution operators corresponding to all local terms are multiplied sequentially.

[0039] When the local terms commutate with each other, the above approximation becomes an exact equation; when the local terms do not commutate with each other, the above approximation is a first-order approximation error introduced by single-step Trotterization.

[0040] It should be noted that the single-step Trottering decomposes the many-body coupling terms contained within a single Hamiltonian time evolution window into several two-qubit or single-qubit gate sequences that can be directly executed on quantum hardware, thus lowering the implementation threshold of the quantum comb-unitary operator on physical hardware. The Trotter approximation error introduced when the local terms do not commute can be defined through an end-to-end reliability certificate, ensuring that this approximation error does not change the final prediction label.

[0041] S40. Configure independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and map each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window.

[0042] The purpose of this step is to introduce an independently optimizable evolution time parameter for each Hamiltonian time evolution window, and to constrain these parameters to a physically reasonable positive interval through bounded nonlinear mapping, so that each evolution window can sample the dynamic characteristics of the target Hamiltonian with different durations, thereby capturing richer spectral information.

[0043] Furthermore, the step of configuring independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and mapping each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window, specifically includes: S41. Configure an independent real-number trainable parameter for each of the Hamiltonian time evolution windows; S42. Set the lower and upper limits of the evolution time; S43. Map each of the real-number trainable parameters to a value between 0 and 1 using the sigmoid function, and linearly transform the mapped values ​​to the bounded positive interval according to the lower limit and the upper limit to obtain the evolution time value of each evolution window.

[0044] In this embodiment, the system configures an independent real-number trainable parameter for each Hamiltonian time evolution window. ( =1,…,S). These real-valued trainable parameters can be randomly sampled during the initialization phase and updated through optimization algorithms during the training phase. To prevent excessively large evolution time values ​​from causing decoherence or excessively small values ​​from causing insufficient dynamic information, the system sets a lower limit for the evolution time. and upper limit value (Typical heuristic values ​​are a lower limit of 0.02 and an upper limit of 0.5), and each real-valued trainable parameter is passed through the sigmoid function. Mapped to values ​​between 0 and 1, where The input is the sigmoid function; then, based on the lower and upper limits, the mapped value is linearly transformed to a bounded positive interval. , Within this range, the evolution time values ​​for each evolution window are obtained: ; in, Indicates the first The evolution time value of each evolution window.

[0045] It should be noted that, through the aforementioned bounded nonlinear mapping, the evolution time of each evolution window is independently constrained within a physically reasonable range, preventing the evolution time from collapsing into a trivial constant solution or diverging to physically meaningless values. The independently tuned evolution time, allocating different durations to each evolution window, provides a richer frequency dictionary compared to a uniform duration scheduling strategy for all evolution windows. This allows the quantum comb-unitary operator to capture the dynamic characteristics of the target Hamiltonian at different time scales.

[0046] S50. After applying the quantum comb unitary operator to the initial state, measure the expected value of the single-qubit Pauli Z operator of each qubit and the expected value of the Pauli ZZ operator of adjacent qubits to form a readout vector. Input the readout vector into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs.

[0047] The purpose of this step is to encode the multi-body phase characteristic information of the target quantum system into the local observable of the quantum state using a quantum comb unitary operator, then extract this information using single-bit and nearest-neighbor two-bit measurements, and finally map it into the predicted probability of the phase class by a lightweight classical classifier.

[0048] Further, after applying the quantum comb-unitary operator to the initial state, the expected value of the single-qubit Pauli Z operator and the expected value of the adjacent qubit Pauli ZZ operator are measured to form a readout vector. The readout vector is then input into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. Specifically, this includes: S51. Apply the quantum comb unitary operator to the initial state to obtain the evolved quantum state of the target quantum system; S52. Measure the Pauli Z operator expectation value of each qubit in the evolved quantum state to obtain n single-bit measurement values, where n is the total number of qubits; S53. Measure the Pauli ZZ operator expectation values ​​of adjacent qubits of the evolved quantum state respectively to obtain n-1 adjacent two-qubit measurement values; S54. Concatenate the n single-bit measurement values ​​with the n-1 adjacent double-bit measurement values ​​to form a readout vector with a dimension of 2n-1; S55. Input the readout vector into the classical classifier, and after the hidden layer mapping and activation function processing of the classical classifier, output the predicted probability of the quantum phase category to which the target quantum system belongs.

[0049] In this embodiment, the system first applies the quantum comb-unitary operator constructed in step S20 to the initial state prepared in step S21 to obtain the evolved quantum state. This evolved quantum state encodes the dynamic characteristics of the target quantum system at multiple time scales.

[0050] Subsequently, the system measures the Pauli Z operator expectation value of each qubit in the evolved quantum state. ( This process yields n single-bit measurements, and the Pauli ZZ operator expectation values ​​for adjacent qubits of the evolved quantum state are measured respectively. ( -1), resulting in n-1 adjacent double-bit measurements. The n single-bit measurements are concatenated with the n-1 adjacent double-bit measurements to form a readout vector x of dimension 2n-1. H ): ; in, This represents the readout vector. The Hamiltonian of the target quantum system is represented. Indicates the first The expectation value of the Pauli Z operator for qubits ( ), Indicates the first The ( ) and the ( The expectation value of the Pauli ZZ operator with 100 qubits ( -1), where n is the total number of qubits.

[0051] The readout vector described above has a dimension of 2n-1, which increases linearly with the total number of qubits n, avoiding the exponential measurement overhead required for full quantum state tomography. These observables can be obtained through standard single-qubit readouts on all major quantum computing platforms without the need for auxiliary qubits.

[0052] The system inputs the readout vector into a classical classifier, which is a single-hidden-layer multilayer perceptron (MLP). Its structure is as follows: the input layer has a dimension of 2n-1, the hidden layer has a dimension of 16, and uses the Rectified Linear Unit (ReLU) activation function. The output layer has a dimension of 1 and uses the sigmoid activation function. After processing by the hidden layer mapping and activation function of the classical classifier, the system outputs the predicted probability of the quantum phase category to which the target quantum system belongs. .

[0053] From the predicted probability Reliability score can be defined A reliability score close to 1 indicates that the classifier has high confidence in the prediction results, while a score close to 0 indicates that the current parameter point is near the decision threshold, i.e., near the critical region of phase transition.

[0054] It should be noted that the dimension of the readout vector is only 2n-1, and it only contains the expectation values ​​of the single-bit Pauli Z operator and the nearest-neighbor Pauli ZZ operator. These measurements are all local measurements and do not involve nonlocal string order operators or auxiliary bit-assisted measurements. However, thanks to the Heisenberg conjugation effect generated by the alternating basis transformation and Hamiltonian evolution window in the quantum comb unitary operator, the global characteristic information of the many-body quantum phase has been conjugated and transferred to the aforementioned local readout channel, so that local measurements can capture the phase discrimination information originally encoded in the nonlocal string order or entangled structure.

[0055] S60. Output the quantum phase classification result of the target quantum system based on the predicted probability.

[0056] The purpose of this step is to convert the predicted probabilities output by the classical classifier into discrete quantum phase category labels, thereby completing the final determination of the quantum phase category to which the target quantum system belongs.

[0057] Furthermore, the step of outputting the quantum phase classification result of the target quantum system based on the predicted probability specifically includes: S61. Determine whether the predicted probability is greater than or equal to a preset threshold. S62. When the predicted probability is greater than or equal to the preset threshold, output the first quantum phase category label of the target quantum system; S63. When the predicted probability is less than the preset threshold, output the second quantum phase category label of the target quantum system.

[0058] In this embodiment, the system acquires a preset threshold (typically 0.5) and determines whether the predicted probability is greater than or equal to the preset threshold. When the predicted probability is greater than or equal to the preset threshold, the system determines that the target quantum system belongs to a first quantum phase category and outputs the first quantum phase category label y=1; when the predicted probability is less than the preset threshold, the system determines that the target quantum system belongs to a second quantum phase category and outputs the second quantum phase category label y=0.

[0059] In this embodiment, the Hamiltonian control parameters of the target quantum system are unknown. During the training phase, this invention only needs to adjust the parameters of the trainable parameterized quantum circuit blocks and the real trainable parameters of each evolution window at a few representative parameter points through optimization algorithms, so that the output of the quantum comb unitary operator, after readout and classification, is consistent with the true phase label. After training convergence, for any instance of the target quantum system with unknown parameter configuration, the system only needs to perform a single forward propagation (including quantum comb unitary operator action, local measurement, and classical classifier inference) to output the quantum phase category label, without retraining or resolving the Hamiltonian parameters.

[0060] Figure 3 This is a phase diagram reconstruction result of the quantum comb classifier of this invention under three benchmark Hamiltonian families and four evolution slot numbers. (See diagram below.) Figure 3As shown, the figure consists of twelve subgraphs in four columns and three rows. The columns correspond to four settings with 0, 1, 3, and 5 evolution slots, respectively. The rows, from top to bottom, correspond to three Hamiltonian families: the Cluster-Ising model, the disordered transverse field Ising model, and the XXZ chain. The color of each subgraph represents the predicted phase probability output by the quantum comb classifier. The color bar is located on the right side of the figure, with blue through white to red indicating probabilities from 0 to 1. In the legend at the bottom, the solid black line represents the true phase transition boundary, and the dashed yellow line represents the equally probable contour predicted by the classifier. The grid accuracy and average reliability are marked in the lower right corner of each subgraph. When the number of evolution slots is 0, the predicted phase diagrams of the disordered transverse field Ising model and the XXZ chain are close to random and have almost zero reliability. After adding evolution slots, the phase transition structure is significantly sharpened. The effect tends to saturate near the number of evolution slots of 3, and provides further gains on some benchmarks when the number of evolution slots is 5, indicating that phase recognition is driven by Hamiltonian dynamics rather than static initial state features.

[0061] It should be noted that this invention can assess the confidence level of the classification results using a reliability score. When the reliability score is close to 0, it indicates that the target quantum system is near the phase transition critical point, which can be used to drive the device to perform intensive sampling in this parameter region to achieve automated phase diagram exploration. Furthermore, this invention also includes an end-to-end reliability certificate: if the reliability score is greater than the end-to-end stability threshold... The value of the hyperbolic tangent function, i.e. The error introduced by the Trotter approximation will not change the predicted label, where The end-to-end stability threshold is determined by both the Trotter error bound and the Lipschitz constant of the classical classifier. This is a vector of evolution time values ​​for each evolution window. This certificate allows the model to automatically decide whether to accept predictions based on confidence levels during deployment.

[0062] Furthermore, such as Figure 4 As shown, based on the above-mentioned multi-body quantum phase classification method based on shallow dynamical interleaving, the present invention also provides a multi-body quantum phase classification system based on shallow dynamical interleaving, the multi-body quantum phase classification system based on shallow dynamical interleaving comprising: The dynamics access module 51 is used to obtain dynamics access permissions for the target quantum system. The dynamics access permissions include the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown. The quantum comb construction module 52 is used to prepare the initial state into a direct product state and construct the quantum comb unitary operator of the target quantum system. The quantum comb unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. The Hamiltonian time evolution window is a single instance of performing the Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows. Trotter approximation module 53 is used to approximate the continuous evolution operation of each Hamiltonian time evolution window as a product of local terms using single-step Trotterization; The time parameterization module 54 is used to configure independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and to map each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window. The readout classification module 55 is used to apply the quantum comb unitary operator to the initial state, measure the single-qubit Pauli Z operator expectation value of each qubit and the Pauli ZZ operator expectation value of adjacent qubits, form a readout vector, and input the readout vector into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. The classification output module 56 is used to output the quantum phase classification result of the target quantum system according to the predicted probability.

[0063] Furthermore, such as Figure 5 As shown, based on the above-mentioned multibody quantum phase classification method and system based on shallow dynamics interleaving, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 5 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0064] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a multi-body quantum phase classification program 40 based on shallow dynamical interleaving, which can be executed by the processor 10 to implement the multi-body quantum phase classification method based on shallow dynamical interleaving in this application.

[0065] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the multibody quantum phase classification method based on shallow dynamical interleaving.

[0066] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components of the terminal communicate with each other via a system bus.

[0067] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a multi-body quantum phase classification program based on shallow dynamics interleaving, and the multi-body quantum phase classification program based on shallow dynamics interleaving, when executed by a processor, implements the steps of the multi-body quantum phase classification method based on shallow dynamics interleaving as described above.

[0068] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0069] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0070] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A multi-body quantum phase classification method based on shallow dynamical interleaving, characterized in that, The multi-body quantum phase classification method based on shallow dynamical interleaving includes the following steps: Obtain dynamic access permissions to the target quantum system, the dynamic access permissions including the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown; The initial state is prepared as a direct product state, and the quantum comb unitary operator of the target quantum system is constructed. The quantum comb unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. The Hamiltonian time evolution window is a single instance of performing the Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows. The continuous evolution operation of each Hamiltonian time evolution window is approximated as a product of local terms by using single-step Trotterization; Independent trainable parameters are configured for the evolution time of each Hamiltonian time evolution window, and each trainable parameter is mapped to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window. After applying the quantum comb unitary operator to the initial state, the expected value of the single-qubit Pauli Z operator of each qubit and the expected value of the Pauli ZZ operator of the adjacent qubit are measured to form a readout vector. The readout vector is then input into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. The quantum phase classification result of the target quantum system is output based on the predicted probability.

2. The many-body quantum phase classification method based on shallow dynamical interleaving according to claim 1, characterized in that, The acquisition of dynamic access permissions for the target quantum system includes the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown, specifically including: Establish a communication connection with the target quantum system and obtain first authority to apply a Hamiltonian time evolution operation of a specified duration to the target quantum system; Obtain second permission to perform single-bit projection measurements on each qubit of the target quantum system; While keeping the Hamiltonian control parameters of the target quantum system unknown, the first permission and the second permission are combined to form the dynamic access permission.

3. The many-body quantum phase classification method based on shallow dynamical interleaving according to claim 1, characterized in that, The initial state is prepared as a direct product state, and a quantum comb-unitary operator of the target quantum system is constructed. The quantum comb-unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterizable quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks, specifically including: The initial state of n qubits is prepared as follows The n-fold tensor product of the states, The initial state is the n-fold tensor product of the states, where n is the total number of qubits; Construct each of the trainable parameterized quantum circuit blocks, each of the trainable parameterized quantum circuit blocks consisting of a single-bit rotation layer acting on each qubit and a controlled NOT gate loop acting between adjacent qubits; The quantum comb unitary operator is constructed in the form that the S-th trainable parameterized quantum circuit block, the S-th Hamiltonian time evolution window, the (S-1)-th trainable parameterized quantum circuit block, the (S-1)-th Hamiltonian time evolution window, ..., the 1st trainable parameterized quantum circuit block, the 1st Hamiltonian time evolution window, and the 0th trainable parameterized quantum circuit block act sequentially.

4. The many-body quantum phase classification method based on shallow dynamical interleaving according to claim 1, characterized in that, The method of approximating the continuous evolution operation of each Hamiltonian time evolution window as a product of local terms using single-step Trotterization specifically includes: The Hamiltonian of the target quantum system is decomposed into several local terms; The Hamiltonians in the continuous evolution operators of each Hamiltonian time evolution window are replaced with the sum of each local term, and the continuous evolution operators are approximated as the product of the evolution operators corresponding to each local term.

5. The many-body quantum phase classification method based on shallow dynamical interleaving according to claim 1, characterized in that, The process of configuring independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and mapping each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window, specifically includes: Each Hamiltonian time evolution window is configured with an independent real-number trainable parameter; Set the lower and upper limits of the evolution time; Each of the real-number trainable parameters is mapped to a value between 0 and 1 using the sigmoid function. The mapped values ​​are then linearly transformed into the bounded positive interval based on the lower and upper limits to obtain the evolution time values ​​of each evolution window.

6. The many-body quantum phase classification method based on shallow dynamical interleaving according to claim 1, characterized in that, After applying the quantum comb-unitary operator to the initial state, the expected value of the single-qubit Pauli Z operator and the expected value of the Pauli ZZ operator of adjacent qubits are measured to form a readout vector. The readout vector is then input into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. Specifically, this includes: The quantum comb unitary operator is applied to the initial state to obtain the evolved quantum state of the target quantum system; The Pauli Z operator expectation value of each qubit is measured for the evolved quantum state to obtain n single-bit measurement values, where n is the total number of qubits; The Pauli ZZ operator expectation values ​​of adjacent qubits are measured for the evolved quantum state to obtain n-1 adjacent two-qubit measurement values; The n single-bit measurements are concatenated with the n-1 adjacent double-bit measurements to form a readout vector with a dimension of 2n-1; The readout vector is input into the classical classifier, and after being processed by the hidden layer mapping and activation function of the classical classifier, the predicted probability of the quantum phase category to which the target quantum system belongs is output.

7. The many-body quantum phase classification method based on shallow dynamical interleaving according to claim 1, characterized in that, The step of outputting the quantum phase classification result of the target quantum system based on the predicted probability specifically includes: Determine whether the predicted probability is greater than or equal to a preset threshold; When the predicted probability is greater than or equal to the preset threshold, the first quantum phase category label of the target quantum system is output; When the predicted probability is less than the preset threshold, the second quantum phase category label of the target quantum system is output.

8. A many-body quantum phase classification system based on shallow dynamical interleaving, characterized in that, The shallow dynamical interleaving-based multi-body quantum phase classification system is used to implement the shallow dynamical interleaving-based multi-body quantum phase classification method as described in any one of claims 1-7, wherein the shallow dynamical interleaving-based multi-body quantum phase classification system comprises: A dynamics access module is used to acquire dynamics access permissions to a target quantum system. The dynamics access permissions include the permission to apply Hamiltonian time evolution operations to the target quantum system and the permission to perform local observable measurements on the target quantum system, wherein the Hamiltonian control parameters of the target quantum system are unknown. A quantum comb construction module is used to prepare the initial state into a product state and construct the quantum comb unitary operator of the target quantum system. The quantum comb unitary operator is composed of S Hamiltonian time evolution windows and S+1 trainable parameterized quantum circuit blocks cascaded in an alternating order of evolution windows and circuit blocks. The Hamiltonian time evolution window is a single instance of performing the Hamiltonian time evolution operation on the target quantum system, where S is the preset number of evolution windows. The Trotter approximation module is used to approximate the continuous evolution operation of each Hamiltonian time evolution window as a product of local terms using single-step Trottering; The time parameterization module is used to configure independent trainable parameters for the evolution time of each Hamiltonian time evolution window, and to map each trainable parameter to a bounded positive interval through a bounded nonlinear function to obtain the evolution time value of each evolution window. The readout classification module is used to measure the expected value of the single-qubit Pauli Z operator and the expected value of the adjacent qubit Pauli ZZ operator after the quantum comb unitary operator is applied to the initial state, to form a readout vector, and input the readout vector into a classical classifier to obtain the predicted probability of the quantum phase category to which the target quantum system belongs. The classification output module is used to output the quantum phase classification result of the target quantum system based on the predicted probability.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and a multi-body quantum phase classification program based on shallow dynamics interleaving stored in the memory and executable on the processor. When the multi-body quantum phase classification program based on shallow dynamics interleaving is executed by the processor, it implements the steps of the multi-body quantum phase classification method based on shallow dynamics interleaving as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a multi-body quantum phase classification program based on shallow dynamical interleaving, which, when executed by a processor, implements the steps of the multi-body quantum phase classification method based on shallow dynamical interleaving as described in any one of claims 1-7.

Images