Quantum machine learning using genetic algorithms

HK40137306APending Publication Date: 2026-09-11STANDARD CHARTERED BANK SINGAPORE BRANCH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
HK42026126310
Authority / Receiving Office
HK · HK
Patent Type
Applications
Current Assignee / Owner
Priority Date
2024-02-07
Filing Date
2026-07-17
Publication Date
2026-09-11

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

A method for determining a quantum neural network (QNN) configuration for classification in a problem domain by generating a population of candidate QNN configurations based on a number of feature val
Need to check novelty before this filing date? Find Prior Art

Description

(12) UK Patent Application (19) GB (11) 2700344 (13) A (43) Date of A Publication 21.01.2026 (21) Application No: GB2501736.9 (22) Date of Filing: 06.02.2025 (30) Priority Data: (31) 202400348 (32) 07.02.2024 (33) SG (71) Applicant(s): Standard Chartered Bank, Singapore Branch 8 Marina Boulevard, #27-01 Marina Bay Financial Centre, Singapore, Singapore, 018981 (72) Inventor(s): Yongxin Steve Cao, Elena Strbac (74) Agent and / or Address for Service: Marks & Clerk LLP 40 Torphichen Street, Edinburgh, EH3 8JB (51) INT CL: G06N3 / 126 (2023.01) G06N10 / 60 (2022.01) (56) Documents Cited: None (58) Field of Search: INT CL: Other: No search performed 17(5)(b) (54) Title of the Invention: Quantum machine learning using genetic algorithms Abstract Title: Classification using a quantum genetic algorithm (57) A method for determining a quantum neural network (QNN) configuration for classification in a problem domain by generating a population of candidate QNN configurations based on a number of feature values or parameters and performing a genetic search in the population of the configurations. The genetic search uses a simulation of iteratively evolving the population based on fitness metrics to identify an individual in the population, i.e. a QNN configuration, with the highest fitness metric. Multi-layered hierarchical entanglement relationship between qubits may be used with two qubit groups at each layer, wherein the number of qubits in the two qubit groups may be determined based on a tree split ratio.In this way, a classification algorithm for a specific domain may be provided (e.g., classification between classes A and B). Fig 1. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Abstract This paper proposes a method for determining the configuration of a quantum neural network (QNN) for problem domain classification. This method first generates a population of candidate QNN configurations based on several eigenvalues ​​or parameters, and then performs a genetic search within this population. The genetic search identifies the individual with the highest fitness in the population, i.e., the QNN configuration, by simulating the iterative evolution of the population based on a fitness index. This method can utilize multi-level qubit entanglement relationships, with each level containing two qubit groups, where the number of qubits in the two qubit groups can be determined based on the tree split ratio. In this way, a classification algorithm for a specific domain can be provided (e.g., classification between classes A and B). Fig 1.

Claims

1. A method for determining a target Quantum Neural Network (QNN) configuration for binary classification, the method comprising:receiving a binary classification dataset, the dataset comprising a plurality of feature data and target classification data;generating a population of candidate QNN configurations based on the plurality of features, each candidate QNN configuration defining a QNN comprising a random mapping of each of the plurality of features to at least one qubit of the QNN and entanglement relationships between the qubits;performing a genetic search in the population of QNN configurations for a target QNN configuration by:simulated execution of candidate QNNs initialized using the each QNN configuration in the population of QNN configurations;computing a fitness metric for each of the candidate QNNs based on outputs of simulated execution of the candidate QNNs and the target classification data; and iteratively evolving the population of QNN configurations based on the fitness metrics to identify a target QNN configuration with the highest fitness metric.

2. The method of claim 1, wherein iterative evolution of the population of QNN configurations comprises:selection of a subset of the QNN configurations based on the computed fitness metrics; generation of a subsequent population of QNN configurations by random mutation operations on the subset of the QNN configurations.

3. The method of any one of the preceding claims, wherein each QNN configuration comprises at least one configuration value defining operations performed on qubits in each layer of the QNN.

4. The method of claim 1, wherein the entanglement relationships between the qubits of the QNNs defined by the QNN configurations comprise multi-layered hierarchical entanglement relationships.

5. The method of claim 4, wherein the multi-layered hierarchical entanglement relationships comprise two qubit groups at each layer, wherein number of qubits in the two qubit groups at each layer are determined based on a tree split ratio.

6. The method of claim 1, wherein the method further comprises:generating at least one guiding feature based on one or more of the plurality of features;andeach candidate QNN configuration comprising a mapping of each guiding feature to at least one qubit as defined in the QNN configuration.

7. The method of any one of the preceding claims, wherein the method further comprises random sampling of records in the domain dataset; andsimulated execution of candidate QNNs using the randomly sampled records.

8. The method of any one of the preceding claims, wherein the method further comprises defining a QNN based on the identified target QNN configuration; and deploying the QNN based on the identified target QNN configuration to process live data and generate binary classification inferences.

9. The method of any one of the preceding claims, wherein the number of qubits in each candidate QNN configuration is less than the total number of features in the dataset; and at least a subset of qubits in each QNN configuration encode more than one feature data.

10. The method of claim 2, wherein mutation comprises introducing a random rotation to a subset of qubits of the QNNs defined by the subset of the QNN configurations.

11. A method for determining a target Quantum Neural Network (QNN) configuration for a problem domain, the method comprising:receiving a problem domain dataset, the dataset comprising a plurality of feature data and target variable data;generating a population of candidate QNN configurations based on the plurality of features, each candidate QNN configuration defining a QNN comprising a random mapping of each of the plurality of features to at least one qubit of the QNN and entanglement relationships between the qubits;performing a genetic search in the population of QNN configurations for a target QNN configuration by:simulated execution of candidate QNNs initialized using the each QNN configuration in the population of QNN configurations;computing a fitness metric for each of the candidate QNNs based on outputs of simulated execution of the candidate QNNs and the target variable data; and iteratively evolving the population of QNN configurations based on the fitness metrics to identify a target QNN configuration with the highest fitness metric.

12. The method of claim 11, wherein iterative evolution of the population of QNN configurations comprises:selection of a subset of the QNN configurations based on the computed fitness metrics; generation of a subsequent population of QNN configurations by random mutation operations on the subset of the QNN configurations.

13. The method of claim 11 or claim 12, wherein each QNN configuration comprises at least one configuration value defining operations performed on qubits in each layer of the QNN.

14. The method of claim 11, wherein the entanglement relationships between the qubits of the QNNs defined by the QNN configurations comprise multi-layered hierarchical entanglement relationships.

15. The method of claim 14, wherein the multi-layered hierarchical entanglement relationships comprise two qubit groups at each layer, wherein the number of qubits in the two qubit groups at each layer are determined based on a tree split ratio.

16. The method of claim 12, wherein the method further comprises:generating at least one guiding feature based on one or more of the plurality of features;andeach candidate QNN configuration comprising a mapping of each guiding feature to at least one qubit as defined in the QNN configuration.

17. The method of any one of claims 11 to 16, wherein the method further comprises random sampling of records in the domain dataset; andsimulated execution of candidate QNNs used the randomly sampled records.

18. The method of any one of claims 11 to 17, wherein the method further comprises defining a QNN based on the identified target QNN configuration; anddeploying the QNN based on the identified target QNN configuration to process live data and generate inferences.

19. The method of any one of claims 11 to 18, wherein the number of qubits in each candidate QNN configuration is less than the total number of features in the dataset; and at least a subset5 of qubits in each QNN configuration encode more than one feature data.

20. The method of claim 12, wherein mutation comprises introducing a random rotation to a subset of qubits of the QNNs defined by the subset of the QNN configurations.

21. A quantum computing system, the system comprising:a quantum circuit generated according to a target QNN configuration, wherein the target10 QNN configuration is determined by execution of the method of any one of claims 1 to 20.