Search

CN-121998117-A - Quantum architecture searching method and device based on noise perception evaluation

CN121998117ACN 121998117 ACN121998117 ACN 121998117ACN-121998117-A

Abstract

The invention discloses a quantum architecture searching method and device based on noise perception evaluation, the method comprises the steps of obtaining an operation pool formed by all allowable quantum gate operations according to topological connection and a native gate set of target quantum hardware, initializing an architecture parameter matrix, carrying out micro-sampling on the architecture parameter matrix by adopting a re-parameterization technology to obtain candidate quantum circuits, carrying out performance test on the candidate quantum circuits in a noise-containing environment, calculating an expression capability index and an anti-noise performance index, constructing a multi-objective loss function containing the expression capability index, the anti-noise performance index and a stability regular term, obtaining a local optimal quantum circuit based on multi-objective loss function iteration update, and carrying out quantum gate operation exchange, fusion and elimination on the local optimal quantum circuit iteration until the quantum circuits converge to obtain a final optimal quantum circuit. The invention improves the running efficiency and noise resistance robustness of the variable component sub-algorithm on NISQ equipment.

Inventors

  • YU XUTAO
  • LI SIXUAN
  • ZHANG ZAICHEN
  • LIU YUXIANG

Assignees

  • 东南大学

Dates

Publication Date
20260508
Application Date
20251203

Claims (10)

  1. 1. A quantum architecture searching method based on noise perception evaluation is characterized by comprising the following steps: (1) Obtaining an operation pool formed by all allowable quantum gate operations according to the topological connection and the native gate set of the target quantum hardware, and initializing a framework parameter matrix; (2) Carrying out micro-sampling on the architecture parameter matrix by adopting a re-parameterization technology to obtain candidate quantum circuits; (3) Performing performance test on candidate quantum circuits in a noise-containing environment, calculating an expression capacity index and an anti-noise performance index, constructing a multi-objective loss function comprising the expression capacity index, the anti-noise performance index and a stability regular term, and obtaining a local optimal quantum circuit based on iterative updating of the multi-objective loss function; (4) And carrying out quantum gate operation exchange, fusion and elimination on the locally optimal quantum circuit iteration until the quantum circuit converges to obtain the finally optimized quantum circuit.
  2. 2. The quantum architecture search method based on noise perception assessment according to claim 1, wherein step (1) specifically comprises: determining a connection edge allowing execution of a double-bit gate between quantum bits according to topological connection of target quantum hardware; According to a native gate set supported by target quantum hardware, determining available single-bit gate operation and double-bit gate operation, and combining all allowed single-bit gate operation and double-bit gate operation to form an operation pool; The architecture parameter matrix is initialized.
  3. 3. The quantum architecture search method based on noise perception assessment according to claim 1, further comprising, after step (1): Adopting a low-rank matrix decomposition mode to matrix the architecture parameters Decomposition into two low rank matrices And The product of (a), i.e 。
  4. 4. The quantum architecture search method based on noise perception assessment according to claim 1, wherein step (2) specifically comprises: For architecture parameter matrix Noise is obtained from Gumbel distributed sampling corresponding to each quantum gate operation in the operation pool And using temperature parameters Normalizing the controlled softmax function to obtain the selection probability of each operation Wherein the temperature parameter Gradually decreasing with search iteration; According to the probability of selection And selecting quantum gate operation, and assembling the selected quantum gate operation according to the line depth position sequence to obtain a candidate quantum line.
  5. 5. The quantum architecture search method based on noise perception assessment according to claim 1, wherein step (3) specifically comprises: Executing candidate quantum circuits for multiple times in a noisy environment, and calculating KL divergence between measurement result distribution and ideal Haar random distribution as an expression capability index; The candidate quantum circuit and the inverse circuit are connected in series to form an identical circuit, and the identical circuit is executed in a noise-containing environment, and the probability of returning to an initial state is counted to be used as an anti-noise performance index; Constructing a multi-objective loss function containing an expression capability index, an anti-noise performance index and a stability regular term; And (3) taking the minimum loss obtained by calculating the multi-objective loss function as a target, and iterating through a gradient descent method until convergence to obtain the local optimal quantum circuit.
  6. 6. The quantum architecture search method based on noise perception evaluation according to claim 1, wherein the multi-objective loss function is specifically: , Where C is a multi-objective loss function, 、 And As the weight coefficient of the light-emitting diode, In order to express the capability index, the expression level of the cell is, Is used as an index of the anti-noise performance, Is stable.
  7. 7. The quantum architecture search method based on noise perception assessment according to claim 1, wherein step (4) specifically comprises: performing a gate swap on the locally optimal quantum wire, wherein the gate swap is to move single bit gate operations forward to before double bit gate operations based on the ease of quantum gate relationships; Performing gate fusion on the quantum circuits, wherein the gate fusion is realized by integrating continuous similar rotary gate operations through accumulating rotary angles; Performing gate elimination on the quantum wire, wherein the gate elimination is a deletion reciprocal gate operation pair, a self-reciprocal gate operation pair, a zero angle rotation gate operation and an identity gate operation; gate fusion and gate elimination are performed on the quantum wire loop until the quantum gate operation number and wire depth of the quantum wire are no longer reduced.
  8. 8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the method of any one of claims 1-7.
  9. 9. A computer readable storage medium having stored thereon a computer program/instruction, which when executed by a processor, implements the method of any of claims 1-7.
  10. 10. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the method of any of claims 1-7.

Description

Quantum architecture searching method and device based on noise perception evaluation Technical Field The present invention relates to quantum technologies, and in particular, to a method and apparatus for searching a quantum architecture based on noise perception evaluation. Background In the era of noisy medium-scale quantum computing (NISQ), quantum devices are faced with multiple hardware constraints such as higher noise levels, limited number of qubits, and shorter coherence time. The variable component sub-algorithm (VQA) is one of the most promising applications in the NISQ era, and solves the optimization problem by a mixed mode of a Parameterized Quantum Circuit (PQC) and a classical optimizer. The PQC is composed of a series of quantum gates arranged in a specific sequence, and the architecture design of the PQC comprises the types, the sequences and the connection modes of the gates, so that the expression capability and the trainability of an algorithm are directly influenced. Traditional quantum circuit design relies on manual experience or heuristic rules, which is time-consuming and labor-consuming and difficult to ensure universality on different problems and hardware platforms. Microscale architecture search (DQAS) is a method of automatically discovering quantum wire architecture, relaxing discrete gate selection problems into continuous microscale optimization problems, and performing efficient searches through gradient descent. The existing DQAS method generally adopts Gumbel-Softmax heavy parameterization technology to realize differentiable discrete sampling, and the conversion process from random exploration to deterministic selection is controlled by temperature parameters. However, the existing method only considers the performance index under the ideal noiseless condition when evaluating the candidate line, and ignores the influence of noise factors such as gate errors, decoherence and measurement errors on real hardware. The evaluation mode leads the searched circuit to perform excellently in an ideal environment, but the performance is obviously reduced when the circuit runs on actual noise-containing hardware, and the practicability of a quantum algorithm is seriously affected. Meanwhile, the candidate lines obtained by searching often contain a large number of redundant quantum gates, and a systematic post-processing optimization strategy is lacked, so that efficient deployment on NISQ equipment with limited resources is not facilitated. Disclosure of Invention Aiming at the problems existing in the prior art, the invention aims to provide a quantum architecture searching method and device based on noise perception evaluation, which have higher robustness on noise-containing hardware and higher operation efficiency of a searched quantum circuit. In order to achieve the above object, the present invention provides the following technical solutions: a quantum architecture searching method based on noise perception evaluation comprises the following steps: (1) Obtaining an operation pool formed by all allowable quantum gate operations according to the topological connection and the native gate set of the target quantum hardware, and initializing a framework parameter matrix; (2) Carrying out micro-sampling on the architecture parameter matrix by adopting a re-parameterization technology to obtain candidate quantum circuits; (3) Performing performance test on candidate quantum circuits in a noise-containing environment, calculating an expression capacity index and an anti-noise performance index, constructing a multi-objective loss function comprising the expression capacity index, the anti-noise performance index and a stability regular term, and obtaining a local optimal quantum circuit based on iterative updating of the multi-objective loss function; (4) And carrying out quantum gate operation exchange, fusion and elimination on the locally optimal quantum circuit iteration until the quantum circuit converges to obtain the finally optimized quantum circuit. Further, the step (1) specifically includes: determining a connection edge allowing execution of a double-bit gate between quantum bits according to topological connection of target quantum hardware; According to a native gate set supported by target quantum hardware, determining available single-bit gate operation and double-bit gate operation, and combining all allowed single-bit gate operation and double-bit gate operation to form an operation pool; The architecture parameter matrix is initialized. Further, after the step (1), the method further comprises: Adopting a low-rank matrix decomposition mode to matrix the architecture parameters Decomposition into two low rank matricesAndThe product of (a), i.e。 Further, the step (2) specifically includes: For architecture parameter matrix Noise is obtained from Gumbel distributed sampling corresponding to each quantum gate operation in the operation poolAnd using temperature par