Noise attribution method, apparatus, medium, and program product for quantum error correction simulation

Abstract

The application discloses a noise attribution method, device, medium and program product for quantum error correction simulation; wherein the noise attribution method for quantum error correction simulation comprises the following steps: obtaining a quantum error correction simulation object, determining a plurality of noise events for representing fault occurrence conditions in a quantum error correction simulation process and a mapping relationship between each noise event and a physical noise parameter according to the quantum error correction simulation object; randomly sampling the noise events in a native computing layer to obtain a plurality of samples, determining signed sensitivity of the noise events based on at least part of the samples, and converting the signed sensitivity of the noise events into signed sensitivity of a physical noise position according to the mapping relationship to obtain an attribution result. The above scheme realizes calculation from sampling to an attribution result in a native computing layer, and reduces the overhead of cross-language data transfer and object construction in quantum error correction simulation.

Description

Technical Field

[0001] This invention generally relates to the field of quantum error correction simulation technology, and specifically to a noise attribution method, device, medium, and program product for quantum error correction simulation. Background Technology

[0002] Quantum computing, leveraging properties such as quantum superposition and quantum entanglement, exhibits potential advantages over traditional classical computing in solving certain computationally challenging problems. However, current physical qubits are highly susceptible to decoherence from the external environment and noise in control circuits. To achieve fault-tolerant quantum computing, quantum error correction (QEC) technology must be introduced. QEC involves encoding multiple physical qubits into a single logical qubit and periodically performing synthesis measurements and decoding without destroying the logical quantum state to identify and correct physical faults.

[0003] Large-scale quantum error correction (QEC) simulation is an indispensable tool for evaluating the performance of quantum error-correcting codes, optimizing decoding algorithms, and guiding hardware design before building or tuning real quantum hardware. Existing simulation frameworks typically use high-level languages ​​(such as Python) for experimental control and model representation, while employing low-level native languages ​​(such as C++ or Rust) for high-performance sampling. During simulation, per-sample fault trajectories, detector syndromes, and logic masks frequently involve memory allocation, object encapsulation, and serialization copying between the high-level language and the native low-level languages, resulting in extremely high cross-language data transfer overhead and limiting simulation throughput. Summary of the Invention

[0004] The present invention aims to provide a noise attribution method, apparatus, medium, and program product for quantum error correction simulation, which at least reduces the overhead of cross-language data transfer and object construction in quantum error correction simulation.

[0005] In a first aspect, the present invention provides a noise attribution method for quantum error correction simulation, comprising the following steps: Obtain a quantum error correction simulation object, and determine multiple noise events to characterize the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, as well as the mapping relationship between each noise event and physical noise parameters; The noise event is randomly sampled in the native computing layer to obtain multiple samples. The signed sensitivity of the noise event is determined based on at least a portion of the samples. According to the mapping relationship, the signed sensitivity of the noise event is converted into the signed sensitivity of the physical noise location to obtain the attribution result.

[0006] As an implementation method, the quantum error correction simulation object includes at least a simulation source, physical noise parameters, a decoding configuration, and a simulation configuration; After obtaining the quantum error correction simulation object, the following is also included: The simulation source, the physical noise parameters, the decoding configuration, and the simulation configuration are subjected to normalized serialization processing. Experiment identifiers are generated based on the simulation source after normalized serialization, the physical noise parameters, the decoding configuration, and the simulation configuration.

[0007] As a possible implementation, before randomly sampling the noise event in the native computing layer, the method further includes: The quantum error correction simulation object is compiled to obtain a bit compression sampling plan for generating multiple samples; The bit compression sampling plan is associated with and stored in conjunction with the simulation source and the decoding configuration.

[0008] As an implementation approach, during the next simulation, it is determined whether the current simulation source and decoding configuration are the same as the stored simulation source and decoding configuration. If they are the same, the bit compression sampling scheme associated with the stored simulation source and decoding configuration is reused; otherwise, the current quantum error correction simulation object is compiled to obtain another bit compression sampling scheme.

[0009] As an implementation method, the samples are grouped according to the sample dimension, and the occurrence state of the same noise event in multiple quantum error correction simulation samples is stored as an event state set. The event state set contains multiple occurrence states stored using bit-level encoding.

[0010] As a possible implementation, determining the signed sensitivity of the noise event based on at least a portion of the samples includes: The quantum error correction state of each sample is obtained, and the quantum error correction state is decoded according to the decoding configuration to obtain the logic error state corresponding to each sample. Based on the quantum error correction state and the logical error state, the signed sensitivity of each noise event is determined.

[0011] As an implementation method, the quantum error correction state includes a detector syndrome state and a real logical observable state; Decoding the quantum error-corrected state according to the decoding configuration includes: The logic correction state is obtained based on the detector syndrome state; The logic error state is determined based on the difference between the actual logical observable state and the logic correction state.

[0012] As an implementation method, determining the signed sensitivity of each of the noise events based on the quantum error correction state and the logic error state includes: The score of the noise event is determined based on the occurrence status of the noise event, the logic error status, and the sampling probability corresponding to the noise event in multiple samples. The signed sensitivity of the noise event is determined based on the statistical relationship between the scores corresponding to multiple samples and the logical error state.

[0013] As an implementation method, the probabilistic semantic relationship between each of the noise events is determined, the probabilistic semantic relationship including at least one of independence, mutual exclusion and correlation; If multiple noise events are independent, random sampling is performed on the multiple noise events; If multiple noise events have a mutually exclusive or correlated relationship, when multiple noise events can be converted into events that satisfy a preset probability consistency condition, the multiple noise events are converted into independent relationships and random sampling is performed; when multiple noise events cannot satisfy the preset probability consistency condition, the joint probability relationship between the multiple noise events is retained and random sampling is performed.

[0014] As an implementation method, when the multiple noise events cannot satisfy the preset probability consistency condition, the method further includes: if no explicit configuration allowing approximation is received, random sampling with independent noise events is prohibited.

[0015] As an implementation method, if multiple noise events are associated with the same physical noise location, the absolute value of the signed sensitivity of the physical noise locations corresponding to the multiple noise events is summed and then taken as the noise risk level of the physical noise location.

[0016] As an implementation method, the attribution result includes at least one of the following impact information: qubit identifier corresponding to physical noise parameters, quantum error correction round, gate type, operation type, and physical component information. The impact information is grouped and aggregated to obtain noise attribution results of different granularities.

[0017] In a second aspect, the present invention provides a noise attribution device for quantum error correction simulation, comprising: A preprocessing unit is used to acquire a quantum error correction simulation object, determine multiple noise events to characterize the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, and the mapping relationship between each noise event and physical noise parameters. A computing unit is configured to randomly sample the noise event in the native computing layer to obtain multiple samples, determine the signed sensitivity of the noise event based on at least a portion of the samples, and convert the signed sensitivity of the noise event into the signed sensitivity of the physical noise location according to the mapping relationship to obtain an attribution result.

[0018] Thirdly, the present invention provides a computing device having a computer program stored thereon, the computer program being used to implement the above-mentioned noise attribution method for quantum error correction simulation.

[0019] Fourthly, the present invention provides a non-transitory computer-readable storage medium, wherein when the instructions in the storage medium are executed by a processor of a computing device, the computing device is able to perform the noise attribution method of the above-described quantum error correction simulation.

[0020] Fifthly, the present invention provides a computer program product comprising instructions that, when executed, enable the noise attribution method for quantum error correction simulation as described above.

[0021] The above scheme enables computation from sampling to attribution results in the native computing layer, reducing the overhead of cross-language data transfer and object construction in quantum error correction simulation. Attached Figure Description

[0022] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A noise attribution method for quantum error correction simulation provided in this embodiment of the invention; Figure 2 A block diagram illustrating the principle of a noise attribution device for quantum error correction simulation provided in this embodiment of the invention; Figure 3 A schematic diagram of a computing device provided in an embodiment of the present invention. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0024] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] The terminology involved in this invention will first be explained as follows: Quantum Error Correction (QEC): Using redundant coding (such as surface code) to distribute logical qubit information across multiple physical qubits, physical errors are detected and corrected through stabilizer measurement, protecting quantum information from being destroyed by noise; Quantum error correction circuit description: A textual or programmed representation that precisely describes the initialization, encoding, syndrome extraction, measurement, and other operations of quantum error correction codes (such as surface codes) in the form of gate sequences or pulse sequences; Detection event model: An abstract model decoupled from specific circuit implementation, which transforms physical error propagation into correlation between detectors (stabilizer measurements); Noise events: Random events used to characterize the occurrence of faults during quantum error correction simulation; noise events can be detection event edges. Detection event edge: An edge connecting two detectors represents a specific error pattern (such as a bit flip on a data bit). The detectors at both ends of the edge will be triggered simultaneously when the error occurs. It is the basic unit for the decoder to perform matching. Physical noise parameters: describe the probability of various error processes in quantum error correction circuits; Decoder: An algorithm or software module that infers the most likely physical error based on the detector syndrome and provides a logical correction result; Decoder description: The text that specifies the decoder used to infer the error chain, generally including decoder type, backend, algorithm version, software implementation version, parameters, weight model, detector identifier, and logical observable identifier, etc. Sample configuration: Parameters that control sampling behavior; Normalized serialization: Converting the required data into a standard byte sequence according to uniform, definite rules that do not change with the storage method; Bit compression sampling scheme: a compilation product that converts quantum circuits or event detection models into highly compact bit-level instruction sequences, mapping each possible fault event to one bit or a few bits, so that error trajectories can be generated with extremely fast bit operations during sampling, avoiding the loop overhead of interpreted languages; Logical observable measurement: A Boolean quantity composed of one or more physical measurement results combined according to a specific relationship, used to determine the state of logical quantum information or logical error state; Fault event trajectory: A list / bitmask of all physical errors that actually occurred in a sample; Detector syndrome mask: a binary vector, where each bit corresponds to a detector; Logical observable mask: A mask indicating whether an uncorrectable error has occurred in a logical qubit; Decode correction mask: A bitmask of the recommended correction operations at the decoder output, used to indicate where to apply the X or Z correction gate; Signed sensitivity: A metric that measures the direction and magnitude of the influence of a detected event edge (or physical noise location) on logic failure. A positive value indicates that the presence of the noise source increases the probability of logic failure, while a negative value indicates that the presence of the noise source decreases the probability of logic failure. Compilation cache key: A key composed of the simulation source digest (content of the quantum error correction circuit / event detection model) and the sampling plan mode version. Experimental configurations with the same cache key can share immutable compilation outputs (such as bit compression sampling plans and side indices) and are allowed to have different noise parameters or decoder configurations, thereby achieving "compile once, reuse with parameters multiple times"; Native computing layer: High-performance core modules written in compiled languages ​​such as C++ / Rust, responsible for compilation, sampling, decoding and calling, and attribution calculation; separated from upper-layer interpreted interfaces such as Python, avoiding repeated copying and conversion of data between language boundaries, greatly improving throughput; Forward circuit sampling: When the joint distribution of physical noise cannot be accurately decomposed into independent detection event edges, it is necessary to return to the original quantum circuit and simulate the generation and propagation of noise step by step.

[0026] like Figure 1 As shown in the figure, the noise attribution method for quantum error correction simulation provided by the present invention is characterized by including the following steps: S1: Obtain a quantum error correction simulation object, and determine multiple noise events to characterize the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, as well as the mapping relationship between each noise event and physical noise parameters; When performing quantum error correction simulation, the system receives quantum error correction simulation objects submitted by upper-level users (e.g., through a Python interface). These quantum error correction simulation objects can also be referred to as experimental configurations.

[0027] Quantum error correction simulation objects generally include simulation sources, physical noise parameters, decoding configurations, analysis modes, and simulation configurations. Among them, the simulation source, physical noise parameters, decoding configurations, analysis modes, and simulation configurations are configured as invariant quantum error correction experimental configurations. Each invariant quantum error correction experimental configuration has a unique identifier. As long as the identifier is the same, the compilation results of the invariant quantum error correction experimental configuration can be reused in different rounds of simulation (also known as experiments). Each round of simulation can have different noise parameters or decoder configurations, thereby achieving "compile once, reuse multiple times with parameters" to reduce computational consumption and improve computational speed.

[0028] The simulation source can be a description of a quantum error correction circuit or a model of a detection event.

[0029] A quantum error correction circuit description may include the number of qubits, a time-ordered sequence of gate operations, detector declarations, logical observable declarations, and noise locations. In other examples, the quantum error correction circuit description may also include the type of each operation, measurement bond, basis, Pauli operation, repeating structure, noise locations, and metadata.

[0030] An event detection model includes at least a detector, a logical observable, and event detection edges. In other examples, the event detection model may also include event detection edge probabilities.

[0031] Physical noise parameters can be a uniform noise rate, or different parameters can be set for different physical operations. Physical operations characterize all behaviors triggered by control signals that the hardware can directly emit—including quantum gates, measurements, resets, idle waits, etc.

[0032] Simulation configuration describes the parameters used during simulation runtime, including but not limited to the number of samples, random seed, sample block size (also known as sample group size), sampling mode version, and explicit configurations such as whether approximation processing is allowed.

[0033] The following simulation source uses a quantum error correction circuit as an example for explanation.

[0034] The description of the quantum error correction circuit is compiled to identify the physical operations carrying noise model declarations. These declarations define the noise channel type attached to each physical operation and its corresponding physical noise parameters. Each physical operation carrying a noise model declaration constitutes a physical noise location, which includes the operation type (e.g., single-qubit gate, two-qubit gate, measurement, reset, or idle operation), the target qubit identifier, the noise channel type (e.g., single-qubit depolarization channel DEPOLARIZE1, two-qubit depolarization channel DEPOLARIZE2, bias Pauli channel PAULI_CHANNEL_1 / 2, independent bit flipping channel X_ERROR, etc.), and the corresponding physical noise parameters (e.g., error probability, bias probability, etc.).

[0035] Based on the noise channel type declared by the physical noise location, determine the probabilistic semantic relationships between multiple fault branches contained therein. These probabilistic semantic relationships include at least one of independence, mutual exclusion, and correlation.

[0036] For uniformly depolarized channels (such as DEPOLARIZE1 and DEPOLARIZE2), the error probabilities of each fault branch are symmetrically distributed and can be accurately transformed into the product of multiple independent sampling factors, which can be labeled as independent relationships. For biased Pauli channels (such as PAULI_CHANNEL_1), the fault branches cannot occur simultaneously, which can be labeled as mutually exclusive relationships. For multiple locations where error modes are jointly correlated due to shared physical noise sources, they can be labeled as correlated relationships.

[0037] For each physical noise location, the system compiles and expands to generate multiple noise events (i.e. detection event edges) to characterize the occurrence of faults during quantum error correction simulation, and establishes a mapping relationship between each noise event and physical noise parameters.

[0038] If multiple noise events are mutually exclusive or correlated, and can be converted into events that satisfy a preset probability consistency condition (e.g., the fault branch of a uniform depolarization channel can be precisely decomposed into independent factors), then the multiple noise events are converted to generate multiple independent detection event edges. For each generated detection event edge, the following mapping information is recorded: Physical location label: Identifies the physical operation and target qubit to which the detected event edge belongs; Physical noise parameter identifier: Identifies the physical parameters from which the side sampling probability of this detected event originates; Sampling parameter: Bernoulli sampling probability of the edge of the detected event; Variation relationship: The partial derivatives of the above sampling parameters with respect to the physical noise parameters; Probabilistic semantic label: Marks the edge of the detected event as an independent sample.

[0039] Furthermore, if multiple noise events cannot meet the preset probability consistency condition, the system retains the joint probability relationship between the multiple noise events, does not expand them into independent detection event edges, but marks the physical noise location as the forward circuit sampling location in the compilation product, and performs random sampling according to the original joint probability distribution at runtime.

[0040] Furthermore, for noise channels that are independently distributed (such as X_ERROR), a detection event edge can be directly generated for it, and the sampling parameters (equal to the physical noise parameters themselves) and their variation relationship (derivative is 1) of the detection event edge can be recorded.

[0041] All detected event edges and their mapping relationships are summarized to generate a detected event edge index. The detected event edge index is a pre-allocated contiguous memory layout table, in which each detected event edge corresponds to a fixed storage offset, which is used to store the occurrence state of the noise event in bit-compressed manner during subsequent random sampling.

[0042] The detection event edge index can also contain a preset index relationship between each detection event edge and the physical noise parameter, so that when determining the influence information of the physical noise parameter in the future, the sampling parameters and change relationship of the corresponding detection event edge can be directly located based on the index without the need for data format conversion or cross-module object transfer.

[0043] Based on the aforementioned physical location tags, the qubit identifier, quantum error correction round, gate type, operation type, and physical component information corresponding to each physical noise parameter can be obtained. This information is used as a dimension for grouping and aggregating the impact information of the physical noise parameters when generating subsequent noise attribution results, thus outputting noise attribution results at different granularities.

[0044] As one possible approach, after obtaining the quantum error correction simulation object, it also includes: The simulation source, the physical noise parameters, the decoding configuration, and the simulation configuration are subjected to normalized serialization processing. In some examples, the data undergoing normalized serialization also includes simulation patterns, metadata, etc.

[0045] During the normalization and serialization process, all target information (simulation source, physical noise parameters, decoding configuration, simulation configuration, simulation mode, metadata, etc.) is converted into normalized serialized data.

[0046] For example, different data types in the target information (such as floating-point precision, integer encoding, and string identifiers) can be uniformly converted into a predefined standardized format. For example, but not limited to, standardizing floating-point probability values ​​in physical noise parameters into fixed-precision binary representations, standardizing quantum bit identifiers into continuous integer indices, and standardizing decoder version information into semantic version strings, etc.

[0047] For set-type data in the target information (such as a list of noise channels or a detector declaration sequence), they are deterministically arranged according to predefined sorting rules (such as ascending order by qubit identifier or the order in which the operation appears in the quantum error correction circuit) to eliminate the differences in serialization results caused by different input orders.

[0048] In addition, redundant content that does not affect the simulation semantics, such as comments, whitespace characters, and implicit declarations of default values, is removed from the target information; equivalent but different expressions are normalized into a unique representation.

[0049] Then, the target information, after being processed by format unification, order determination, and normalization, is converted into a byte sequence according to a predefined encoding format (such as binary compact encoding, JSON canonical format, or hash tree structure). This byte sequence completely represents the current configuration state of the simulation object, and the same configuration will inevitably produce the same byte sequence.

[0050] Experiment identifiers are generated based on the simulation source after normalized serialization, the physical noise parameters, the decoding configuration, and the simulation configuration.

[0051] For example, but not limited to, a digest value of the target information after normalized serialization is calculated using a deterministic hash function (such as SHA-256, BLAKE3, etc.), and the digest value is used as the experimental identifier of the quantum error correction simulation experiment.

[0052] The experiment identifier `experiment_id` is defined as `Hash(domain, source_digest, decoder_digest, noise, mode, options, metadata)`. Here, `domain` represents the identifier domain, `source_digest` is the simulation source digest value, `decoder_digest` is the decoder digest, `noise` is the physical noise parameter, `mode` is the simulation mode, `options` are the simulation configuration, and `metadata` is the metadata. The simulation source digest is obtained by hashing the simulation source, and the decoder digest is obtained by hashing the decoding configuration.

[0053] When determining the test identifier, the serialized byte sequence of the target information is concatenated into a unified input buffer in a fixed order. The concatenation order can be predefined as needed (e.g., the order in the test identifier mentioned above; in other examples, it could be a different order) to ensure that different calls produce the same concatenation result.

[0054] A deterministic hash function is applied to the concatenated input buffer to obtain a fixed-length digest value, which serves as the experimental identifier.

[0055] After obtaining the experiment identifier, it can be associated with the corresponding normalized serialized target information, the compilation cache key, and the subsequently generated bit compression sampling plan, and stored in the compilation cache. When the same or equivalent simulation source is received again, the cache can be directly hit based on the same experiment identifier, avoiding repeated compilation.

[0056] The compilation cache key includes the simulation source summary value and the sampling plan mode version.

[0057] By eliminating irrelevant differences in input format, order, and redundant content through normalized serialization, the same experimental identifier is generated for semantically equivalent target information, which greatly improves the hit rate of the compilation cache and avoids the waste of computing resources caused by repeated compilation.

[0058] The generation process of experimental identifiers is completely deterministic. The same target information, that is, the same experimental identifiers are generated at any time and at any computing node, ensures the reproducibility and traceability of simulation results.

[0059] Different decoding or simulation configurations (such as different sampling mode versions, different approximation allowance settings, etc.) will generate different experimental identifiers, ensuring that the compilation products under different configurations are strictly isolated, and avoiding incorrect attribution results caused by configuration confusion.

[0060] Experimental identifiers can serve as unique fingerprints for simulation tasks, enabling rapid retrieval of corresponding compilation artifacts, sampling plans, and attribution results, thus facilitating problem localization and performance analysis.

[0061] As a possible implementation, before randomly sampling the noise event in the native computing layer, the method further includes: The quantum error correction simulation object is compiled to obtain a bit compression sampling plan for generating multiple samples; After the target information is normalized and serialized and the experimental identifier is generated, the native computing layer compiles the quantum error correction simulation object to generate a bit compression sampling plan for multiple samples.

[0062] Specifically, the normalized simulation source is analyzed to obtain the physical operation carrying the noise model declaration. The probabilistic semantic relationship between each fault branch is determined according to the noise channel type. The convertible mutually exclusive or related fault branches are expanded into independent detection event edges according to the preset probability consistency condition. The non-convertible ones are marked as the sampling position of the forward circuit.

[0063] All detection event edges are arranged into a continuous list according to a predefined traversal order (such as ascending order by qubit identifier, or by detector declaration order, etc.). Each detection event edge is assigned a globally unique edge index (EdgeID) for address location in subsequent bit storage layout.

[0064] Based on a preset sample block size (e.g., a sample block contains 1024 samples; other numbers are also possible, without specific limitations), a contiguous memory storage area is pre-allocated for each detection event edge. For each detection event edge e, its bit storage area contains W machine words (e.g., 64-bit unsigned integers), and the k-th bit of each machine word corresponds to the occurrence state of that detection event edge in the k-th sample. The storage areas of all detection event edges are arranged contiguously in memory with fixed offsets, forming a preset index relationship, i.e., bit layout information.

[0065] Then, based on the probabilistic semantic relationship of the detected event edges, a corresponding sampling instruction sequence is generated. For each detected event edge marked as independent, a Bernoulli sampling instruction is generated, which includes the sampling parameter q of that edge. e It also includes the storage offset of the noise channel in memory; for the physical location marked as a forward sampling point, a forward sampling instruction is generated, which contains the complete parameters of the original noise channel and a description of the joint probability distribution.

[0066] At least the edge index, preset index relationship, and sampling instruction sequence should be packaged into a bit-compressed sampling plan.

[0067] The bit compression sampling plan is associated with and stored in conjunction with the simulation source and the decoding configuration.

[0068] In a simulation project, multiple rounds of simulation may be performed, and a unified simulation source can be used for these multiple rounds of simulation.

[0069] In multi-round simulations, during the next simulation, it is determined whether the current simulation source and decoding configuration are the same as the stored simulation source and decoding configuration. For example, if the next simulation is the 50th round, it is determined whether the simulation source and decoding configuration of the 50th round are the same as the stored simulation source and decoding configuration. If they are the same, the bit compression sampling scheme associated with the stored simulation source and decoding configuration is reused; otherwise, compilation is performed based on the current quantum error correction simulation object to obtain a different bit compression sampling scheme. As mentioned above, the simulation source and decoding configuration will have an experimental identifier, which allows direct cache hits based on the same experimental identifier, avoiding repeated compilation.

[0070] As an implementation method, the samples are grouped according to the sample dimension, and the occurrence state of the same noise event in multiple quantum error correction simulation samples is stored as an event state set. The event state set contains multiple occurrence states stored using bit-level encoding.

[0071] Specifically, after randomly sampling the edges of detected events in the native computing layer, multiple samples are obtained. Each sample records whether each detected event edge occurred in the current sampling round. The sample set is divided into at least one sample block along the sample index direction, and each sample block contains a preset number of consecutive samples. For example, if each sample block contains 1024 samples, then samples 0 to 1023 constitute sample block 0, samples 1024 to 2047 constitute sample block 1, and so on. Using sample dimension to group and obtain sample blocks allows subsequent state determination, decoding, and attribution calculations to be executed continuously in the same computing context on a sample block basis. This fully utilizes the CPU's SIMD (Single Instruction Multiple Data) instruction set or the GPU's parallel computing capabilities, avoiding the interpretability overhead of processing samples one by one.

[0072] For each detected event edge, the occurrence states of the detected event edge are collected from multiple samples within the same sample block to form the event state set corresponding to the detected event edge. The multiple occurrence states in the event state set are stored in a pre-allocated contiguous memory area using bit-level encoding.

[0073] S2: Randomly sample the noise event in the native computing layer to obtain multiple samples, determine the signed sensitivity of the noise event based on at least a portion of the samples, and convert the signed sensitivity of the noise event into the signed sensitivity of the physical noise location according to the mapping relationship to obtain the attribution result.

[0074] Specifically, determining the signed sensitivity of the noise event based on at least a portion of the samples includes: The quantum error correction state of each sample is obtained, and the quantum error correction state is decoded according to the decoding configuration to obtain the logic error state corresponding to each sample. The quantum error correction state of each sample is read from the aforementioned bit storage area. This quantum error correction state includes both the detector syndrome state and the real logical observable state.

[0075] Specifically, for the kth sample in the current sample block, according to the above detection event edge index, the occurrence states of each detection event edge associated with detector i in sample k are XORed to obtain the detector syndrome state; according to the logical observable declaration in the simulation source, the values ​​of each related measurement bit constituting the logical observable in sample k are XORed to obtain the real logical observable state.

[0076] The step of decoding the quantum error-corrected state according to the decoding configuration includes: The logic correction state is obtained based on the detector syndrome state; After obtaining the detector syndrome state of each sample, the detector syndrome state is processed based on the preset decoding configuration to obtain the logical correction state.

[0077] Specifically, based on the detection event edge index and the preset index relationship, the detector syndrome state corresponding to each sample in the current sample block is read. Each detector syndrome state is obtained by XORing the occurrence states of several related noise event edges and is stored in a pre-allocated machine word in single-bit form. Furthermore, the single-bit stored detector syndrome states are extracted in batches according to the input format required by the decoder (such as sparse matrix rows, graph node list, or tensor network input tensor) to form decoded input data.

[0078] The above input data is passed to the corresponding decoder through the decoder plugin interface with the version number.

[0079] After receiving the detector syndrome state, the decoder constructs a corresponding decoding graph or equivalent model based on a preset error-correcting code topology. For topological quantum error-correcting codes such as surface codes and color codes, the decoding graph is typically composed of detectors as nodes, with the weights of detection event edges determined by the logarithmic ratio of the occurrence probabilities of those edges. For example, the decoder searches the decoding graph for error pattern estimates that minimize the detector syndrome matching cost and generates correction suggestions for each logical observable accordingly. These correction suggestions can be output in bit-level encoded form. For a logical observable, if the decoder determines that a flip needs to be applied to the observable to cancel the predicted error, the corresponding correction bit is set to 1; otherwise, it is set to 0.

[0080] The logic error state is determined based on the difference between the actual logical observable state and the logic correction state.

[0081] Based on a pre-defined index relationship, the actual logical observable state and the logical correction state of the same logical observable in the same sample are located to their corresponding bits within the same machine word. Since both are stored using the same bit-level encoding method, the alignment operation can be completed through direct memory address offset, without the need for data rearrangement or format conversion. A bit-by-bit XOR operation is performed on the actual logical observable state and the logical correction state to obtain the logical error state.

[0082] Based on the quantum error correction state and the logical error state, the signed sensitivity of each noise event is determined.

[0083] Specifically, the score of the e-th detected event edge in the k-th sample is calculated according to the following formula: s_e,k =X_ e,k / q_ e,k -(1-X_ e,k) / (1-q_ e,k); Where X_e,k is the occurrence state of noise event e (Bernoulli variable), taking the value of 0 or 1; q_e,k is the sampling probability of noise event e.

[0084] The signed sensitivity is calculated using the following formula: g_e = (1 / N) SUM_k (L_k-b)s_e,k; Where N is the total number of samples; L_k is the logic error state of the k-th sample; b is the baseline of the logic error rate (usually the average of L_k across all samples); and s_e,k is the score of the e-th detected event edge in the k-th sample.

[0085] As an implementation method, when the multiple noise events cannot satisfy the preset probability consistency condition, the method further includes: if no explicit configuration allowing approximation is received, random sampling with independent noise events is prohibited.

[0086] Specifically, a pre-defined probability consistency condition is applied to multiple retrieval event edges that have mutually exclusive or related relationships. The pre-defined probability consistency condition means that the joint probability distribution formed by the multiple detection event edges can be equivalently transformed into the product form of a set of independent Bernoulli variables, and the distribution before and after the transformation remains consistent across all statistical moments (i.e., the transformed independent sampling results are unbiased and do not differ from the original joint distribution in terms of expectation, variance, and higher-order moments).

[0087] When multiple detected event edges are determined to fail to meet the preset probability consistency condition, the system further checks whether the current simulation configuration contains an explicit configuration that allows approximation. An explicit configuration that allows approximation means that the user, through explicit parameter declarations in the simulation configuration, actively indicates that, within a certain accuracy loss range, detected event edges that do not meet the preset probability consistency condition can be forcibly approximated as independent noise events for sampling.

[0088] If the explicit configuration is detected, approximate sampling is performed; if the system does not detect the explicit configuration, random sampling is prohibited from being performed on independent detection events.

[0089] The approximate sampling includes performing an approximate transformation on noise events that do not meet the preset probability consistency condition according to a preset approximation strategy (such as forcibly normalizing the probabilities of each branch of the bias Pauli channel to a uniform distribution, or merging low-probability branches into high-probability branches), generating approximate independent noise events, and performing random sampling based on the approximate sampling parameters.

[0090] Prohibiting random sampling of independent detection event edges includes refusing to convert multiple detection event edges that do not meet the preset probability consistency condition into independent noise events without receiving an explicit configuration that allows approximation, and prohibiting random sampling of the multiple detection event edges in the manner of independent Bernoulli sampling.

[0091] As an implementation method, if multiple noise events are associated with the same physical noise location, the signed sensitivity of the physical noise locations corresponding to the multiple noise events is summed and then the absolute value is taken as the noise risk level of the physical noise location. This method preserves the directional cancellation of different fault branches on the logic failure index; it avoids the systematic amplification of physical operations involving multiple branches caused by first taking the absolute value of each branch and then summing them.

[0092] As an implementation method, the attribution result includes at least one of the following impact information: qubit identifier corresponding to physical noise parameters, quantum error correction round, gate type, operation type, and physical component information. The impact information is grouped and aggregated to obtain noise attribution results of different granularities.

[0093] The noise attribution results of different granularities described here can include edge-level, position-level, gate-level, round-level, qubit-level, and detector-level attribution results. Classifying the attribution results according to different granularities facilitates result review. See the table below for details:

[0094] To facilitate the review of attribution results, the attribution results may also include experimental identifiers, simulation source summaries, decoder identity and version, software version, result mode, sample block configuration, probabilistic semantics, number of samples, number of errors, number of discards, and runtime.

[0095] Secondly, such as Figure 2 As shown, an embodiment of the present invention provides a noise attribution device for quantum error correction simulation, comprising: Preprocessing unit 1 is used to acquire a quantum error correction simulation object, determine multiple noise events to characterize the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, and the mapping relationship between each noise event and physical noise parameters. The computing unit 2 is used to randomly sample the noise event in the native computing layer to obtain multiple samples, determine the signed sensitivity of the noise event based on at least a portion of the samples, and convert the signed sensitivity of the noise event into the signed sensitivity of the physical noise location according to the mapping relationship to obtain the attribution result.

[0096] The noise attribution device for quantum error correction simulation is used to implement the above method. The specific steps are described in the method description and will not be repeated here.

[0097] Thirdly, embodiments of the present invention also provide a computing device having a computer program stored thereon, the computer program being used to implement the noise attribution method based on the above-described quantum error correction simulation.

[0098] Among them, computing devices include workstations, servers, computing clusters, or classical computing devices connected to a quantum processor control system; the processor used can be a central processing unit, a graphics processing unit, or other processors that support bitwise operations.

[0099] The following is for reference. Figure 3 It shows a schematic diagram of a computing device suitable for implementing embodiments of the present invention.

[0100] like Figure 3 As shown, the computing device includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 502 or a program loaded from storage section 508 into random access memory (RAM) 503. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0101] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed.

[0102] In particular, in accordance with the embodiments of the noise attribution method for quantum error correction simulation disclosed in this invention, reference is made above. Figure 1 The described process can be implemented as a computer software program. For example, embodiments of the noise attribution method for quantum error correction simulation disclosed in this invention include a computer program product comprising a computer program tangibly contained on a machine-readable medium, the computer program containing instructions for performing... Figure 1 The program code for the method. In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable media 511.

[0103] It should be noted that the computer-readable storage medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. The computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0104] In this invention, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium may also be any computer-readable storage medium other than a computer-readable storage medium, which can transmit, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wireless, wireline, optical fiber, RF, etc., or any suitable combination thereof.

[0105] 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 the noise attribution methods for quantum error correction simulations of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing the 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, or sometimes 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.

[0106] The units or modules described in the embodiments of the present invention can be implemented in software or hardware. The described units or modules can also be housed in a processor; for example, a processor may be described as including an acquisition model, a first calculation module, and a second calculation module. The names of these units or modules do not necessarily limit the specific unit or module itself. For example, a preprocessing unit may be described as "for acquiring a quantum error correction simulation object, determining multiple noise events characterizing the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, and the mapping relationship between each noise event and physical noise parameters."

[0107] Fourthly, embodiments of the present invention also provide a non-transitory computer-readable storage medium, which may be the computer-readable storage medium included in the aforementioned apparatus in the above embodiments; or it may be a standalone computer-readable storage medium not assembled into the device. The computer-readable storage medium stores one or more programs, which are used by one or more processors to execute the noise attribution method for quantum error correction simulation described in the present invention.

[0108] Fifthly, embodiments of the present invention also provide a computer program product, characterized in that the computer program product includes instructions, which, when executed, enable the noise attribution method of quantum error correction simulation as described above to be implemented.

[0109] It should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., used above to indicate orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, unless otherwise stated, "a plurality of" means two or more.

[0110] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the specific combination of the above-described technical features, but also includes other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in this invention.

Claims

1. A noise attribution method for quantum error correction simulation, characterized in that, Includes the following steps: Obtain a quantum error correction simulation object, and determine multiple noise events to characterize the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, as well as the mapping relationship between each noise event and physical noise parameters; The noise event is randomly sampled in the native computing layer to obtain multiple samples. The signed sensitivity of the noise event is determined based on at least a portion of the samples. According to the mapping relationship, the signed sensitivity of the noise event is converted into the signed sensitivity of the physical noise location to obtain the attribution result.

2. The noise attribution method for quantum error correction simulation according to claim 1, characterized in that, The quantum error correction simulation object includes at least a simulation source, physical noise parameters, decoding configuration, and simulation configuration; After obtaining the quantum error correction simulation object, the following is also included: The simulation source, the physical noise parameters, the decoding configuration, and the simulation configuration are subjected to normalized serialization processing. Experiment identifiers are generated based on the simulation source after normalized serialization, the physical noise parameters, the decoding configuration, and the simulation configuration.

3. The noise attribution method for quantum error correction simulation according to claim 2, characterized in that, Before randomly sampling the noise events in the native computing layer, the method further includes: The quantum error correction simulation object is compiled to obtain a bit compression sampling plan for generating multiple samples; The bit compression sampling plan is associated with and stored in conjunction with the simulation source and the decoding configuration.

4. The noise attribution method for quantum error correction simulation according to claim 3, characterized in that, In the next simulation, it is determined whether the current simulation source and decoding configuration are the same as the stored simulation source and decoding configuration. If they are the same, the bit compression sampling scheme associated with the stored simulation source and decoding configuration is reused. Otherwise, it is compiled based on the current quantum error correction simulation object to obtain another bit compression sampling scheme.

5. The noise attribution method for quantum error correction simulation according to any one of claims 1-4, characterized in that, The samples are grouped according to the sample dimension, and the occurrence state of the same noise event in multiple quantum error correction simulation samples is stored as an event state set. The event state set contains multiple occurrence states stored using bit-level encoding.

6. The noise attribution method for quantum error correction simulation according to any one of claims 2-4, characterized in that, The determination of the signed sensitivity of the noise event based on at least a portion of the samples includes: The quantum error correction state of each sample is obtained, and the quantum error correction state is decoded according to the decoding configuration to obtain the logical error state corresponding to each sample. Based on the quantum error correction state and the logic error state, the signed sensitivity of each noise event is determined.

7. The noise attribution method for quantum error correction simulation according to claim 6, characterized in that, The quantum error correction state includes the detector syndrome state and the real logical observable state; Decoding the quantum error-corrected state according to the decoding configuration includes: The logic correction state is obtained based on the detector syndrome state; The logic error state is determined based on the difference between the actual logical observable state and the logic correction state.

8. The noise attribution method for quantum error correction simulation according to claim 6, characterized in that, The determination of the signed sensitivity of each noise event based on the quantum error correction state and the logical error state includes: The score of the noise event is determined based on the occurrence status of the noise event, the logic error status, and the sampling probability corresponding to the noise event in multiple samples. The signed sensitivity of the noise event is determined based on the statistical relationship between the scores corresponding to multiple samples and the logical error state.

9. The noise attribution method for quantum error correction simulation according to any one of claims 1-4, characterized in that, Determine the probabilistic semantic relationships between the noise events, wherein the probabilistic semantic relationships include at least one of independence, mutual exclusion, and correlation; If multiple noise events are independent, random sampling is performed on the multiple noise events; If multiple noise events have a mutually exclusive or correlated relationship, and if multiple noise events can be converted into events that satisfy a preset probability consistency condition, then the multiple noise events are converted into an independent relationship and random sampling is performed. When multiple noise events fail to meet the preset probability consistency condition, random sampling is performed while preserving the joint probability relationship between the multiple noise events.

10. The noise attribution method for quantum error correction simulation according to claim 9, characterized in that, The method further includes, when multiple noise events fail to meet the preset probability consistency condition, prohibiting random sampling with independent noise events if no explicit configuration allowing approximation is received.

11. The noise attribution method for quantum error correction simulation according to any one of claims 1-4, characterized in that, If multiple noise events are associated with the same physical noise location, the absolute value of the summation of the signed sensitivities of the physical noise locations corresponding to the multiple noise events is used as the noise risk level of the physical noise location.

12. The noise attribution method for quantum error correction simulation according to any one of claims 1-4, characterized in that, The attribution results include at least one of the following impact information: quantum bit identifier corresponding to physical noise parameters, quantum error correction round, gate type, operation type, and physical component information. The impact information is grouped and aggregated to obtain noise attribution results of different granularities.

13. A noise attribution device for quantum error correction simulation, characterized in that, include: A preprocessing unit is used to acquire a quantum error correction simulation object, determine multiple noise events to characterize the fault occurrence during the quantum error correction simulation process based on the quantum error correction simulation object, and the mapping relationship between each noise event and physical noise parameters. A computing unit is configured to randomly sample the noise event in the native computing layer to obtain multiple samples, determine the signed sensitivity of the noise event based on at least a portion of the samples, and convert the signed sensitivity of the noise event into the signed sensitivity of the physical noise location according to the mapping relationship to obtain an attribution result.

14. A computing device, characterized in that, It stores a computer program for implementing the noise attribution method for quantum error correction simulation according to any one of claims 1-12.

15. A non-transitory computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the computing device, the computing device is able to perform the noise attribution method for quantum error correction simulation according to any one of claims 1-12.

16. A computer program product, characterized in that, The computer program product includes instructions that, when executed, cause the noise attribution method for quantum error correction simulation as described in any one of claims 1-12 to be implemented.

Images