Method and device for the computer-implemented prediction of the performance of a quantum computer
Patent Information
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- SIEMENS AG
- Filing Date
- 2024-08-09
- Publication Date
- 2026-05-06
Smart Images
Figure EP2024072545_13022025_PF_FP_ABST
Abstract
Description
[0001] Description
[0002] Method and device for computer-implemented
[0003] Predicting the performance of a quantum computer
[0004] The invention relates to a method and a device for computer-implemented prediction of the performance of a quantum computer with at least one quantum processing unit, which are characterized by respective performance parameters.
[0005] Quantum computing represents a new paradigm in data processing that uses the fundamental principles of quantum mechanics to perform calculations. While a conventional computer works with bits, each of which can assume one of two discrete states, 0 or 1, a quantum computer uses so-called quantum bits (qubits). A qubit can assume not only one of two quantum states, | 0> or | 1>, but also any superposition (linear combination) of these two states. Such a linear combination is also called a superposition or superposition state.
[0006] Algorithms and applications that utilize quantum mechanical resources can be represented simply and efficiently using so-called quantum circuits. A quantum circuit is a computational routine consisting of coherent quantum operations on quantum bits.
[0007] A quantum circuit is executed on a quantum computer and can be supplemented by further instructions on a classical computer. In this case, it is called a hybrid algorithm. Such an algorithm can also be iterative, meaning it can consist of repeating classical and quantum components. Quantum circuits enable a quantum processing unit to receive classical information and output a classical solution, with the quantum processing unit using quantum principles such as interference and entanglement to perform the calculation.
[0008] Currently existing quantum computers, which comprise one or more quantum processing units, use very different technologies to represent qubits and their interactions. Some of the best known technologies are superconducting circuits, ion traps, and photonic circuits. Other approaches are based on neutral atom traps, diamond-based qubits, or quantum dots.
[0009] The technology chosen to implement the qubits determines the technology required to control and read out the qubits in the quantum processing unit (QPU). While the properties of quantum processing units in terms of the number of qubits and error rates are likely to improve significantly in the future, properties such as coherence time, gate operation time and connectivity are more dependent on the choice of technology used, but can also improve over time. The coherence time describes how long information can be stored and used in a qubit before a so-called decoherence occurs and the stored information is lost. The gate operation time essentially determines how long it takes to execute a quantum circuit. Connectivity describes how many other qubits a particular qubit can directly interact with.can interact .
[0010] While superconducting qubits have short gate operation times in the range of a few nanoseconds, the coherence time before decoherence is comparatively short, in the microsecond range. Their connectivity is limited to neighboring qubits, where in a square lattice geometry, a qubit can have at most four neighbors.
[0011] In contrast, the gate operation times for qubits represented by ion traps are comparatively slow (in the range of microseconds), but these qubits, on the other hand, have a comparatively long coherence time in the range of minutes and an all-to-all connectivity.
[0012] In addition to the differences mentioned, the scalability and thus the costs of implementing a quantum circuit as well as the operating conditions under which the respective quantum processing units are operated are highly dependent on the technology.
[0013] Different types of quantum processing units can implement different models of quantum computation. Quantum computation can, for example, be gate-based, measurement-based, or use so-called topological qubits. Quantum computers whose quantum processing units are based on photonic circuits typically use measurement-based quantum computation.
[0014] The increasing availability and performance of quantum computers make them interesting for a variety of applications. These range from quantum simulation (materials research, catalysts, pharmaceuticals, etc.) to optimization tasks (portfolio allocation, production, and logistics) to solving differential equations and machine learning.
[0015] Currently, available quantum computers are located almost exclusively in laboratories under controlled environmental conditions, supervised by technical experts. The computing power of quantum computers can be accessed remotely. There are efforts to install quantum computers on user premises in the future, particularly due to latency requirements (e.g., in financial applications and real-time optimizations) or for data protection and / or security reasons. In such applications, quantum computers will likely be located in data centers or business premises.
[0016] A practical problem is that quantum computers can easily be disturbed by environmental influences. Depending on the specific implementation of the quantum computer, e.g., using superconducting qubits, ion traps, or integrated photonic circuits, quantum computers are sensitive to certain types of external influences, such as vibrations, electromagnetic fields, temperature fluctuations, or cosmic radiation.
[0017] To fully exploit the power of quantum computers, a placement that minimizes the negative effects mentioned above is therefore required. For example, temperature fluctuations must be kept to a minimum or vibrations must be kept at bay. Assessing the impact of various environmental influences on a specific type of quantum computer requires a detailed understanding of its physical architecture, which represents a major obstacle for many quantum computer users. The same applies to the recalibration of the quantum computer, which must be performed regularly.
[0018] There is a need to be able to estimate the performance of quantum computers at any location, and in particular to take into account the effects of nearby interfering factors, such as machines.
[0019] The object of the invention is to provide a method and a device that enable a prediction of the performance of a quantum computer having at least one quantum processing unit. These objects are achieved by a method according to the features of claim 1, a device according to the features of claim 8, and a computer program product according to the features of claim 10. Advantageous embodiments emerge from the dependent claims.
[0020] According to a first aspect of the invention, a method for computer-implemented prediction of the performance of a quantum computer with at least one quantum processing unit is proposed. The at least one quantum processing unit is characterized by respective performance parameters. In particular, the performance parameters comprise one or more of the following parameters: a processor performance; a main memory; a number of qubits; properties of the qubits (gate operation times, coherence time before the occurrence of decoherence; connectivity to other qubits), a networking rate of the qubits; an error rate for respective gate operations; a coherence time.
[0021] The method for computer-implemented predictions is preferably to be understood as meaning that the prediction is carried out in a computer-implemented manner. In a preferred embodiment of the invention, the method for computer-implemented predictions according to the invention can therefore also be referred to as a computer-implemented method for predicting the performance of a quantum computer with at least one quantum processing unit.
[0022] It is understood that the quantum processing unit is a part of the quantum computer, so the quantum processing unit can alternatively always be referred to as the quantum processing unit of the quantum computer.
[0023] For planned operation of the quantum computer at a predetermined location, the following steps are carried out: In a step i), at least one environmental parameter is read in. The term “environmental parameter” refers to a digital environmental parameter. The term “reading in an environmental parameter” means that the environmental parameter is received by a processor which carries out the method according to the invention. The at least one environmental parameter is or was recorded using one or more sensors. The term “sensor” refers to a recording unit which is designed to measure a specific environmental parameter and to provide the measured value as a digital value which represents the environmental parameter for further processing.
[0024] The sensor(s) are installed, in particular, at or near the specified location. The sensors can be configured to measure any type of environmental parameter. In particular, one or more sensors can be designed to measure vibrations, temperatures, electromagnetic fields, or radiation. One or more of the aforementioned sensor types can be used.
[0025] In a step ii), at least one state parameter of the at least one quantum processing unit is determined by processing the at least one environmental parameter by a trained data-driven model. The at least one environmental parameter is fed into the trained data-driven model as digital input information, and the trained data-driven model supplies the at least one state parameter as digital output information. The at least one state parameter describes a performance of the at least one quantum processing unit.
[0026] The method according to the invention enables a simple estimation of the performance of the quantum computer in different environments. The performance of the at least one quantum processing unit of the quantum computer results from the combination of a simulated operation of the quantum computer, taking into account the input data, with the at least one digital environmental parameter.
[0027] As explained, the method uses a trained data-driven model. The model was trained with training data that includes a variety of quantum computer performance values depending on different environmental conditions.
[0028] In principle, any known data-driven model that has been trained using machine learning methods can be used in the method of the present invention. In a preferred embodiment, the trained data-driven model is a neural network. Nevertheless, other data-driven models can also be implemented in the method of the present invention, such as models based on decision trees or support vector machines.
[0029] In a practical embodiment, the determination of the at least one state parameter of the at least one quantum processing unit is repeated if a measure describing the performance is undershot. In other words, in particular, an increased error rate leads to a repeated execution of the "simulation" in order to increase the accuracy of the performance parameter, i.e., the at least one state parameter, under given environmental parameters.
[0030] In a further preferred embodiment, the determination of the at least one state parameter of the at least one quantum processing unit is repeated for a plurality of predetermined locations if a measure describing the performance of the at least one state parameter is not reached. With the aid of this embodiment, if an initially planned location turns out to be unsuitable due to the performance measure of the at least one state parameter, an attempt is made to take other locations into account with regard to the effects of the environmental influences present there. For example, an acceptable performance measure can be achieved by a slight change to the planned location which is exposed to lower vibrations and / or noise and the like.
[0031] A further advantageous embodiment provides for self-calibration to be performed for the at least one quantum processing unit of the quantum computer if a performance-defining measure of the at least one state parameter is undershot. This makes it possible to detect fundamental recalibrations and maintenance work: For example, modifications to the quantum computer hardware can be incorporated into the simulation of the trained data-driven model to predict possible side effects. This function can also be used to train users in the use of the quantum computer.
[0032] A further advantageous embodiment provides that the at least one state parameter includes an error rate for respective gate operations. Furthermore, other state parameters can also be considered.
[0033] It is also expedient if the at least one environmental parameter is varied within a predetermined bandwidth around the respective measured value and for each variation the at least one state parameter is determined. This makes it possible to determine the extent to which certain changes in the environmental parameters already have a detrimental effect on the performance of the quantum computer. This makes it possible to examine predetermined locations whose environmental parameters are subject to certain fluctuations in advance for their suitability for placing the quantum computer. According to a second aspect of the present invention, a device for computer-implemented prediction of the performance of a quantum computer with at least one quantum processing unit is proposed. The at least one quantum processing unit is characterized by respective performance parameters.The device comprises a processor configured to carry out the steps of the method according to one or more embodiments of the invention for planned operation of the quantum computer at a predetermined location.
[0034] According to a third aspect of the present invention, a computer program product is proposed with program code stored on a machine-readable carrier for carrying out a method according to one of several embodiments of the present invention.
[0035] The invention is explained in more detail below using an exemplary embodiment in the drawing.
[0036] Fig. 1 shows a schematic representation of a quantum computer with a quantum processing unit comprising a processor for carrying out an embodiment of the present invention.
[0037] Fig. 1 shows a quantum computer QC with, for example, a single quantum processing unit QPU . The quantum processing unit QPU can be based on any known technology to represent qubits and their interactions . For example, the quantum processing unit QPU could be based on superconducting circuits, ion traps or photonic circuits . The type of qubits determines the technology required to control and read the qubits in the quantum processing unit . Properties such as coherence time, gate operation times and connectivity depend on this. The quantum processing unit QPU is characterized by performance parameters . The performance parameters include, for example, processor power, RAM, number of qubits and the properties of the qubits, in particular a networking rate, a coherence time and an error rate for respective gate operations .In addition to the parameters mentioned, the performance parameters may also include other parameters that characterize the quantum processing unit QPU.
[0038] The quantum computer QC with its quantum processing unit QPU is to be installed at a given location PL, e.g. a data center, which is exposed to certain environmental influences. The performance of the quantum processing unit QPU of the quantum computer QC is highly dependent on the environmental conditions at the planned location PL. For example, the performance of the quantum computer QC and its quantum processing unit QPU can react particularly to certain types of external noise such as vibrations, electromagnetic fields, temperature fluctuations or cosmic radiation. The effects on the performance of the quantum computer depend heavily on the technology used by the quantum processing unit QPU, i.e. whether the quantum processing unit QPU is based on superconducting qubits, ion traps or integrated photon circuits, etc.
[0039] In order to be able to predict the performance of the quantum computer QC at the planned location PL, the computer-implemented method described below is carried out. Environmental parameters EPI, EP2 are recorded at the planned location PL. The environmental parameters EPI, EP2 can be of the same or different types. The environmental parameters EPI, EP2 are recorded using sensors ES1, ES2, which are designed to record, for example, vibrations, electromagnetic fields, temperatures or radiation, etc. The number of sensors used to record environmental parameters can be greater or smaller than the number 2 shown here. The environmental parameters EPI, EP2 represent digital values of the metrologically recorded environmental values.
[0040] The environmental parameters EPI, EP2 are transmitted via a suitable communication connection to a processor PR, which implements a trained data-driven model MO and receives the digital environmental parameters EPI, EP2 as input information and provides at least one state parameter CPI, CP2 as digital output information. Each state parameter CPI, CP2 describes a performance of the at least one quantum processing unit QPU.
[0041] The trained data-driven model MO described in this paper is based, for example, on a neural network that was previously trained using training data. The training data includes a variety of performance parameters of the quantum processing unit QPU along with information about environmental parameters EPI, EPI, and state parameters CPI, CP2 that describe performance.
[0042] In the embodiment shown in Fig. 1, the state parameters that describe the performance of the quantum processing unit QPU are output, for example, on a user interface (not shown). The user interface includes, in particular, a display. The user interface provides information for a human user.
[0043] Alternatively or additionally, the output information of the state parameters CPI and CP2 can be used to repeat the determination of the state parameters if a performance-defining measure is undershot. This can, for example, increase the influence of environmental parameters in the case of high error rates.
[0044] Alternatively or additionally, the determination of the
[0045] State parameters CP1, CP2 are repeated for a plurality of predefined locations PLi (where i describes a plurality of different locations PL) if a performance-descriptive measure of at least one state parameter CP1, CP2 is undershot. Based on the plurality of repeated "simulations," the location can then be identified at which the at least one state parameter CPI, CP2 has a measure that describes the best possible performance.
[0046] According to a further expedient embodiment, a self-calibration is carried out for the at least one quantum processing unit QPU if a measure describing the performance of the at least one state parameter CP1, CP2 is undershot.
[0047] It is also expedient, when determining the state parameter(s) CPI, CP2, to repeat the determination of the state parameters, particularly iteratively, using environmental parameters that are varied within a specified range around the respective measured value. This allows, for example, acceptable fluctuation ranges for the respective environmental parameters to be determined.
[0048] The invention thus proposes a digital representation of the quantum computer, which contains a simulation of the basic function of the quantum computer depending on the environmental conditions. The environmental conditions are monitored using a number of sensors. The status of the quantum computer is then determined by combining the simulations, taking into account the measured values of the sensors, with direct measurements of the relevant performance indicators of the quantum computer. For example, the effects of temperatures and / or vibrations on the actual error rates of the qubits can be determined.
[0049] The use of a digital representation of the
[0050] Quantum computers in different environments to be simulated leads to cost savings, as trial and error in setting up the quantum computer in a potentially noisy environment is avoided and recalibration and maintenance by experts can be kept to a minimum.
[0051] The use of a digital representation of a technical entity is a well-understood technique used in industry. It helps to simulate the effects of the environment on the quantum computer and their impact on the results of quantum calculations. This achieves the following advantages:
[0052] An optimal placement of the quantum computer can be found even before installation at a planned location (e.g., data center). The digital representation takes into account individual noise channels present at the planned location and derives their impact on the specific type of quantum computer used. Knowledge of these impacts can help find the best placement for the quantum computer and determine measures to protect the quantum computer, such as magnetic shielding, etc. The same applies in the opposite direction, as effects of the operation of the quantum computer on nearby machines can be predicted, e.g., with regard to electric fields, vibrations due to coolant flow, and the like.
[0053] Using a digital representation of the quantum computer in conjunction with sensors that measure environmental conditions as well as the quantum computer's internal performance indicators enables condition monitoring and can indicate the need for recalibration or maintenance. This information can be processed in a way that is usable even by non-expert users.
[0054] Potential distortions of calculation results due to
[0055] Perturbations can be detected, and their impact on the quality of the results can be estimated. This gives users insight into how much confidence they can place in the results of a quantum computer. Regardless of the grammatical gender of a particular term, this includes people with male, female, or other gender identities.
Claims
Patent claims 1. A method for computer-implemented prediction of the performance of a quantum computer (QC) with at least one quantum processing unit (QPU), wherein the at least one quantum processing unit is characterized by respective performance parameters, wherein for a planned operation of the quantum computer at a predetermined location (PL), the following steps are carried out: i) reading in at least one environmental parameter (EPI, EP2), wherein the at least one environmental parameter was recorded with one or more sensors (ESI, ES2), wherein the sensor or sensors (ESI, ES2) are installed at or in the area of the predetermined location (PL);ii) determining at least one state parameter (CPI, CP2) of the at least one quantum processing unit (QPU) by processing the at least one environmental parameter (EPI, EP2) by a trained data-driven model (MO), wherein the at least one environmental parameter (EPI, EP2) is fed into the trained data-driven model (MO) as digital input information and the trained data-driven model (MO) supplies the at least one state parameter (CPI, CP2) as digital output information, wherein the at least one state parameter (CP1, CP2) describes a performance of the at least one quantum processing unit (QPU); 2. Method according to claim 1, characterized in that the trained data-driven model (MO) is a neural network.
3. Method according to claim 1 or 2, characterized in that the determination of the at least one state parameter (CP1, CP2) of the at least one quantum processing unit (QPU) is repeated if a measure describing the performance is undershot.
4. Method according to one of the preceding claims, characterized in that the determination of the at least one state parameter (CPI, CP2) of the at least one quantum processing unit (QPU) is repeated for a plurality of predetermined locations (PLi) if a measure of the at least one state parameter (CPI, CP2) describing the performance is undershot.
5. Method according to one of the preceding claims, characterized in that a self-calibration is carried out for the at least one quantum processing unit (QPU) of the quantum computer (QC) if a measure of the at least one state parameter (CP1, CP2) describing the performance is undershot.
6. Method according to one of the preceding claims, characterized in that the at least one state parameter (CP1, CP2) comprises an error rate for respective gate operations.
7. Method according to one of the preceding claims, characterized in that the at least one environmental parameter (EPI, EP2) is varied within a predetermined bandwidth around the respective measured value and the at least one state parameter (CPI, CP2) is determined for each variation.
8. Device (10) for computer-implemented prediction of the performance of a quantum computer (QC) with at least one quantum processing unit (QPU), wherein the at least one quantum processing unit is characterized by respective performance parameters (LP), wherein the device comprises a processor (PR) which is configured to carry out the following steps for a planned operation of the quantum computer at a predetermined location (PL): i) reading in at least one environmental parameter (EPI, EP2), wherein the at least one environmental parameter was recorded with one or more sensors (ESI, ES2), wherein the or the sensors (ESI, ES2) are installed at or in the area of the predetermined location (PL); ii) determining at least one state parameter (CPI, CP2) of the at least one quantum processing unit (QPU) by processing the at least one environmental parameter (EPI, EP2) by a trained data-driven model (MO), wherein the at least one environmental parameter (EPI, EP2) is fed into the trained data-driven model (MO) as digital input information and the trained data-driven model (MO) supplies the at least one state parameter (CPI, CP2) as digital output information, wherein the at least one state parameter (CPI, CP2) describes a performance of the at least one quantum processing unit (QPU).
9. The apparatus of claim 8, wherein the apparatus is configured to perform a method according to any one of claims 2 to 7.
10. A computer program product comprising program code stored on a machine-readable medium for carrying out a method according to any one of claims 1 to 7 when the program code is executed on a computer.