A method for quantum flip warning and threshold setting of qubits in a quantum computer

Abstract

The present invention discloses a method for quantum bit quantum flip warning and threshold setting of a quantum computer, which sets an initial threshold for quantum bit flip of the quantum computer based on a quantum flip prediction model; uses a threshold prediction model to dynamically set the quantum flip threshold to achieve quantum flip warning and replacement of the initial threshold; when the threshold prediction model has used up all the quantum flip thresholds in the historical database that conform to the thresholds predicted by the quantum flip prediction model, a threshold adjustment model is constructed to generate a new dynamic threshold to replace the current threshold. The present invention uses a Bayesian threshold prediction model to adjust the monitoring threshold to solve the problem that the method of steady-state distribution of the quantum flip prediction model based on the Markov chain cannot perform calculations on the continuous-time Markov chain, resulting in prediction distortion of the network warning model.

Description

Technical Field

[0001] The present invention belongs to the technical field of quantum computers, and particularly relates to a method for warning of qubit quantum flipping and threshold setting in a quantum computer. Background Art

[0002] With the continuous development of the big data field, more and more concepts have been proposed and applied to production. Quantum computers can be used for the search of a large amount of data. Suppose a traditional computer needs to execute tens of millions of instructions to complete a certain big data search, and the efficiency is extremely low. While a quantum computer often only needs to execute thousands of instructions to complete the same task, thus bringing a brand-new revolutionary change to the big data industry. A quantum computer is a machine that can perform quantum computing. It realizes mathematical and logical operations, processes and stores information through the laws of quantum mechanics. It uses quantum states as memory units and information storage forms, and quantum communication and quantum computing based on quantum dynamics evolution as the basis for information transfer and processing. In a quantum computer, the sizes of various components of its hardware reach the atomic or molecular level. A quantum computer is a physical system that can store and process information represented by qubits.

[0003] So far, quantum systems do not interact with the outside world in any unnecessary way, that is, they almost only deal with the dynamics of closed quantum systems. Although fascinating conclusions can be drawn for the information processing tasks that can be realized in principle in this ideal system, this observation is affected by the fact that there is no completely closed system in the real world, except for the entire universe. Real systems suffer from unnecessary interactions with the outside world. These unnecessary interactions appear as noise in quantum information processing systems. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for warning of qubit quantum flipping and threshold setting in a quantum computer in view of the above-mentioned deficiencies of the prior art. The qubits of the quantum computer are on a disk memory. The qubit starts from state 0 or 1, but after a long time, scattered magnetic fields are very likely to cause the qubit to be disturbed and may flip its state. The Markov chain model is used to correct the qubit, that is, to correct the state after flipping according to the state before the previous flip.

[0005] To achieve the above technical purpose, the technical solution adopted by the present invention is as follows:

[0006] A method for warning of qubit quantum flipping and threshold setting in a quantum computer, comprising:

[0007] Step 1, set the initial threshold for qubit flipping of the quantum computer based on a quantum flipping prediction model;

[0008] Step 2: Dynamically set the quantum flip threshold using a threshold prediction model to achieve quantum flip warning and initial threshold replacement;

[0009] Step 3: After the threshold prediction model has used all the quantum flip thresholds in the historical database that match the thresholds predicted by the quantum flip prediction model, construct a threshold adjustment model to generate a new dynamic threshold to replace the current threshold.

[0010] Specific measures taken to optimize the above technical solution also include:

[0011] The above Step 1 includes:

[0012] First, create multiple test bits for the target bit, and each test bit has the superposition characteristics and pre-inversion state of the target bit;

[0013] Second, construct a quantum flip prediction model, use the qubit flips of different numbers of test bits for flip prediction respectively, and use the stored historical bit flip data for quantum flip prediction model analysis to predict the flip of the qubit and obtain the prediction probability;

[0014] According to the prediction probability, determine whether the target bit flips or not, perform weighted averaging on multiple groups of data probability values, and obtain a better flip probability as the threshold for qubit flip, which is the initial threshold for qubit flip of the quantum computer.

[0015] The formula of the above quantum flip prediction model is: X(k + 1) = X(k) × P

[0016] In the formula: X(k) represents the state vector of the trend analysis and prediction object at time t = k, P represents the one-step transition probability matrix, and X(k + 1) represents the state vector of the trend analysis and prediction object at time t = k + 1.

[0017] The above Step 2 uses Bayesian method to construct a threshold prediction model.

[0018] In the above Step 2, each threshold that has ever successfully flipped in the historical database quantum flip threshold except the one used by the current qubit is put into the Bayesian threshold prediction model one by one to obtain the threshold usage probability. The threshold with the largest value is used as the steady-state distribution of the Markov chain network warning model, and then replaces the initial threshold in use to complete the continuous operation of the threshold prediction model.

[0019] The above Step 3 uses ridge regression method to construct a threshold adjustment model.

[0020] The above-mentioned generation of a new dynamic threshold to replace the current threshold in Step 3 is specifically:

[0021] By conducting simulation flip tests on ten groups of qubits with different quantities, setting the initial threshold for each group to fluctuate by 10% above and below the current threshold, forming a 20% upper and lower fluctuation threshold range, and dividing it into 10 parts with a 2% segmentation;

[0022] Taking the condition of adding or subtracting 2% of the current threshold for each part to obtain historical data, combining it with the threshold adjustment model for calculation to get the exercise fitting value, and then performing a difference calculation with the actual fitting value obtained by combining the data of the last qubit flip in history with the threshold adjustment model. Mark the threshold of the group of data where the difference between the exercise fitting value and the actual fitting value is greater than 10%;

[0023] Successively conduct model training and difference comparison on the 10 groups of different threshold data, mark the thresholds, and generate a new dynamic threshold by weighted averaging multiple marked thresholds to replace the current threshold.

[0024] The present invention has the following beneficial effects:

[0025] The present invention highlights the status of artificial intelligence in the threshold setting of quantum computers, and creatively adopts the Bayesian threshold prediction model to adjust the monitoring threshold to solve the problem that the method of the steady-state distribution of the quantum flip prediction model based on the Markov chain cannot perform calculations on continuous-time Markov chains, resulting in prediction distortion of the network warning model.

[0026] The Markov chain and the Bayesian network are used to comprehensively analyze and predict the qubit flip from both vertical and horizontal aspects, innovatively dealing with the deficiency of the Markov chain in handling the lack of upper-layer indicators. The Markov chain is a method for exploring the probability distribution of variables in future time determined by samples, which is a longitudinal prediction method. The Bayesian qubit flip threshold shows the mutual influence relationship between flip prediction and threshold indicators, which is a horizontal prediction method. Combining these two methods has the advantage of solving the problem of the lack of non-bottom-layer indicator data in the multi-layer indicator system to achieve macroscopic qubit flip prediction. The inverse deduction function of the Bayesian qubit flip threshold also provides a basis for network fault risk control.

[0027] In addition, when the threshold predicted by the inverse deduction function of the Bayesian qubit flip threshold reaches a situation where there is no available threshold, to prevent the occurrence of a steady state. A threshold fluctuation range is formed, and according to the ridge regression method, a threshold adjustment model training result is used to obtain the weighted average of the test fitting value and the actual fitting value to generate a new dynamic threshold as the backup threshold when there is no available threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is the schematic diagram of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0029] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0030] As Figure 1 shown, a method for quantum flip warning and threshold setting of qubits in a quantum computer includes:

[0031] Step 1: Set the initial threshold for qubit flipping of the quantum computer based on the quantum flip prediction model;

[0032] First, create multiple test qubits for the target qubit. Each test qubit has the superposition characteristics and pre-inversion state of the target qubit. Secondly, construct the [quantum flip prediction model], and use the qubit quantum of different numbers of test qubits for flip prediction respectively, and use the stored historical qubit flip data for [quantum flip prediction model] analysis to predict the qubit flip. According to the predicted probability, obtain the final state of the target qubit, that is, whether it is likely to flip or not. Weight the average of the test probability values of multiple groups of data to obtain a better flip probability as the threshold for qubit quantum flipping, as the initial threshold for qubit flipping of the quantum computer.

[0033] Quantum flip prediction model formula: X(k + 1) = X(k) × P

[0034] In the formula: X(k) represents the state vector of the trend analysis and prediction object at time t = k, P represents the one-step transition probability matrix, and X(k + 1) represents the state vector of the trend analysis and prediction object at time t = k + 1.

[0035] Initial transition probability of historical qubit flipping [0.3, 0.7]

[0036] State transition probability of current qubit flipping from 0 to 1 [0.6, 0.4]

[0037] State transition probability of current qubit flipping from 1 to 0 [0.3, 0.7]

[0038] The rectangular set is:

[0039] Calculated through the model: X(k + 1) = X(k) × P

[0040] Probability of qubit flipping from 0 to 1 with the next period threshold unchanged: 0.3×0.6 + 0.3×0.7 = 0.39

[0041] Probability of qubit flipping from 1 to 0 with the next period threshold unchanged: 0.3×0.4 + 0.7×0.7 = 0.61

[0042] Probability of qubit flipping with the next period threshold unchanged [0.39 0.61]

[0043] Step 2: Dynamically set the quantum flip threshold using a threshold prediction model to achieve quantum flip early warning and initial threshold replacement;

[0044] Regarding the distortion of the flip prediction probability caused by the stable release of the

Quantum Flip Prediction Model

[0045] Construct a Bayesian

Threshold Prediction Model

[0046] First, access the historical alarm database for combined analysis to obtain the model parameters

prior probability

conditional probability

adjustment factor

[0047] Secondly, put

prior probability

conditional probability

adjustment factor

Quantum Flip Prediction Model

Threshold Prediction Model

[0048] Model formula and example:

[0049] P(A|B) = (P(B|A) * P(A)) / (P(B|A)P(A) + P(B|A')P(A'))

[0050]

Prior probability

Conditional probability

Adjustment factor

[0051] P(A) is the total number of faults using the current threshold ignoring other factors / the total number of historical faults. For example: 40%;

[0052] P(A') = 1 - P(A), which is 60% here;

[0053] P(B|A) is the probability of the result of the total number of times the current threshold has been used during the continuous learning process of the

Network Early Warning Model

[0054] P(B|A') is the probability that the threshold in the threshold database has appeared in the historical fault database. If all historical thresholds have been applied in the historical fault database, it is 100% here;

[0055] P(B) is to directly consider the threshold usage probability formula ignoring other factors

[0056] P(B) = P(B|A)P(A) + P(B|A')P(A'), which is 0.5 * 0.4 + 1 * 0.6 = 0.8 here;

[0057] Then according to Bayes' formula, it can be calculated as follows, that is

[0058] P(A|B) = (0.5 * 0.4) / (0.8) = 0.25

[0059] Put each threshold that has ever been successfully flipped in the historical database quantum flip threshold except for the currently used qubit into the Bayesian [threshold prediction model] one by one to obtain the threshold usage probability. The threshold with the largest value is used as the stationary state distribution of the Markov chain [network warning model]. After that, the calculation of the continuous-time Markov chain cannot be realized. Thus, it replaces the initial threshold in use. Complete the continuous operation of the [threshold prediction model].

[0060] Step 3: When the threshold prediction model has used all the thresholds in the historical database quantum flip threshold that match the prediction of the quantum flip prediction model, construct a threshold adjustment model to generate a new dynamic threshold to replace the current threshold.

[0061] Use the ridge regression method to construct the [threshold adjustment model]. Through the simulation flip test of ten groups of qubits with different numbers, set the initial threshold of each group to fluctuate by 10% above and below the current threshold, forming a 20% upper and lower fluctuation threshold interval, and divide it into 10 parts with 2%. Take the historical data with the condition that each part is the current threshold plus or minus 2% of the threshold, combine it with the [threshold adjustment model] operation to obtain the exercise fitting value, and then calculate the difference with the actual fitting value obtained by combining the data of the last quantum bit flip in history with the [threshold adjustment model]. If the difference between the exercise fitting value and the actual fitting value of a certain group is greater than 10%, it means that the threshold fluctuation of that group of data is more likely. Mark the threshold of this group of data, and do not operate otherwise. Perform model training and difference comparison on the threshold data of 10 groups in turn, and mark the threshold. Generate a new dynamic threshold by weighted averaging multiple marked thresholds to replace the current threshold.

[0062] Formula of the [threshold adjustment model]: ||Xθ - y|| 2 + ||Γθ|| 2

[0063] The formula for preventing overfitting is: θ(a) = (X T X + aI) -1 X T y

[0064] Among them, X represents the input; y represents the predicted result of the output; || represents the regularization operation; I represents the identity matrix; θ is the fitting hyperparameter; Γ is the weight constant; a is the weight of the identity matrix; θ(a) represents finding the value of θ under the condition that a is determined.

[0065] Examples of the quantum flip process are as follows:

[0066] The state of a single qubit that is 30% 0 and 70% 1 can be spread across three qubits. In this way, as a group, these qubits are in a state where 30% of all three qubits are 0 and 70% are 1. This larger but equivalent quantum state helps researchers correct errors.

[0067] These qubits can be connected through two gates in a quantum circuit. One gate checks the "parity" of the first and second physical bits (note: here, parity means whether the two bits are in the same state. For example, for 00 and 11, the parity is the same, while for 01 and 10, it is different) - whether they are the same or different - and the other gate checks the "parity" of the first and third physical bits. When no error occurs (i.e., the qubits are in the superposition state |000> + |111>), the parity-measuring gates can indicate that the parity of the first-second and first-third bits are both the same. However, if the first physical bit accidentally flips, causing the system state to become |100> + |011>, these two gates will detect that the parity of both pairs of qubits is different. If the second qubit flips, resulting in the system state becoming |010> + |101>, the parity-measuring gates will find that the parity of the first-second bits is different, but the parity of the first-third bits is the same. Similarly, if the third qubit flips, the detection result will be that the first-second bits are the same, and the first-third bits are different. Such unique results tell what corrective measures should be taken, if necessary, to flip back the first, second, or third qubit without the entire logical qubit collapsing.

[0068] The above is only the preferred embodiment of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A method for quantum flip warning and threshold setting of qubits in a quantum computer, characterized in that, Including: Step 1: Set the initial threshold for qubit flipping of a quantum computer based on a quantum flipping prediction model; Step 2: Dynamically set the quantum flipping threshold using a threshold prediction model to achieve quantum flipping early warning and replacement of the initial threshold; In Step 2, a Bayesian method is used to construct the threshold prediction model. In Step 2, the thresholds of each qubit that has ever been successfully flipped except for the currently used qubit in the historical database of quantum flipping thresholds are put into the Bayesian threshold prediction model one by one to obtain the threshold usage probability. The threshold with the largest value is used as the steady-state distribution of the Markov chain network early warning model and then replaces the initial threshold in use to complete the continuous operation of the threshold prediction model; Step 3: When the threshold prediction model has used all the thresholds in the historical database of quantum flipping thresholds that conform to the prediction of the quantum flipping prediction model, construct a threshold adjustment model to generate a new dynamic threshold to replace the current threshold; In Step 3, a ridge regression method is used to construct the threshold adjustment model.

2. The method for quantum flip warning and threshold setting of qubits in a quantum computer according to claim 1, characterized in that, Step 1 includes: First, create multiple test qubits for the target qubit, and each test qubit has the superposition characteristics and pre-inversion state of the target qubit; Second, construct a quantum flipping prediction model, use the qubit quantum of different numbers of test qubits for flipping prediction respectively, and use the stored historical qubit flipping data for analysis of the quantum flipping prediction model to perform flipping prediction on the qubit and obtain the prediction probability; According to the prediction probability, determine whether the target qubit flips or not, perform weighted averaging on multiple groups of data probability values, and obtain a better flipping probability as the threshold for qubit flipping of the quantum computer, that is, the initial threshold for qubit flipping of the quantum computer.

3. A method for quantum flip warning and threshold setting of qubits in a quantum computer according to claim 1, characterized in that, The formula of the quantum flipping prediction model is: X(k + 1)=X(k)×P In the formula: X(k) represents the state vector of the trend analysis and prediction object at time t = k, P represents the one-step transition probability matrix, and X(k + 1) represents the state vector of the trend analysis and prediction object at time t = k + 1.

4. A method for quantum flip warning and threshold setting of qubits in a quantum computer according to claim 1, characterized in that, The generation of a new dynamic threshold to replace the current threshold in Step 3 is specifically: Through simulation flipping tests on ten groups of qubits with different numbers, the initial threshold is set to fluctuate by 10% above and below the current threshold for each group, forming a 20% upper and lower fluctuation threshold interval, and dividing it into 10 parts with 2%; Obtain historical data based on the condition that each part is the current threshold plus or minus 2% of the threshold, combine it with the threshold adjustment model for operation to obtain the exercise fitting value, and then perform a difference calculation with the actual fitting value obtained by combining the threshold adjustment model with the data of the most recent qubit flipping in history. Mark the threshold of the group of data where the difference between the exercise fitting value and the actual fitting value is greater than 10%; Perform model training and difference comparison on the ten groups of different threshold data in turn, mark the thresholds, and generate a new dynamic threshold by weighted averaging multiple marked thresholds to replace the current threshold.

Images