Patent Application
20200380397
This application claims the benefit of French Patent Application No. 1873616 filed Dec. 20, 2018, the disclosure of which is herein incorporated by reference in its entirety.
The invention falls within the field of methods for analyzing simulations of the execution of a quantum circuit.
For such an analysis, it is known to repetitively perform a very large number of simulations, in order to obtain the values of all possible quantum states.
It is also known to take intermediate measurements when simulating the execution of the quantum circuit.
These known solutions do not enable a good compromise between on the one hand simplifying the analysis of the execution of a quantum circuit and on the other hand completing the analysis performed.
The invention at least partially remedies the shortcomings of the prior art and improves the compromise mentioned above.
It involves being able to analyze the execution of a quantum circuit, particularly for the purposes of debugging that quantum circuit.
To do so, according to the invention, it will be beneficial both to be able to track the execution of that quantum circuit step-by-step and also to extract in a simple and complete manner, one which is transparent to the analyst of the execution of that quantum circuit, all or some of the values of quantum state vectors for certain qubits of particular interest to that analyst, all in order to facilitate both the focusing on the part of the execution of that quantum circuit considered to be particularly interesting, while thereafter facilitating the extraction of data related to that part of the execution, the analysis of the execution of the quantum circuit while preserving the completeness of that analysis, i.e. without losing or sacrificing any or all of the quantum states considered interesting by the analyst, which could damage the completeness of the analysis and/or the running of the analysis, potentially leading to a substantially less effective debugging of the execution of that quantum circuit.
To achieve this good compromise between substantially simplifying the analysis of the execution of that quantum circuit and ensuring the completeness of that analysis, the invention proposes a particular combination of a post-selection and a quantum iteration, both performed on a simulated execution rather than an actual execution.
The invention enables the analyst of the execution of the quantum circuit to easily and effectively focus on the part of the execution of that quantum circuit that interests him or her, all while making the extraction of the interesting data obtained simpler or more transparent to him or her, hence providing a precise focus on what interests him or her, with a simpler way to extract only what interests him or her.
Likewise, the great simplification achieved through the use of a quantum iterator enables the user to carry out more numerous and more detailed post-selection operations, without making the time spent or the difficulty of analysis prohibitive.
A method is proposed for analyzing a simulation of the execution of a quantum circuit, comprising:
According to another aspect, a method is proposed for analyzing a simulation of the execution of a quantum circuit, comprising:
The features described in the following paragraphs may optionally be implemented. They may be implemented independently of one another, or in combination with one another:
Preferably, the method comprises, between the post-selection step and the analysis step:
To achieve this good compromise between substantially simplifying the analysis of the execution of that quantum circuit and that analysis being very useful, the invention proposes a particular combination of a post-selection and a setting filtration, both performed on a simulated execution rather than an actual execution.
The method therefore enables the analyst of the execution of the quantum circuit to focus on a part of the execution of the quantum circuit that interests him or her and on the type of process for that execution part, increasing the options to debug the quantum circuit.
Preferably, the method comprises, between the post-selection step and the analysis step, and before the step of retrieval:
This makes it possible to retrieve a quantum state vector that has been filtered beforehand. The analysis of the quantum state vector may thereby take place immediately after that quantum state vector has been retrieved, without requiring any additional operations on that quantum state vector.
Preferably, the method comprises:
The iterator therefore makes it possible to extract additional information about the quantum state vector, in a simple and unsupervised manner. This makes it possible to further improve the completeness of the analysis of the execution.
Preferably, the step of post-selecting one or more particular values of one or more qubits is carried out at the output of the measurement of said qubit(s).
As the intermediate measurements or the final measurement are an integral part of the execution, the post-selection step may advantageously be performed after one of those measurements. This makes it possible to analyze the simulation step-by-step.
Preferably, the step of post-selecting one or more particular values of one or more qubits is carried out after a quantum gate operation on said qubit(s).
Thus, knowing the sequence of operations during execution, the post-selection step may be performed at any time during execution. This enables a complete, non-degraded analysis of the quantum states deemed interesting at any moment in the simulation.
Preferably, the method comprises:
This enables a very sizable memory gain, especially in the application terminal, whose memory may be fairly limited.
Preferably, the size of the buffer memory, in the application terminal and in the simulator, is large enough for the time taken to process the contents of that buffer memory by the application terminal to be longer than the time taken to transfer the data communicated between the simulator and the application terminal.
This sizable memory gain does not or almost does not come at the cost of any slowdown in processing the data within the application terminal.
Preferably, the size of the buffer memory is between 1 MB and 100 MB, and very preferably between 3 MB and 20 MB.
The memory range is thereby optimized for a compromise between memory gain and preserving processing speed.
Preferably, the step of analysis is followed by a step of debugging the execution of the quantum circuit.
This therefore enables a beneficial application of the analysis step, improving the quality and efficiency of the analyzed quantum circuit.
Preferably, the method comprises, between the post-selection step and the analysis step:
Multiple filtration steps may therefore be performed on the quantum state vector. The filtering of the quantum states of the quantum state vector based on the various parameters makes it possible to further increase the completeness of the analysis, by focusing on the quantum states deemed interesting. The various types of filtration may be combined and/or stacked with one another.
Preferably, the method comprises, after the setting filtration step:
The step of renormalizing advantageously makes it possible to have a coherent filtered quantum state vector, particularly when quantum states of the quantum state vector were removed from the quantum state vector during the filtration steps.
Preferably, the method is integrated into an application programming interface (API) capable of accessing a server including a quantum circuit execution simulator and a module for communicating with a remote application terminal separate from the simulator.
This makes it possible to improve ease of use for a user-analyst from their application terminal.
Preferably, the number of qubits in the quantum circuit is between 2 and 1000, preferably between 10 and 50.
The analysis method of the invention operates even with large quantum circuits, or with very large ones.
Preferably, the analysis method is implemented in python language. The language python is well-suited to this type of method for analyzing the operation of a quantum circuit.
The invention and variants thereof may generally make it possible to offer a method for analyzing a simulation of the execution of a quantum circuit, enabling a complete and simplified analysis, particularly for debugging the execution of the quantum circuit, and in particular for offering a better compromise between on the one hand the simplification of the analysis of the execution of a quantum circuit and on the other hand the completeness of the analysis performed.
Such a solution makes it possible to solve the problems posed by known solutions.
Other features, details, and benefits of the invention will become apparent upon reading the detailed description below and upon analyzing the attached drawings, in which:
Reference is now made to
The system 1 particularly comprises a simulator 2 of a quantum circuit. More specifically, the simulator 2 makes it possible to simulate a quantum processor.
As an output of the simulator 2, a quantum state vector is obtained. The quantum state vector comprises 2n quantum states, n being the number of qubits that the simulation relies upon.
For example, the quantum state vector comprises between 22 and 21000. In other words, the quantum circuit comprises between 2 and 1000 qubits. Preferably, the quantum state vector comprises for example between 210 and 250. In other words, the quantum circuit comprises between 10 and 50 qubits.
The simulator 2 comprises for example a memory 3. The memory 3 may particularly be able to store the quantum state vector obtained. The memory 3 therefore comprises a large storage capacity in order to be able to store a quantum state vector comprising between 22 and 21000 and preferably between 210 and 250 possible quantum states.
The simulator 2 is for example stored on a server 4.
The server 4 comprises for example a communication module 5a, making it possible to enter into communication with an application terminal 6 via a communication module 5b of the application terminal 6.
The application terminal 6 is for example remote from the server 4. The application terminal 6 particularly enables an analyst to access the quantum state vector in order to analyze it, without storing it.
To that end, an iterator 7 may be implemented within the system 1. The iterator 7 enables an interaction between the application terminal 6 and the simulator 2.
More specifically, the iterator 7 makes it possible to avoid duplicating all the information, in other words, the quantum state vector obtained at the output of the simulator 2, to the application terminal 6. In other words, the iterator 7 enables the retrieval of the information at the output of the simulator 2, on the application terminal 6. The iterator 7 therefore enables an analyst to access the information at the output of the simulator 2 in a simple way.
More specifically, the iterator 7 loops onto each quantum state of the quantum state vector obtained at the output of the simulator 2, so that each quantum state can be retrieved on the application terminal.
The iterator 7 may further comprise filters F1, F2, and F3 to be applied to the quantum state vector. Those filters are described in greater detail with reference to
Furthermore, the system 1 may comprise a buffer memory 8a implemented in the simulator 2, and a buffer memory 8b implemented in the application terminal 6.
The information obtained at the output of the simulator 2 is for example transmitted to the application terminal 6, by means of the iterator 7, via the buffer memories 8a, 8b.
The iterator 7 is stored in the simulator 2 and is in communication with the buffer memory 8a.
The use of buffer memories 8a, 8b makes it possible to greatly reduce the memory size needed to access the information obtained at the output of the simulator 2.
The size of the buffer memories 8a, 8b may be chosen so that the time taken to process the content of the buffer memory 8b by the application terminal 6 is greater than the time taken to transfer the information obtained at the output of the simulator 2 to the application terminal 6.
Thus, the size of the buffer memories 8a, 8b may be chosen to be 1 MB and 100 MB, and preferably between 3 MB and 20 MB.
The result is that the size of the buffer memories 8a, 8b is at least 10 times, and preferably at least 100 times smaller than the memory 3 storing the information obtained at the output of the simulator 2.
Thus, at the end of the simulation, a quantum state vector 2n is obtained. The quantum state vector 2n is stored in the memory 3 of the simulator 2. The memory 3 communicates with the iterator 7. More specifically, the memory 3 transmits sections of the quantum state vector 2n to the iterator 7. The iterator 7 is configured to process those sections of the quantum state vector 2n, particularly by means of the filters F1, F2, and F3. The iterator 7 is configured to send the sections of the quantum state vector to the buffer memory 8a of the simulator 2 which transmits them, via the communication interfaces 5a, 5b, to the application terminal 6. The buffer memory 8b of the application terminal 6 receives the sections of the quantum state vector, so that they can be analyzed by an analyst directly on the application terminal. Communication between the server 4 and the application terminal 6 is for example done by radio waves or over the Internet.
According to one embodiment variant, the iterator 7 is stored within the buffer memory 8a of the simulator 2.
During the step of simulation S1, the simulation of the quantum circuit is launched. When simulating the execution of a quantum circuit, the quantum state vector obtained is made up of a plurality of qubits having a quantum state 0 and 1.
With reference to
Thus, when the qubit 2 is measured, a probability p is associated with one of the two quantum states 0 or 1, and a probability p−1 is associated with the other quantum state 0 or 1.
With each measurement, a new probability is associated with each quantum state 0 or 1 of the measured qubits.
Thus, when the qubit n is measured, a probability p is associated with one of the two quantum states 0 or 1 for the first qubit, and a probability 1-p is associated with the other one of the two quantum states 0 or 1. Likewise, for the nth qubit, a probability p is associated with one of the two quantum states 0 or 1 of the nth qubit, and a probability 1-p is associated with the other one of the two quantum states 0 or 1 of the nth qubit.
The probabilities p, p and p are generally different, but may, in some cases, be equal.
During the simulation, intermediate measurements are taken for each qubit. Those intermediate measurements influence the quantum state of the measured qubits, leading to a qubit to be set to a quantum state between 0 and 1.
The simulation therefore makes it possible to obtain an execution tree, illustrated in
At the end of the execution, a plurality of quantum state vectors, comprising 2n quantum states, is obtained.
The triangles symbolize quantum operations that can be carried out on qubits. By quantum operations it is meant for example applying quantum gates to the qubits, which may or may not modify the values of the qubits.
In the post-selection step S2, one or more particular values of one or more qubits are post-selected at one or more steps of the simulation. Particular value means a particular quantum state.
In other words, it is possible to post-select one or more branches of the execution. More specifically, it is possible to post-select a quantum state from one or more qubits at the time of execution.
Post-selection may amount to choosing the values of the quantum states of the quantum state vector.
For a given simulation, the post-selections of values of qubits may be successive. The post-selections may also be repetitive. For example, it is possible to post-select the same qubit value several times. For example, it is possible to post-select different values for the same qubit several times. For example, it is possible to select different values of different qubits several times.
In the remainder of the description, the term post-selection is used, although multiple post-selections may be performed, as described above.
Post-selection is performed at multiple steps of the simulation. For example, post-selection is performed at the output of the measurement of the qubits.
Post-selection is performed following a quantum gate operation on the qubit.
Next, a setting filtration step S3 is carried out. The setting filtration step sets the value of one or more quantum states of the quantum state vector(s) derived from post-selection(s) of qubits.
The terms quantum state vector derived from post-selection(s) of qubits and quantum state vector of the post-selected qubit(s) are used interchangeably in the remainder of the description.
The setting filtration step S3 is for example performed by the iterator 7, and particularly by applying the filter F1 depicted in
The setting filtration step S3 is for example located between the step of post-selection S2 and the step of retrieval S5, described below.
The setting filtration step S3 particularly comprises the setting of quantum state values of the quantum state vector derived from post-selection.
The setting filtration step particularly enables the iterator 7 to not need to browse the entire quantum state vector derived from post-selection of qubits when it is retrieved on the application terminal 6.
More specifically, if, out of one quantum state vector derived from post-selection of qubits of n quantum states, three quantum states are set during the setting filtration step, a vector of size 2n-3 is retrieved by means of the iterator 7, rather than a quantum vector of size 2n as before.
The setting filtration step therefore further makes it possible to reduce the size of the information to be transmitted to the application terminal.
According to one embodiment, an additional filtration step S4 is performed.
The additional filtration step S4 is implemented by the iterator 7.
For example, the additional filtration step S4 may comprise threshold filtration and/or value filtration of the quantum states of the quantum state vector derived from the post-selection of qubits.
Threshold filtration particularly comprises the defining of a probability threshold for which the quantum states associated with a probability that fall below that threshold are not retained.
Threshold filtration is performed by applying the filter F2 illustrated in
Value filtration particularly comprises the defining of a quantum state value for which the quantum states of the quantum state vector derived from post-selection of qubits having that value are not retrieved.
Value filtration is performed by applying the filter F3 illustrated in
This additional step of filtration S4 makes it possible, on the one hand, to reduce the size of the information to be sent to the application terminal 6 by eliminating certain quantum states of the quantum state vector derived from post-selecting qubits, and, on the other hand, to increase the completeness of the analysis by focusing only on the quantum states deemed interesting.
The filtration types may be combined and/or stacked for multiple quantum states of the quantum state vector.
According to one embodiment, the steps of setting filtration S3 and additional filtration S4 may be followed by a step of renormalizing the filtered quantum state vector, during which the probability of each quantum state saved in the filtered quantum state vector is modified so that the sum of the probabilities of each quantum state is equal to 1.
Next, a step of retrieval S5, by the iterator 7, of all or some of the quantum states of the quantum state vector(s) derived from the post-selections of qubits and filtered is carried out.
The step of retrieval S5 may also comprise the retrieval, by the iterator 7, of all or some of the quantum states of the quantum state vector(s) derived from the post-selection of qubits and derived from different types of filtration, described with reference to the steps of setting filtration S3 and additional filtration S4.
The step of retrieval S5 may further comprise the retrieval of each of the probabilities associated with each quantum state of the retrieved quantum state vector, whether renormalized or not.
In particular, the iterator 7 is able to identify a parameter representative of the probability of each quantum state of the quantum state vector, between the step of post-selection and the step of retrieval, for example before the step of setting filtration.
The iterator 7 is for example able to read the quantum state vector, and to associate a probability with each quantum state. The iterator is further able, based on a complex model of the matrix stabilizer type, to recalculate the elements of the matrix in order to put them into the form of a readable quantum state vector and to associate a probability with each quantum state of that quantum state vector.
The iterator 7 is configured to iterate on each of the quantum states of the quantum state vector obtained after step S4, and potentially on the probabilities associated with them, so that an analyst can access it from the application terminal 6 for a step of analysis S6.
The step of analysis S6 comprises for example the comparison of the quantum states of the quantum state vector retrieved with the quantum states of the quantum state vector expected following the steps of post-selection S2 and setting filtration S3.
In the step of debugging S7, the data from the analysis may be used for the purposes of debugging the execution of the quantum circuit.
The method is implemented on an application programming interface (or API) that can access the server 4.
In one variant, the threshold filtration particularly comprises the defining of a probability threshold for which the quantum states associated with a probability that fall above that threshold are not retained.
In one variant, the value filtration particularly comprises the defining of a quantum state value for which only the quantum states of the quantum state vector derived from post-selection of qubits having that value are retrieved.
The invention may have application particularly in the field of quantum computing and debugging the execution of quantum circuits.
The invention is not limited to the examples described above, which are only given as examples, but rather encompasses all variants that a person skilled in the art may envision within the scope of the protection sought.
| Number | Date | Country | Kind |
|---|---|---|---|
| 1873616 | Dec 2018 | FR | national |
