Method and apparatus for preprocessing input data of quantum computing circuit

Abstract

Provided are a method and an apparatus for preprocessing input data of a quantum computing circuit. A quantum computing system to which a preprocessing apparatus is applied according to an embodiment of the present invention comprises: the preprocessing apparatus for normalizing input data by using an exponential function, a normalization characteristic of which is adjusted by an exponential parameter; and a quantum computing circuit for performing a quantum computing operation on the basis of the input data normalized by the preprocessing apparatus. Accordingly, adaptive normalization for input data is possible through parameter adjustment according to the application of the quantum computing circuit, thereby improving the application performance of the quantum computing circuit.

Description

Method and device for preprocessing input data of a quantum computing circuit

[0001] This patent application is related to the 2022 Ministry of Trade, Industry and Energy's Energy Technology Development Project, the SoC Development Project for Low-Power Energy Harvesting Smart Sensors, and the 2022 Korea Electronics Technology Institute's Basic Research Project, the Quantum Computing SoC Architecture and S / W Platform Solution Technology Development Project.

[0002] The present invention relates to quantum computing technology, and more particularly, to a preprocessing method for normalizing data input to a quantum computing circuit and a quantum computing system applying the same.

[0003] Preprocessing input data is essential for algorithms, functions, and systems to perform their functions. Classical computers allow for the free implementation of preprocessing, including a variety of linear and nonlinear operations.

[0004] However, the input of the quantum circuit used in the quantum computer has a constraint that the sum of its squares must be 1, so normalization using RMS (Root Mean Square) is essential during preprocessing.

[0005] Meanwhile, for edge detection, there are various preprocessing methods such as BLUR, MIN-MAX SCALER, or HISTOGRAM EQUAIZER, and these can improve edge detection performance. However, it is very difficult and impossible to apply preprocessing methods to edge detection in quantum computers, and due to the large number of qubits and the resulting noise, a simplified QHED (Quantum Hadamard Edge Detection) algorithm is used to tile a large image into small image pieces and perform edge detection with HADAMARD + ROLLING CIRCUIT.

[0006] However, although adaptive edge detection according to location is possible by performing normalization of tiled images, there is a problem in that detection performance is greatly reduced when the edge boundary is not clear due to the nature of distance-based normalization.

[0007] The present invention has been devised to solve the above problems, and the purpose of the present invention is to provide a method and device capable of adaptive normalization of input data according to the application of a quantum computing circuit and capable of demonstrating better application performance in conjunction with a preprocessing gate within a quantum computing circuit.

[0008] A quantum computing system according to one embodiment of the present invention for achieving the above object includes a preprocessing device for normalizing input data using an exponential function whose normalization characteristic is adjusted by an exponent parameter; and a quantum computing circuit for performing quantum computing operations based on the input data normalized by the preprocessing device.

[0009] The preprocessing device is,

[0010] A first normalization module that normalizes input data according to the following formula;

[0011] Normalized data =

[0012]

[0013] Xi is the input data, and T can be a tunable parameter.

[0014] The preprocessing unit may further include a second normalization module that normalizes the output of the first normalization module to RMS (Root Mean Square).

[0015] A quantum computing circuit may include an input embedding module that embeds input data normalized by a preprocessing device into a quantum state vector; a preprocessing module that preprocesses a state vector passed through the input embedding module; an application circuit that performs a quantum computing operation on the state vector preprocessed by the preprocessing module; and an observation module that observes a state vector on which an operation is completed in the application circuit and outputs a result.

[0016] The preprocessing module is

[0017] The amplitude of the state vector is transformed by applying the transformation matrix given in the following formula,

[0018]

[0019] T may be a parameter for scaling.

[0020] The quantum computing system according to the present invention may further include a post-processing device that post-processes the computational results by the quantum computing circuit.

[0021] The preprocessing unit and the postprocessing unit can execute a program implemented on a classical computer as a processor, and the quantum computing circuit can execute a program for simulating a quantum computer implemented on a classical computer as a processor.

[0022] The input data is image data, and the application circuit can detect edges through quantum computing operations on the state vector for the image data preprocessed by the preprocessing module.

[0023] The preprocessing unit can simultaneously adjust the blur and sharpness of the image by adjusting T.

[0024] According to another aspect of the present invention, a quantum computing method is provided, characterized in that the method includes: a step of normalizing input data using an exponential function whose normalization characteristic is adjusted by a parameter of an exponent by a preprocessing device; and a step of performing a quantum computing operation by a quantum computing circuit based on the input data normalized by the preprocessing device.

[0025] According to another aspect of the present invention, a preprocessing device for a quantum computing circuit is provided, characterized in that it includes a first normalization module for normalizing input data using an exponential function whose normalization characteristic is adjusted by a parameter of the exponential; and a second normalization module for normalizing the output of the first normalization module to RMS (Root Mean Square) and inputting the normalized output to a quantum computing circuit.

[0026] According to another aspect of the present invention, a preprocessing method for a quantum computing circuit is provided, comprising: a first normalization step of normalizing input data using an exponential function whose normalization characteristic is adjusted by a parameter of the exponential; and a second normalization step of normalizing the output of the first normalization step to RMS (Root Mean Square) and inputting the normalized output to a quantum computing circuit.

[0027] As described above, according to embodiments of the present invention, adaptive normalization of input data is possible through parameter adjustment according to the application of the quantum computing circuit, thereby improving the application performance of the quantum computing circuit.

[0028] Additionally, according to embodiments of the present invention, the preprocessing device can further improve the application performance of the quantum computing circuit by linking with a preprocessing gate that performs adaptive amplification within the quantum computing circuit.

[0029] Figure 1 is a diagram illustrating a quantum computing system to which the present invention can be applied.

[0030] FIG. 2 is a diagram illustrating the configuration of a quantum computing system to which an input data preprocessing device according to one embodiment of the present invention is applied.

[0031] Figure 3 shows the normalization characteristics of Soft-rms according to the change in T value.

[0032] Figure 4 shows the normalization characteristics of Soft-rms and RMS.

[0033] Figures 5 and 6 are diagrams showing the change in amplitude of the state vector according to the change in the T value.

[0034] FIG. 7 is a diagram illustrating the configuration of a quantum computing-based edge detection system to which an input data preprocessing device according to another embodiment of the present invention is applied, and

[0035] Figure 8 shows the simulation results for verifying edge detection performance.

[0036] Hereinafter, the present invention will be described in more detail with reference to the drawings.

[0037] An embodiment of the present invention presents a method and device for preprocessing input data for a quantum computing circuit. This technology normalizes input data for a quantum computing circuit using a novel technique and links it to qubit gates for preprocessing after input to the quantum computing circuit.

[0038] Figure 1 illustrates a quantum computing system to which the present invention can be applied. As illustrated, the quantum computing system to which the present invention can be applied comprises an RMS module (10) and a quantum computing circuit (20).

[0039] The RMS module (10) normalizes input data (X) to RMS and inputs it to a quantum computing circuit (120). The quantum computing circuit (20) embeds the normalized input data into a quantum state vector, performs a quantum computing operation, observes the execution result, and outputs the result (O).

[0040] FIG. 2 is a diagram illustrating the configuration of a quantum computing system to which an input data preprocessing device according to one embodiment of the present invention is applied. As illustrated, the quantum computing system according to the embodiment of the present invention is configured to include a preprocessing device (110) and a quantum computing circuit (120).

[0041] The preprocessing device (110) is a device for performing preprocessing on data input to a quantum computing circuit (120), and is configured to include a soft-rms module (111) and an RMS module (112). The soft-rms module (111) and the RMS module (112) can be implemented as hardware, but can also be implemented as a program on a classical computer (a general computer, not a quantum computer) and executed through a processor.

[0042] The soft-rms module (111) preprocesses the input data (X) by normalizing it according to Equation 1 below.

[0043] [Formula 1]

[0044]

[0045]

[0046] T is a parameter for adjusting the normalization characteristics with a scaling factor. The normalization characteristics of the Soft-rms module (111) according to the change in the T value are illustrated in Fig. 3. When the input data (X) is image data, it is possible to adjust the blur and sharpness of the image simultaneously by adjusting T.

[0047] The RMS module (112) normalizes the output of the Soft-rms module (111) to RMS and inputs it to the quantum computing circuit (120).

[0048] In normalizing input data, the Soft-rms module (111) and the RMS module (112) have the characteristic of converting large input values ​​to larger values ​​and converting small input values ​​to smaller values, as shown in Fig. 4.

[0049] Meanwhile, the RMS module (112) is not an essential component and can be omitted. In other words, the preprocessing device (110) can be implemented with only the Soft-rms module (111).

[0050] The quantum computing circuit (120) is a device for performing quantum computing operations on data that is normalized and input from the preprocessing device (110), and is configured to include an input embedding module (121), a Soft-U module (122), an application circuit (123), and an observation module (124).

[0051] The constituent modules (121, 122, 123, 14) of the quantum computing circuit (120) can be implemented as hardware, and can also be implemented as a program (simulator) for simulating a quantum computer on a classical computer and executed through a processor.

[0052] The input embedding module (121) embeds the input data normalized in the preprocessing device (110) into a quantum state vector, and the Soft-U module (122) preprocesses by amplifying the state vector transmitted through the input embedding module (121).

[0053] The Soft-U module (122) is a rotation and reverse rotation gate, not a qubit gate used in existing quantum computing circuits, and transforms the input state vector by applying the transformation matrix given in Equation 2 below to the input state vector.

[0054] [Formula 2]

[0055]

[0056]

[0057] When parameters φ, λ in the U gate are set to π and θ is expressed as πT as a parameter, a Soft-U gate presented in the embodiment of the present invention is obtained, and the Soft-U module (122) converts the input state vector by adjusting the amplitude of the input state vector using the above Soft-U gate.

[0058] At this time, the scaling factor T value of the Soft-U gate is adjusted to set the scale of the Soft-U gate. The Soft-U gate is a rotation (counterclockwise) gate, and the rotation angle is determined according to the T value. As shown in FIGS. 5 and 6, the range of the T value is set to a limited value depending on the application. In the examples of FIGS. 5 and 6, it must be set to a value of 0.5 or less.

[0059] If the Soft-U gate is obtained by setting the parameters φ and λ in the U gate to 2π and 0, respectively, the Soft-U gate is a rotation gate in the opposite direction (clockwise) that transforms the input state vector by adjusting the amplitude of the input state vector.

[0060] Accordingly, the Soft-U module (122) adjusts large values ​​to large amplitudes and small values ​​to small amplitudes, similar to the Soft-rms module (111) of the aforementioned preprocessing device (110).

[0061] The application circuit (123) is a module for performing quantum computing operations on a state vector adaptively amplified by the Soft-U module (122). The application circuit (123) is designed to perform various applications such as image processing and voice processing.

[0062] The observation module (124) observes (measures) the state vector in which the operation is completed in the application circuit (123) and outputs the result.

[0063] So far, a preferred embodiment of a quantum computing system with an input data preprocessing device has been described in detail.

[0064] Specifically, a method for improving the performance of an application implemented with a quantum computing circuit (120) by normalizing input data with a Soft-rms module (111) in a classical computer-based preprocessing device (110) before embedding a quantum state vector in a quantum computing circuit (120) is presented.

[0065] In addition, a rotation or reverse rotation gate, which is not a gate used in a conventional quantum computing circuit, is added to the quantum computing circuit (120), so that the application performance of the quantum computing circuit (120) can be further enhanced by combining it in pairs with a soft-rms module (111).

[0066] In the above embodiment, the scaling factor T can be adjusted by the user to a value suitable for the application implemented with the quantum computing circuit (120), and in particular, the sharpness of a rapidly changing section can be adjusted.

[0067] FIG. 7 is a diagram illustrating the configuration of a quantum computing-based edge detection system to which an input data preprocessing device is applied according to another embodiment of the present invention. As illustrated, the edge detection system according to an embodiment of the present invention is configured to include a preprocessing device (210), a quantum computing circuit (220), and a postprocessing device (230). The postprocessing device (230) can be implemented in hardware, but can also be implemented as a program on a classical computer and executed through a processor.

[0068] The preprocessing unit (210) is configured to include a tiling module (211) and a Soft-rms module (212). The tiling module (211) divides a 128×128 input image into 16 32×32 images. The Soft-rms module (212) performs the aforementioned Soft-rms-based normalization on the images divided by the tiling module (211). In the embodiment of the present invention, the scaling factor T of the Soft-rms module (212) is assumed to be 0.8, but of course, the value can be changed according to the specifications and needs.

[0069] The quantum computing circuit (220) is configured by adding Soft-U modules (221, 222) to the existing QHED (Quantum Hadamard Edge Detection) circuit. The Soft-U modules (221, 222) perform preprocessing through adaptive amplification on the state vector passing through the H gate as described above and then transfer the result to the amplitude permutation module.

[0070] The post-processing device (230) is configured to detect edges of an image by combining edge information regarding vertical and horizontal, and includes a NMS (Non-Maximum Suppression) module (231) for setting only the edge with the highest accuracy among the detected edges, and a Hysteresis edge tracking module (232) for removing noise.

[0071] The edge detection system presented in Fig. 7 can improve image edge detection performance by adding a Soft-rms module (212) to a preprocessing device (210) and adding Soft-U modules (221, 222) to a quantum computing circuit (220).

[0072] The performance verification results of the edge detection system according to an embodiment of the present invention are presented in Fig. 8. Fig. 8 shows the image edge detection results by 1) a conventional technique applying RMS normalization to the QHED algorithm, 2) a technique applying soft-RMS normalization to the QHED algorithm, and 3) a technique applying soft-RMS normalization and soft-U to the QHED algorithm for the left image.

[0073] According to Fig. 8, it can be confirmed that the existing technique detected the edges in the center of the image well, but did not detect the outer edges with low contrast ratio well. On the other hand, when soft-RMS was applied, it could be confirmed that the cameraman and the camera stand were well detected through sharpness amplification, but the edges with low contrast ratio in the background were not well detected. On the other hand, when soft-U was applied further than soft-RMS and sharpness was greatly amplified, the cameraman and the camera stand as well as the surrounding background were well detected, and it could be confirmed that the outlines of the camera and the cameraman were well connected.

[0074] Through this, we can confirm that the effect is better when applying normalization by soft-rms than the existing technology, and that the detection effect is further improved when applying amplification preprocessing by soft-U.

[0075] Unlike existing techniques that perform preprocessing using blur, which generally blurs the image, the edge detection system according to an embodiment of the present invention can simultaneously control blur and sharpness using a scaling factor to better detect the boundary, thereby improving detection performance for edges of images that are covered by shadows or are excessively blurry.

[0076] Meanwhile, it goes without saying that the technical idea of ​​the present invention can also be applied to a computer-readable recording medium containing a computer program that performs the functions of the device and method according to the present embodiment. In addition, the technical idea according to various embodiments of the present invention can be implemented in the form of computer-readable code recorded on a computer-readable recording medium. The computer-readable recording medium can be any data storage device that can be read by a computer and store data. For example, the computer-readable recording medium can be a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical disk, a hard disk drive, etc. In addition, the computer-readable code or program stored on the computer-readable recording medium can be transmitted through a network connected between computers.

[0077] In addition, although the preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and various modifications can be made by a person having ordinary skill in the art to which the present invention pertains without departing from the gist of the present invention as claimed in the claims, and such modifications should not be understood individually from the technical idea or prospect of the present invention.

Claims

1. A preprocessing device that normalizes input data using an exponential function whose normalization characteristics are adjusted by the parameters of the exponential function; and A quantum computing system comprising: a quantum computing circuit that performs quantum computing operations based on input data normalized by a preprocessing device.

2. In claim 1, The preprocessing device is, A first normalization module that normalizes input data according to the following formula; Normalized data = A quantum computing system characterized in that Xi is input data and T is a tunable parameter.

3. In claim 2, The preprocessing device is, A quantum computing system further comprising a second normalization module that normalizes the output of the first normalization module to RMS (Root Mean Square).

4. In claim 2, Quantum computing circuits, An input embedding module that embeds input data normalized by a preprocessing unit into a quantum state vector; A preprocessing module that preprocesses the state vector passed through the input embedding module; An application circuit that performs quantum computing operations on a state vector preprocessed by a preprocessing module; A quantum computing system characterized by including an observation module that observes a state vector in which an operation is completed in an application circuit and outputs a result.

5. In claim 4, The preprocessing module is The amplitude of the state vector is transformed by applying the transformation matrix given in the following formula, A quantum computing system characterized in that T is a parameter for scaling.

6. In claim 5, A quantum computing system further comprising a post-processing device for post-processing the results of a calculation by a quantum computing circuit.

7. In claim 6, Pre-processing devices and post-processing devices, Executes a program implemented on a classical computer with a processor, Quantum computing circuits, A quantum computing system characterized by executing a program for simulating a quantum computer implemented on a classical computer on a processor.

8. In claim 5, The input data is, It is image data, The application circuit is, A quantum computing system characterized by detecting edges through quantum computing operations on a state vector of image data preprocessed by a preprocessing module.

9. In claim 6, The preprocessing device is, A quantum computing system characterized by simultaneously adjusting the blur and sharpness of an image by adjusting T.

10. A step in which the preprocessing device normalizes the input data using an exponential function whose normalization characteristic is adjusted by the parameter of the exponential; and A quantum computing method, characterized in that it includes a step of performing a quantum computing operation based on input data normalized by a preprocessing device by a quantum computing circuit.

11. A first normalization module that normalizes input data using an exponential function whose normalization characteristics are adjusted by the parameters of the exponential function; and A preprocessing device for a quantum computing circuit, characterized by including a second normalization module that normalizes the output of the first normalization module to RMS (Root Mean Square) and inputs it to a quantum computing circuit.

12. A first normalization step for normalizing the input data using an exponential function whose normalization characteristics are adjusted by the parameters of the exponential function; and A preprocessing method for a quantum computing circuit, characterized by including a second normalization step for normalizing the output of the first normalization step to RMS (Root Mean Square) and inputting it to a quantum computing circuit.

Images