CN117172325A - Data coding method - Google Patents
Data coding method Download PDFInfo
- Publication number
- CN117172325A CN117172325A CN202311268565.3A CN202311268565A CN117172325A CN 117172325 A CN117172325 A CN 117172325A CN 202311268565 A CN202311268565 A CN 202311268565A CN 117172325 A CN117172325 A CN 117172325A
- Authority
- CN
- China
- Prior art keywords
- data
- quantum
- qubit
- training
- ith
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 239000002096 quantum dot Substances 0.000 claims abstract description 28
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 25
- 239000000758 substrate Substances 0.000 claims abstract description 13
- 102000004169 proteins and genes Human genes 0.000 claims abstract description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 description 41
- 239000013598 vector Substances 0.000 description 18
- 238000004364 calculation method Methods 0.000 description 11
- 230000000875 corresponding effect Effects 0.000 description 11
- 230000006870 function Effects 0.000 description 10
- 150000001875 compounds Chemical class 0.000 description 9
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 150000001413 amino acids Chemical class 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- HCHKCACWOHOZIP-UHFFFAOYSA-N Zinc Chemical compound [Zn] HCHKCACWOHOZIP-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 229910052796 boron Inorganic materials 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 229910052731 fluorine Inorganic materials 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 150000002632 lipids Chemical class 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 229910052757 nitrogen Inorganic materials 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 229910052698 phosphorus Inorganic materials 0.000 description 1
- 230000006916 protein interaction Effects 0.000 description 1
- 238000000455 protein structure prediction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 229910052717 sulfur Inorganic materials 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N10/00—Quantum computing, i.e. information processing based on quantum-mechanical phenomena
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Mathematics (AREA)
- Condensed Matter Physics & Semiconductors (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Complex Calculations (AREA)
- Semiconductor Memories (AREA)
Abstract
The embodiment of the application relates to a data coding method. A method of data encoding comprising: describing the data as a general expansion under the computing substrateWherein the data contains C elements corresponding to the substrate |x respectively 1 >……|x C >The method comprises the steps of carrying out a first treatment on the surface of the Performing one or more quantum gate operations on each qubit, wherein each qubit comprises an ith qubitAnd i=1, 2, … …, n; traversing each qubit in turn to do the following: for the ith qubit, calculate 2 i A parameter expressed as p i,d Where d=1, 2,3, … …,2 i The saidWherein the data is at least one of text, protein sequence, molecular structure, and gene. The data coding method provided by the embodiment of the application can effectively solve the problems encountered in the traditional technology.
Description
Information of divisional application
The application is a divisional application of an application patent application with the application date of 2022, 2 and 1, the application number of 202210115012.3 and the application name of 'training quantum circuit and data embedding method'.
Technical Field
Embodiments of the present application relate generally to the field of quantum computing, and more particularly, to a method of data encoding.
Background
In recent years, machine learning has been applied in a large number of fields. As the core of artificial intelligence, with the continuous increase of data analysis requirements of various industries in the big data age, how to perform deep analysis on complex and diverse data based on machine learning, and more efficient utilization of information becomes the main direction of machine learning research in the current big data environment. To further optimize traditional machine learning, research has begun to be conducted to assist machine learning with quantum circuits. However, current methods of training quantum wires are less.
The present application thus proposes a method of data encoding.
Disclosure of Invention
It is an aim of embodiments of the present application to provide a method of data encoding that reduces the amount of parameters to be trained compared to conventional models, and also significantly reduces the number of storage media, namely qubits, used.
An embodiment of the present application provides a method of data encoding, including: describing the data as a general expansion under the computing substrateWherein the data contains C elements corresponding to the substrate |x respectively 1 >……|x C >The method comprises the steps of carrying out a first treatment on the surface of the Performing one or more quantum gate operations on each qubit, wherein each qubit comprises an i-th qubit, and i = 1,2, … …, n; traversing each qubit in turn to do the following: for the ith qubit, calculate 2 i Parameters ofThe parameter is denoted as p i,d Where d=1, 2,3, … …,2 i Said->Wherein the data is at least one of text, protein sequence, molecular structure, and gene.
Another embodiment of the present application also provides a quantum computer configured to implement the method of data encoding described above.
In yet another embodiment of the present application, a computer readable storage medium is provided, on which a computer program is stored, wherein the processor executes the computer program to implement the method for encoding data as described above.
Compared with the prior art, the data coding method provided by the embodiment of the application can effectively improve the data processing speed and accuracy.
Drawings
The drawings that are necessary to describe embodiments of the present application or the prior art will be briefly described below in order to facilitate the description of the embodiments of the present application. It is apparent that the drawings in the following description are only a few embodiments of the application. It will be apparent to those skilled in the art that other embodiments of the drawings may be made in accordance with the structures illustrated in these drawings without the need for inventive faculty.
Fig. 1 is a schematic diagram of a method of training a quantum wire according to some embodiments of the application.
Fig. 2-4 are schematic structural views of entanglement layers according to some embodiments of the application.
Fig. 5 is a schematic diagram of a method of data embedding according to some embodiments of the present application.
Fig. 6 is a schematic representation of molecular structures according to some embodiments of the applications.
Detailed Description
For a better understanding of the spirit of embodiments of the present application, a further description of some preferred embodiments of the application is provided below.
Embodiments of the present application will be described in detail below. Throughout the present specification, the same or similar components and components having the same or similar functions are denoted by similar reference numerals. The embodiments described herein with respect to the drawings are of illustrative nature, of diagrammatic nature and are provided for the basic understanding of the present application. The embodiments of the present application should not be construed as limiting the application.
In addition, for ease of description, "first," "second," "third," etc. may be used herein to distinguish between different components of a figure or series of figures. The terms "first," "second," "third," and the like are not intended to describe corresponding components.
Before describing the technical scheme of the application, some key terms related in the application are explained first:
1. quantum computing: based on the computation mode of quantum logic, the basic unit for storing data is a quantum bit.
2. Quantum bit: basic unit of quantum computation. Conventional computers use 0 and 1 as basic units of binary. Except that quantum computation can handle both 0 and 1, so that the system can be in a linear superposition of 0 and 1: i psi>＝α|0>+β|1>This side α, β represents the complex probability amplitude of the system at 0 and 1. Their modulo square |alpha| 2 ，|β| 2 Representing probabilities at 0 and 1, respectively.
3. Quantum circuit: a representation of a quantum general-purpose computer represents a hardware implementation of the corresponding quantum algorithm/program under a quantum gate model. If the quantum circuit contains adjustable parameters for controlling the quantum gate, the quantum circuit is called parameterized quantum circuit.
4. Quantum gate: often using matrix representation, gates operating on n qubits can be represented by 2 n x 2 n Is a unitary matrix representation of (c). The number of qubits in one gate input and output must be equal. The operation of a quantum gate may be represented by multiplying a matrix representing the quantum gate with a vector representing the state of a qubit.
5. A revolving door: the turnstile is one of quantum gates, and is a group of three 2×2 unitary hermitian matrices (also called unitary matrices). Wherein the rotary x-gate isThe rotary y door is +.> The rotary z door is +.>
6. Quantum classical hybrid computation: the inner layer calculates corresponding physical quantity or loss function by using the quantum circuit, and the outer layer adjusts the calculation range of the variation parameters of the quantum circuit by using a traditional classical optimizer, so that the advantage of quantum calculation can be exerted to the maximum extent, and the method is believed to be one of important directions for potential demonstration of quantum advantage.
The application develops a method for training the quantum circuit for data embedding by using the quantum calculation method, and effectively improves the model learning capacity and the running efficiency by using the quantum calculation method.
Fig. 1 is a schematic diagram of a method of training a quantum wire according to some embodiments of the application. The method includes constructing a quantum wire 101 to be trained; and training with data in a target, which may include a training set or a test set, to update the quantum wire 101. Wherein training quantum wire 101 includes: encoding the data into quantum states (100 in fig. 1) using encoding line u (x); and applying the current quantum wire to the quantum state.
In other implementations of the application, the method of training a quantum wire may further include performing measurements on a computing substrate; and inputting the measured result to the classical neural network.
As shown in fig. 1, the quantum circuit 101 may be a variational circuit, and its effect may be represented by an operator V (θ), which may be trained into a suitable unitary transformation by using the training method of the present application.
The data in the target may be composed of one or more first data in the first database, for example, the data in the target may be data missing one or more first data. The first database is composed of all of the first data.
The method for training the quantum circuit can better reflect the relevance between different first data or between the data and the first data. In particular to the calculation of complex problems, the quantum machine learning is used, so that the calculation speed can be increased, and the calculation precision can be improved.
In some embodiments, the method of training a quantum wire further comprises: the first database is preprocessed. The preprocessing may include mapping each first data in the first database to a different number and mapping each first data to a computing base of the hilbert space by number.
For example, each first data is assigned a digital number (optionally, according to an existing assignment rule, or the co-occurring first data may be encoded into similar numbers, according to different methods) in a certain way, and once agreed, the numbers cannot be changed, and the numbers form the input x of the encoding line U (x) and also determine the quantum state of the encoded output.
In some embodiments, each first data in the first database (set the size of the first database as V) may be mapped to numbers of 0,1,2,3 … … V-1 one by one in some manner, optionally, co-occurring first data may be encoded into similar numbers, and then mapped to 2 of n qubits according to the numbers n The base of the individual hilbert spaces (requirement 2 n Not less than V, i.e. n not less than log 2 V), e.g., 0 first data corresponds to |0>State vector, first data number 634 corresponds to |634>First data of 999 corresponds to |999>And so on.
In some embodiments, after the first database preprocessing is completed, for one input, e.g., any data in the target, the input is provided with C words corresponding to the base |x respectively 1 >、|x 2 >、......、|x C >U (x) will handle qubits(qubit) from the initial state (use |0 for simplicity>Instead of) Transforming into an equal weight superposition state: u (x) in fig. 1 is constructed in turn to encode data in the target used as quantum states.
Unlike the traditional model in which the first data is mapped into a vector of a vector space, the method of mapping each first data into a vector of a Hilbert space can be obtained by training a parameterized quantum circuit, and the data is mapped into the vector of the Hilbert space by training a quantum circuit with a set of suitable unitary transformation parameters, so that the projection of the vector on each computing substrate can reflect the relevance of the data and corresponding data (such as missing one or more first data).
In some embodiments, encoding the data used into quantum states includes: encoding the data used in an equi-weighted superimposed state, e.g.And performing one or more quantum gate operations on each qubit, wherein each qubit comprises an i-th qubit, and i = 1,2, …, n.
In some embodiments, the data may be described as computing an under-substrate general expansionTraversing each qubit in turn to do the following: for the ith qubit, calculate 2 i A parameter, wherein said parameter may be +.>Wherein d=1, 2,3, …,2 i A kind of electronic deviceSequentially performing 2 on the ith qubit i-1 A controlled revolving door. Furthermore, the indices a=0, 1,2, …,2 can be used i-1 -1 denotes these controlled rotation gates, the control qubits being the first i-1 qubits, controlled rotation gate a performing: if the first i-1 qubits are at |a>Then the ith qubit is rotated, e.g., 0>Rotate to +.>Otherwise nothing is done. U (x) was constructed therefrom.
Although the above procedure involves a controlled gate count proportional to 2 n But due to most of C k =0 so most p i,d =0, the number of actually required controlled gates is much smaller than 2 n 。
The method for encoding the data into the quantum state can encode more information by using fewer quantum bits, and prepares the input data into the quantum state by converting the input data into a quantum superposition state so as to facilitate the quantum computation of the next step. Of course, the encoding of the data used into the quantum state is not limited to the above-described encoding method, and the encoding line U (x) may be constructed according to specific input data.
For example, according to the specific situation of the input data, when a line with 4 qubits is used, the corresponding label is 2, and since the binary representation of 2 is 0010, only the parameters of the controlled rotation gate in U (x) need to be set to 0, pi and 0 respectively, so that the input data is encoded.
In some embodiments, quantum wire 101 may be composed of two sub-modules, one being an entanglement module composed of N entanglement layers [ e.g. L in FIG. 1 ] i (θ)]And secondly, a single-bit rotating module consisting of a single-bit arbitrary rotating gate.
Entanglement module: the training device consists of N entanglement layers, the value of the number of layers N can be selected according to the requirements, and generally, the larger the number of layers N is, the better the training effect is. The effect of the entanglement layer is to apply a two-bit gate, such as a Controlled Not (CNOT) gate, between different quantum bits, thereby generating entanglement. Each entanglement layer can be divided into two parts: parameterized single bit rotation modules and fixed-structure two bit manipulation modules (fixed-structure refers to no adjustable parameters).
Fig. 2-4 are schematic structural views of entanglement layers according to some embodiments of the application. As shown in fig. 2-4, wherein the parameterized single bit manipulation module is composed of 3 rotational manipulations about the Y-axis, Z-axis, and Y-axis, respectively, the adjustable parameter is the angle of rotation. This is an euler rotation decomposition, meaning that such parameterized rotations can be combined into any single bit rotation operation. While two-bit operation modules have a variety of possible designs. As shown in fig. 2-4, the two-bit operation module is three typical entanglement layer structures, respectively, using nearest neighbor CNOT ring structure (105 in fig. 2), next-nearest neighbor CNOT ring structure (106 in fig. 3), and a combination of both (107 in fig. 4). The number of two-bit gates required for these examples is all in O (Nlog 2 V), V is the size of the first database and N is the number of layers of the entanglement. Of course, the entanglement modules of the application are not limited to the examples given.
Single bit rotation module: the module consists of n single bit arbitrary rotation gates for n qubits. The optional revolving door is combined by 3 rotation operations around the Y axis, the Z axis and the Y axis respectively, and the adjustable parameter is the rotation angle. This is an euler rotation angle decomposition, and such parameterized rotations can be combined into arbitrary single bit rotation operations.
The parameters in V (θ) all appear as the rotation angle of a single bit rotation gate when it is desired to measure a desired value of a certain mechanical quantity<M>For a certain parameter theta j The parameter shift rule may be applied and calculated as mentioned above for measuring the expected value and the partial derivative of the expected value.
In some embodiments, the method of training a quantum wire further comprises: a number of measurements (102 in fig. 1) are made on the computing substrate to obtain a model predicted value, e.g., an expected value, and the difference of the model predicted value and a reference value is compared, where the reference value corresponds to the data used.
In some embodiments, the method of training a quantum wire further comprises: the quantum wire 101 is updated according to the loss function using a gradient descent method, for example, the parameter θ in V (θ) is updated.
The data used can be trained repeatedly according to the required precision, the collapse probability distribution on the calculation substrate is obtained through multiple measurements and is used as a model predicted value, such as the probability distribution of what the missing one or more first data are likely to be, the true probability distribution in the first database is used as a reference value, a loss function (such as a cross entropy function) is freely selected to measure the difference between the model predicted value and the reference value, and the loss function and the gradient thereof are calculated according to the expected value and the partial derivative thereof.
For example, as shown in FIG. 1, M is any measurable physical quantity, e.g. using<M>To represent the expected values of a plurality of M to be measuredThe parameter shift rule can be applied, the remaining parameters are kept unchanged, use +.>Instead of theta j Repeating the operation of the line a number of times (the number of times is determined by the accuracy requirement) to measure a desired value +.>The other parameters are kept unchanged, use +.>Instead of theta j Repeating the operation of the line a number of times (the number of times is determined by the accuracy requirement) to measure a desired value +.>Will->Andthe partial derivative to be measured is obtained by subtracting and dividing by 2>
In some embodiments, due to the few parameters used in the quantum circuit, a classical neural network may be further used to perform fitting processing to further improve training efficiency. For example, the measured result, such as a desired value, may be input into a classical neural network, such as one or more perceptrons, e.g. as input to a multi-layer perceptron made up of one or more fully connected layers, from which the probability distribution of data corresponding to the data used in the object, such as "what the missing one or more first data is", is predicted.
In some embodiments, the target may be a training set, and multiple training is performed with different data in the training set to update the quantum wire until the entire training set is traversed, and the entire training set is iteratively trained R times (R is freely selectable) to improve training quality.
In some embodiments, the target may also be a test set, and the method of training the quantum wire further includes determining whether training is complete: and calculating a loss function according to the difference between the predicted value and the reference value and judging whether the loss function is smaller than a selected threshold value or not. If the calculated loss functions for selected data in the test set are all less than some selected threshold, the training is ended, otherwise the training of the quantum wire 101 is continued until the test of the test set can be passed.
The embodiment of the application also provides a data embedding method. Fig. 5 is a schematic diagram of a method of data embedding according to some embodiments of the present application. As shown in fig. 5, the method may include: encoding input data into quantum states (200); constructing a quantum wire 201; and applying the quantum wire 201 to the quantum state, wherein the quantum wire 201 comprises a trained variational wire.
In some embodiments of the present application, the method of training quantum wire 101 described above may be used to obtain a constructed quantum wire 201.
In some embodiments of the present application, encoding input data into quantum states is similar to the method described above for encoding input data into quantum states.
In some embodiments, after measuring the quantum state after the quantum circuit 201 acts, the result may be input into one or more perceptrons, for example, as an input of a multi-layer perceptrons composed of one or more fully connected layers, to complete the embedding of data.
The corresponding data most correlated to the input data is known by taking multiple measurements 202. Because the data embedding method uses a quantum computing method, the computing speed and accuracy are greatly improved.
The method for training the quantum circuit and embedding the data greatly reduces the parameter quantity required to be trained, namely the number of the used storage media, namely the quantum bits, and meanwhile, as the quantum state is normalized naturally, the normalization exponential function layer (softmax) probability normalization operation which consumes very much calculation force in the traditional method can be omitted, so that the method for embedding the data is more convenient and accurate.
The training quantum circuit and the data embedding method can be applied to a plurality of different fields. For example, in the field of Natural Language Processing (NLP), a quantum circuit with proper unitary transformation parameters can be obtained according to the method for training the quantum circuit, so that words and texts are mapped into state vectors of Hilbert space, and the projection of the state vectors on various computing substrates can reflect the relevance of words and texts.
Similarly, in the learning of chemical molecular characteristics, the vector of the substructures of the molecules in the chemical space can be learned by using the training quantum circuit and the data embedding method, and the vector coded by the whole compound can be finally obtained by summing the substructures. By mapping the compound substructures into the state vectors of the Hilbert space, the projection of the state vectors onto the respective computing substrates may reflect the relevance of the respective compound substructures. The method overcomes the defects of the traditional molecular characteristic representation, such as sparsity and bit conflict, and can be used for downstream tasks such as activity prediction, similarity retrieval and the like.
Fig. 6 is a schematic diagram of a molecular structure 10 according to some embodiments of the present application.
According to some embodiments of the application, the preprocessing of the molecular substructure library further comprises the steps of: the collection of compounds can be downloaded from the ZINK and chumbL database, de-duplicated for the known collection of compounds, the chemical informatics tool rdkit is used to obtain each atomic signature 11 for the compounds that remain in the collection of compounds with a molecular weight of 12-600, a heavy atom number of 3-50, a lipid water distribution coefficient clogP of-5 to 7, and compounds that contain only the elements H, B, C, N, O oxygen, F, P, S, cl and Br, the remainder being deleted, then the smiles for each compound are normalized, using the Morgan algorithm, with parameter radii set to 0 and 1, and then all atomic signatures are combined in sequence into the molecular structure 10, as shown in FIG. 6. Wherein an atomic signature for fewer than three occurrences may be replaced with "UNSEEN".
Similarly, in the protein sequence feature learning method, amino acids are mapped to the state vectors of the hilbert space, so that the projection of the state vectors on the respective calculation substrates can reflect the correlation of the respective amino acids. The protein sequence generated by model pre-training effectively extracts a series of physical and chemical properties, and can be used for downstream tasks such as protein family classification, protein structure prediction, protein interaction and the like.
It should be understood that, although the above embodiments apply the methods of training quantum circuits and data embedding in language processing, molecular feature learning and protein sequence feature learning, respectively, this is merely an exemplary embodiment for illustrating some applications provided by the present application and should not be construed as limiting the scope of the present application. According to other embodiments of the present application, the method of training quantum circuits and data embedding can be applied to other fields, such as biology, medicine, finance, etc., and the method of gene learning can also be used.
In some embodiments, the method of gene learning further comprises: and obtaining a model predicted value according to the output of the neural network, and comparing the difference between the model predicted value and a reference value, wherein the reference value is related to the used gene, for example, the reference value is the true probability distribution of the other gene in the coexpression gene pair corresponding to the used gene.
For genes in the target used, for example, one gene in a set of co-expressed gene pairs, the output of the neural network may be a probability distribution of the true other gene in the set of co-expressed gene pairs.
After two genes are input into the gene embedding model, respective corresponding vectors are obtained and combined, for example, the two vectors are overlapped, and interaction between the two genes can be predicted by inputting the two vectors into a gene interaction neural network model (such as a two-layer neural network)
The embodiment of the application also provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the method for training the quantum circuit and embedding the data.
The embodiment of the application also provides a quantum computer capable of realizing the training quantum circuit and the data embedding method, which is configured to realize the training quantum circuit and the data embedding method. For example, the quantum computer may include a processing unit and a memory device, wherein the memory device includes software capable of performing the above-described methods of training quantum wires and data embedding.
While the technical content and features of the present application have been disclosed above, those skilled in the art may make various substitutions and modifications based on the teachings and disclosure of the present application without departing from the spirit of the present application. Accordingly, the scope of the present application should not be limited to the embodiments disclosed, but should include various alternatives and modifications without departing from the application and be covered by the claims of the present application.
Claims (6)
1. A method of data encoding, comprising:
describing the data as a computing-under-the-substrate general expansionWherein the data contains C elements corresponding to base |x respectively 1 >……|x C >；
Performing one or more quantum gate operations on each qubit, wherein each qubit comprises an i-th qubit, and i = 1,2, … …, n; and
traversing each qubit in turn to do the following: for the ith qubit, calculate 2 i A parameter expressed as p i,d Where d=1, 2,3, … …,2 i The said
Wherein the data is at least one of text, protein sequence, molecular structure, gene.
2. The method of claim 1, further comprising: taking the ith quantum bit as a target bit and the first i-1 quantum bits as control bits, sequentially executing 2 i-1 A controlled revolving door.
3. The method of claim 2, wherein the indices a = 0,1,2, … …,2 are used i-1 -1 represents the controlled rotation gate, the control qubit being the first i-1 qubits, controlled rotation gate a performing: if the first i-1 qubits are at |a>The ith qubit is rotated.
4. The method of claim 3, wherein the rotating comprises rotating |0>Rotated as
5. A quantum computer configured to implement the method of any of the preceding claims 1-4.
6. A computer readable storage medium having stored thereon a computer program, wherein a processor executes the computer program to implement the method of any of the preceding claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311268565.3A CN117172325A (en) | 2022-02-01 | 2022-02-01 | Data coding method |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210115012.3A CN114444701B (en) | 2022-02-01 | 2022-02-01 | Training quantum circuit and data embedding method |
| CN202311268565.3A CN117172325A (en) | 2022-02-01 | 2022-02-01 | Data coding method |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210115012.3A Division CN114444701B (en) | 2022-02-01 | 2022-02-01 | Training quantum circuit and data embedding method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117172325A true CN117172325A (en) | 2023-12-05 |
Family
ID=81372415
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210115012.3A Active CN114444701B (en) | 2022-02-01 | 2022-02-01 | Training quantum circuit and data embedding method |
| CN202311268565.3A Pending CN117172325A (en) | 2022-02-01 | 2022-02-01 | Data coding method |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210115012.3A Active CN114444701B (en) | 2022-02-01 | 2022-02-01 | Training quantum circuit and data embedding method |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN114444701B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115691669B (en) * | 2023-01-03 | 2023-03-17 | 北京中科弧光量子软件技术有限公司 | Protein structure classification system based on quantum convolution neural network |
| CN115910220B (en) * | 2023-01-03 | 2023-05-09 | 北京中科弧光量子软件技术有限公司 | Quantum computer-based protein amino acid property encoding method and system |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6675565B1 (en) * | 2019-02-21 | 2020-04-01 | Ｍｄｒ株式会社 | Quantum computing device |
| CN110674921A (en) * | 2019-07-11 | 2020-01-10 | 中国科学技术大学 | Method for constructing quantum feedforward neural network based on classical training |
| US11797872B2 (en) * | 2019-09-20 | 2023-10-24 | Microsoft Technology Licensing, Llc | Quantum bit prediction |
| CN113012066B (en) * | 2021-03-15 | 2022-06-17 | 中南大学 | Quantum line learning method and image blurring method based on parametric vitreous color sampling |
| CN113592095B (en) * | 2021-08-04 | 2024-01-16 | 北京量子信息科学研究院 | Model training method and device based on quantum computation |
-
2022
- 2022-02-01 CN CN202210115012.3A patent/CN114444701B/en active Active
- 2022-02-01 CN CN202311268565.3A patent/CN117172325A/en active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| CN114444701A (en) | 2022-05-06 |
| CN114444701B (en) | 2023-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114444701B (en) | Training quantum circuit and data embedding method | |
| US11797872B2 (en) | Quantum bit prediction | |
| KR20210107584A (en) | Method for managing training data | |
| Wang et al. | Repetitive reprediction deep decipher for semi-supervised learning | |
| US11460982B1 (en) | Number embedding application system | |
| Zhao et al. | A review of quantum neural networks: methods, models, dilemma | |
| Wang et al. | Development of variational quantum deep neural networks for image recognition | |
| CN114444664A (en) | Attention model and neural network model based on quantum computation | |
| US11636374B2 (en) | Exponential spin embedding for quantum computers | |
| CN114283903A (en) | Method for predicting protein drug binding force based on quantum gated recurrent neural network | |
| CN114444723A (en) | Model for data reconstruction | |
| Altares-López et al. | AutoQML: Automatic generation and training of robust quantum-inspired classifiers by using evolutionary algorithms on grayscale images | |
| Basterrech et al. | Evolutionary Echo State Network: A neuroevolutionary framework for time series prediction | |
| CN114511092B (en) | Graph attention machine mechanism implementation method based on quantum circuit | |
| Kashyap et al. | Quantum convolutional neural network architecture for multi-class classification | |
| Lin et al. | Continuation path learning for homotopy optimization | |
| CN114511097A (en) | Mutual learning method and system based on quantum circuit | |
| Vasuki et al. | Overview of Quantum Computing in Quantum Neural Network and Artificial Intelligence | |
| Bai et al. | Superposition-enhanced quantum neural network for multi-class image classification | |
| Padha et al. | Quantum deep neural networks for time series analysis | |
| Zankov et al. | Conjugated quantitative structure‐property relationship models: Prediction of kinetic characteristics linked by the Arrhenius equation | |
| CN117994397B (en) | Digital human text action generation method, device, computer equipment and storage medium | |
| US20230368036A1 (en) | Physics-informed multimodal autoencoder | |
| Pande | A Comprehensive Review of Data Encoding Techniques for Quantum Machine Learning Problems | |
| CN114496114A (en) | Model and method for molecular structure reasoning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |