CN117936103B - Intelligent AI acupuncture model training system and method based on neural network - Google Patents
Intelligent AI acupuncture model training system and method based on neural network Download PDFInfo
- Publication number
- CN117936103B CN117936103B CN202410332261.7A CN202410332261A CN117936103B CN 117936103 B CN117936103 B CN 117936103B CN 202410332261 A CN202410332261 A CN 202410332261A CN 117936103 B CN117936103 B CN 117936103B
- Authority
- CN
- China
- Prior art keywords
- matrix
- patient
- target
- data
- covariance matrix
- 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.)
- Active
Links
- 238000001467 acupuncture Methods 0.000 title claims abstract description 126
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 44
- 238000012549 training Methods 0.000 title claims abstract description 44
- 239000011159 matrix material Substances 0.000 claims abstract description 486
- 238000003062 neural network model Methods 0.000 claims abstract description 22
- 230000009467 reduction Effects 0.000 claims description 46
- 239000013598 vector Substances 0.000 claims description 39
- 230000006870 function Effects 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000010801 machine learning Methods 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 32
- 238000011084 recovery Methods 0.000 description 32
- 238000005516 engineering process Methods 0.000 description 5
- 238000011282 treatment Methods 0.000 description 5
- 201000010099 disease Diseases 0.000 description 4
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 4
- 230000006461 physiological response Effects 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 230000002596 correlated effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000001747 exhibiting effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000011514 reflex Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
- 238000011269 treatment regimen Methods 0.000 description 1
Landscapes
- Finger-Pressure Massage (AREA)
Abstract
The invention relates to the technical field of machine learning, in particular to an intelligent AI acupuncture model training system and method based on a neural network, comprising the following steps: acquiring acupuncture data of a patient, acquiring a covariance matrix of the patient, and acquiring a matrix block of the covariance matrix; obtaining the stability of the matrix block according to the difference between the element values in the matrix block; acquiring the stability of the element value, and acquiring an estimated value of the element value according to the stability of the element value; obtaining a final covariance matrix of each patient according to the estimated quantity; thereby obtaining a dimension-reducing matrix; training the neural network according to the dimension-reducing matrix, and obtaining a reference index of the rehabilitation date of the patient according to the trained neural network model. According to the invention, the dimension of the covariance matrix related to the patient is reduced, the key acupuncture information of the patient is reserved, the training speed of the neural network model is improved, and the reference index of the rehabilitation date of the patient is obtained through the trained neural network.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to an intelligent AI acupuncture model training system and method based on a neural network.
Background
Acupuncture has a long history as a traditional Chinese medicine therapy. Along with the development of modern science and technology, people have a deeper understanding of acupuncture. The intelligent AI acupuncture model based on the neural network is to simulate the relevant characteristics and effects of acupuncture by using the neural network technology, thereby realizing automation, intellectualization or auxiliary decision in acupuncture treatment.
When the intelligent AI acupuncture model is trained, due to the fact that training data are few, repeatability of acupuncture data in different dimensions, such as acupuncture point positions, using technology, physiological responses of patients and the like, the data are highly correlated with each other, so that data features are highly repeated, training time is long when the AI acupuncture model is trained by using a neural network, and training cost is high and efficiency is low.
Disclosure of Invention
In order to solve the problems, the invention provides an intelligent AI acupuncture model training system and method based on a neural network.
The intelligent AI acupuncture model training system and method based on the neural network adopt the following technical scheme:
The embodiment of the invention provides an intelligent AI acupuncture model training method based on a neural network, which comprises the following steps:
acquiring acupuncture data of a plurality of patients, wherein the acupuncture data comprises a plurality of data; encoding each data in the acupuncture data of each patient to obtain row vectors corresponding to each data in the acupuncture data of each patient; obtaining a numerical matrix of each patient according to the acupuncture data of each patient and row vectors corresponding to each data in the acupuncture data; obtaining a covariance matrix of each patient according to the numerical matrix of each patient;
Equally dividing the covariance matrix of each patient to obtain a plurality of matrix blocks of the covariance matrix of each patient; according to the difference between the element values in the matrix blocks, the stability of each matrix block in the covariance matrix of each patient is obtained; according to the stability of the matrix blocks, a plurality of unstable matrix blocks and a plurality of stable matrix blocks of the covariance matrix of each patient are obtained; according to the stability of the matrix block and the element values in the matrix block, the stability of each element value in the covariance matrix of each patient is obtained;
According to the element values in the unstable matrix block, the element values in the stable matrix block and the stability of the element values, obtaining an estimated value of each element value in the covariance matrix of each patient; obtaining an updated value of each element value in the covariance matrix of each patient according to the estimated quantity of the element value; replacing elements in the covariance matrix according to the updated values of the element values to obtain a final covariance matrix of each patient; performing dimension reduction on the final covariance matrix of each patient to obtain a dimension reduction matrix of each patient;
training the neural network according to the dimension-reducing matrix, and obtaining a reference index of the rehabilitation date of the patient according to the trained neural network model.
Preferably, the method for obtaining the stability of each matrix block in the covariance matrix of each patient according to the difference between the element values in the matrix blocks comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; marking any matrix block of the target covariance matrix as a target matrix block;
；
In the method, in the process of the invention, For the/>, in the target matrix blockIndividual element values,/>For the/>, in the target matrix blockIndividual element values,/>To take absolute value,/>For the number of element values in the target matrix block,/>Is the stability of the target matrix block.
Preferably, the method for obtaining the unstable matrix blocks and the stable matrix blocks of the covariance matrix of each patient according to the stability of the matrix blocks comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix;
Presetting a first threshold value, acquiring stability of each matrix block in a target covariance matrix, performing linear normalization processing on stability of all matrix blocks in the target covariance matrix, taking the obtained result as normalized stability of each matrix block in the target covariance matrix, taking the matrix blocks with normalized stability smaller than the first threshold value in the target covariance matrix as unstable matrix blocks of the target covariance matrix, and taking the matrix blocks with normalized stability larger than or equal to the first threshold value in the target covariance matrix as stable matrix blocks of the target covariance matrix.
Preferably, the obtaining the stability of each element value in the covariance matrix of each patient according to the stability of the matrix block and the element values in the matrix block comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; marking any matrix block of the target covariance matrix as a target matrix block; obtaining a Z score of each element value in the target matrix block through a Z-score algorithm; and carrying out linear normalization processing on the Z scores of all the element values in the target matrix block, and taking the product of the normalized Z score of each element value in the target matrix block and the stability of the target matrix block as the stability of each element value in the target matrix block.
Preferably, the estimating method for each element value in the covariance matrix of each patient according to the element value in the unstable matrix block, the element value in the stable matrix block and the stability of the element value includes the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; recording any unstable matrix block of the target covariance matrix as a target unstable matrix block;
；
In the method, in the process of the invention, For the/>, in the target unstable matrix blockStability of individual element values; /(I)Is an exponential function with a natural constant as a base; /(I)For the number of element values in the target unstable matrix block,/>For the/>, in the target unstable matrix blockIndividual element values,/>Is a normalization function; /(I)For the/>, in the target unstable matrix blockAn estimate of the individual element values;
for a stable matrix block of the target covariance matrix, the estimator of each element value in the stable matrix block is 0.
Preferably, the obtaining the updated value of each element value in the covariance matrix of each patient according to the estimated value of the element value includes the following specific steps:
Marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; 1 is combined with the first of the target covariance matrices Sum of the estimates of the individual element values, and the/>, in the target covariance matrixThe product of the individual element values as the/>, in the target covariance matrixUpdated values of the individual element values.
Preferably, the replacing the elements in the covariance matrix according to the updated values of the element values to obtain a final covariance matrix of each patient includes the following specific steps:
Marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; and acquiring an updated value of each element value in the target covariance matrix, replacing each element value in the target covariance matrix with a corresponding updated value to obtain a matrix, and recording the matrix as a final covariance matrix of the target patient.
Preferably, the step of obtaining the numerical matrix of each patient according to the acupuncture data of each patient and the row vectors corresponding to each data in the acupuncture data comprises the following specific steps:
Marking any patient as a target patient, acquiring row vectors corresponding to each data in the acupuncture data of the target patient, and longitudinally splicing the row vectors corresponding to all data in the acupuncture data of the target patient to obtain an initial matrix of the target patient; acquiring all numerical data in acupuncture data of a target patient; and marking the product result of all the numerical data in the acupuncture data of the target patient as the comprehensive product parameter of the target patient, multiplying the comprehensive product parameter of the target patient with the initial matrix of the target patient to obtain a matrix, and marking the matrix as the numerical matrix of the target patient.
Preferably, the encoding each data in the acupuncture data of each patient to obtain a row vector corresponding to each data in the acupuncture data of each patient includes the following specific steps:
And marking any One patient as a target patient, marking any One data in the acupuncture data of the target patient as target data, and Encoding the target data by using an One-hot Encoding algorithm to obtain row vectors corresponding to the target data.
The invention also provides an intelligent AI acupuncture model training system based on the neural network, which comprises a memory and a processor, wherein the processor executes a computer program stored in the memory to realize the steps of the method.
The technical scheme of the invention has the beneficial effects that: after acupuncture data of a plurality of patients are obtained, each data in the acupuncture data of each patient is encoded to obtain row vectors corresponding to each data in the acupuncture data of each patient, so that text information in the acupuncture data can be converted into numerical information, further, a covariance matrix constructed subsequently can better represent relevant information of acupuncture of the patients, the covariance matrix of each patient is obtained through row vectors corresponding to each data in the acupuncture data and the acupuncture data of each patient, a plurality of matrix blocks of the covariance matrix of each patient are obtained through equally dividing the covariance matrix of each patient, stability analysis is carried out on matrix blocks of the covariance matrix of each patient, relevant information of acupuncture of the patient can be reflected better, dimension reduction is carried out on the follow-up matrix, training speed of a neural network is improved, stability of element values in the matrix blocks is obtained through differences among element values in the matrix blocks, stability of the element values in the matrix blocks is further obtained, stability of each element value in the covariance matrix is further obtained, the covariance matrix is further obtained through element values in the unstable matrix blocks, the dimension reduction of each element in the covariance matrix block is further, the dimension of each element in the covariance matrix is further obtained, the element in the covariance matrix is further, the final training of the neural network is further, the time of the neural network is further obtained through the dimension reduction of each element in the covariance matrix is further, the final training of the neural network is further, the training time of the neural network is further obtained, and the training time of the neural network is further finally obtained through the training of the element is reduced.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart illustrating a training method of an intelligent AI acupuncture model based on a neural network according to a second embodiment of the present invention;
FIG. 2 is a schematic diagram of an initial matrix of acquisition target patients according to a second embodiment of the present invention;
Fig. 3 is a flowchart of obtaining a reference index of a patient's rehabilitation date according to a second embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention to achieve the preset purpose, the following detailed description is given below of a neural network-based intelligent AI acupuncture model training system and method according to the present invention, with reference to the accompanying drawings and the preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the intelligent AI acupuncture model training system and method based on the neural network provided by the invention with reference to the accompanying drawings.
The first embodiment provided by the invention is as follows:
step S101, acupuncture data of a plurality of patients are acquired.
The acupuncture data of the patient includes a plurality of data, such as acupuncture point data, disease type data, treatment regimen data, and heart rate data of the patient.
Step S102, each data containing text information in the acupuncture data is encoded, and the covariance matrix of each patient is obtained by combining each data containing numerical information in the acupuncture data.
It should be noted that, each data containing text information in the acupuncture data is encoded, where the text information refers to that the information in the data is text, for example, acupuncture point location data and disease type data, and specifically, the text information data; the row vector corresponding to each data in the acupuncture data can be obtained through encoding, a first matrix can be constructed, each data containing numerical value information in the acupuncture data is multiplied with the first matrix to obtain a second matrix, wherein each data of the numerical value information refers to a numerical value, for example, heart rate data of a patient, and the information in the data is a numerical value; and obtaining a covariance matrix through the second matrix.
Step S103, equally dividing the covariance matrix of each patient to obtain a plurality of matrix blocks of the covariance matrix of each patient, obtaining the stability of each matrix block in the covariance matrix of each patient according to the difference between the element values in the matrix blocks, obtaining the stability of each element value in the covariance matrix of each patient according to the stability of the matrix blocks and the element values in the matrix blocks, obtaining the estimated value of each element value in the covariance matrix of each patient according to the element values and the stability of the element values in the matrix blocks, obtaining the final covariance matrix of each patient according to the estimated value of the element values, and performing dimension reduction on the final covariance matrix to obtain the dimension reduction matrix of each patient.
The stability of the element values is obtained by multiplying the Z-score of the element values in the matrix block by the stability of the corresponding matrix block.
Step S104, collecting acupuncture data of a large number of patients, acquiring a dimension reduction matrix of each patient in the large number of patients, marking the dimension reduction matrix as a first dimension reduction matrix, and taking all the first dimension reduction matrices as a data set of a neural network model; the neural network model in this embodiment is specifically ResNet to 50.
Each first dimension-reducing matrix in the data set is artificially marked, and the marking method comprises the following steps: when the recovery date of the patient is less than or equal to one month, marking a first dimension-reduction matrix of the patient with the recovery date less than or equal to one month as a higher recovery degree; when the recovery date of the patient is more than one month and less than six months, marking the first dimension-reduction matrix of the patient with the recovery date of more than one month and less than six months as a medium recovery degree; when the recovery date of the patient is greater than or equal to six months, the first dimension-reduction matrix of the patient with the recovery date greater than or equal to six months is marked as a lower recovery degree.
Inputting each first dimension reduction matrix in the data set into a neural network model, wherein a loss function of the neural network model is a cross entropy loss function, and outputting to obtain the probability that each first dimension reduction matrix is of a higher rehabilitation degree, the probability that each first dimension reduction matrix is of a medium rehabilitation degree and the probability that each first dimension reduction matrix is of a low rehabilitation degree; and when the loss function converges, obtaining a trained neural network model.
Marking any patient as a target patient; the dimension-reducing matrix of the target patient is marked as a target dimension-reducing matrix; the target dimension reduction matrix is input into a neural network model after training, the probability of obtaining the higher recovery degree, the probability of obtaining the middle recovery degree and the probability of obtaining the lower recovery degree of the target dimension reduction matrix are output, and the probability of obtaining the higher recovery degree, the probability of obtaining the middle recovery degree and the probability of obtaining the lower recovery degree of the target dimension reduction matrix are used as reference indexes of the recovery date of a target patient.
The covariance matrix of the patient is obtained, the dimension of the covariance matrix is reduced, the neural network is trained according to a large number of dimension reduction matrices, and the reference index of the rehabilitation date of the patient is obtained.
Further, the invention provides an embodiment II for describing the intelligent AI acupuncture model training method based on the neural network in detail.
The second embodiment provided by the invention is as follows:
Referring to fig. 1, a flowchart of a training method of an intelligent AI acupuncture model based on a neural network according to a second embodiment of the invention is shown, and the method includes the following steps:
Step S001, acquiring acupuncture data of a plurality of patients; encoding each data in the acupuncture data of each patient to obtain row vectors corresponding to each data in the acupuncture data of each patient; obtaining a numerical matrix of each patient according to the acupuncture data of each patient and row vectors corresponding to each data in the acupuncture data; and obtaining a covariance matrix of each patient according to the numerical matrix of each patient.
It should be noted that, the main purpose of this embodiment is to input the reduced-dimension acupuncture data into the neural network for training and subsequent processing, so that the acupuncture data with different dimensions need to be acquired and preprocessed before the acupuncture data is input into the network.
Specifically, acupuncture data of a plurality of patients are obtained, specifically as follows:
Obtaining acupuncture data for a number of patients, including but not limited to: acupuncture point position data, disease type data, treatment manipulation data, heart rate data of a patient and the like, wherein the treatment manipulation data comprise twisting, lifting and inserting, breathing, following reinforcing and reducing manipulations, mountain fire burning, natural cooling manipulations and the like.
Further, each data in the acupuncture data of each patient is encoded to obtain a row vector corresponding to each data in the acupuncture data of each patient, which is specifically as follows:
Marking any One patient as a target patient, marking any One data in acupuncture data of the target patient as target data, and Encoding the target data by utilizing an One-hot Encoding algorithm to obtain row vectors corresponding to the target data; it should be noted that, the One-hot Encoding algorithm is utilized to encode the target data, so as to obtain the existing method of the row vector corresponding to the target data as the One-hot Encoding algorithm, and the One-hot Encoding algorithm creates a virtual variable and converts the virtual variable into binary vectors, for example, three acupuncture points (acupuncture point location data) are assumed to be respectively 。/>Can be encoded as/>，/>Can be encoded as/>，/>Can be encoded as; It should be noted that, the target data is directed at the text information in the acupuncture data, and the numerical information is not subjected to the independent heat coding treatment; if the length of the row vector corresponding to each data is different, 0 is supplemented after the row vector, so that the subsequent splicing matrix operation is facilitated.
Further, according to the acupuncture data of each patient and the row vector corresponding to each data in the acupuncture data, a numerical matrix of each patient is obtained, which is specifically as follows:
acquiring row vectors corresponding to each type of data in the acupuncture data of a target patient, longitudinally splicing the row vectors corresponding to all types of data in the acupuncture data of the target patient according to the acquisition sequence to obtain a matrix with increased row numbers and unchanged column numbers, and recording the matrix as an initial matrix of the target patient, referring to fig. 2, fig. 2 is a schematic diagram of the initial matrix of the target patient, fig. 2 contains three types of schematic row vectors corresponding to the data, namely row vectors of acupuncture point data, row vectors of disease type data and row vectors of treatment method data, and longitudinally splicing the row vectors to obtain a large matrix with increased row numbers; it should be noted that, the initial matrix is a square matrix, if the number of elements in different column vectors is different, 0 is added in front, so that the number of elements in the column vectors is the same, and then the square matrix is obtained; acquiring all numerical data in acupuncture data of a target patient; it should be noted that, the numerical data refers to specific numerical information, such as heart rate data of the patient in the acupuncture data; and marking the product result of all the numerical data in the acupuncture data of the target patient as the comprehensive product parameter of the target patient, multiplying the comprehensive product parameter of the target patient with the initial matrix of the target patient to obtain a matrix, and marking the matrix as the numerical matrix of the target patient.
Further, a covariance matrix of each patient is obtained according to the numerical matrix of each patient.
It should be noted that, the covariance matrix of each patient is obtained according to the numerical matrix of each patient as the existing method, which is not described in detail in this embodiment.
To this end, a covariance matrix for each patient is obtained.
Step S002, equally dividing the covariance matrix of each patient to obtain a plurality of matrix blocks of the covariance matrix of each patient; according to the difference between the element values in the matrix blocks, the stability of each matrix block in the covariance matrix of each patient is obtained; according to the stability of the matrix blocks, a plurality of unstable matrix blocks and a plurality of stable matrix blocks of the covariance matrix of each patient are obtained; and obtaining the stability of each element value in the covariance matrix of each patient according to the stability of the matrix block and the element values in the matrix block.
It should be noted that, the covariance matrix is obtained as described above, since the scale of the acupuncture data is not large, the available data is small, the repeatability of the acupuncture data in different dimensions, such as the acupuncture point position, the use technique, and the physiological response of the patient, is highly correlated with each other, so that the data features are highly repeated, that is, the same patient has higher element value repeatability under the row vector where the patient is located, and the repeatability results in the covariance matrix feature exhibiting less stability.
It should be noted that, the covariance matrix stability due to the scene limitation is low, and the stability refers to the number of zero elements in the matrix and the number of corresponding high-repeatability element values, and the stability is lower as the number is larger. In order to eliminate inaccurate dimension reduction results caused by the low stability, local stability of different areas in the matrix needs to be detected and acquired, element estimation amount determination is carried out according to the local stability, and the higher the local stability is, the lower the element estimation amount is, and otherwise, the higher the element estimation amount is; further, after determining the local stability of the different matrix areas, if the area stability is low, the missing value needs to be estimated, so as to improve the stability of the matrix. In order to obtain a reasonable element deficiency value estimator, the reasonable element deficiency value estimator is characterized in that on the basis of guaranteeing the original matrix element value characteristics, effective values can be added, the stability of a local area in a matrix can be increased after the effective values are added, the data quality after vector dimension reduction is improved, and the training and recognition efficiency of a neural network is improved.
It should be noted that, due to the limitation of the data characteristics of the scene, the more the number of zero values of the local area in the covariance matrix obtained according to the acupuncture data, the more the zero values in the corresponding scene, namely, the matrix; the more the element values are similar, the more the corresponding scene is: the more highly correlated such data are to each other, such as acupoint location, technique of use, and patient physiological response, such that the data features are highly repetitive, the higher the matrix instability. Thus, the higher the instability of the element in this local region.
Specifically, the covariance matrix of each patient is equally divided, and a plurality of matrix blocks of the covariance matrix of each patient are obtained, which are specifically as follows:
And marking the covariance matrix of the target patient as a target covariance matrix.
The size of the target covariance matrix is recorded asThe target covariance matrix is obtained through equal-proportion blockingThe size is/>Matrix block of/>A preset block proportion; in this embodiment,/>In this embodiment, the interpolation filling data is performed on the portion of the target covariance matrix beyond the boundary by using the quadratic linear interpolation method.
Further, according to the difference between the element values in the matrix blocks, the stability of each matrix block in the covariance matrix of each patient is obtained, specifically as follows:
and marking any matrix block of the target covariance matrix as a target matrix block.
；
In the method, in the process of the invention,For the/>, in the target matrix blockIndividual element values,/>For the/>, in the target matrix blockIndividual element values,/>To take absolute value,/>For the number of element values in the target matrix block,/>Is the stability of the target matrix block.
It should be noted that, the element value refers to a specific numerical value in the matrix,The method is characterized in that the method is used for representing the abnormality between element values in a target matrix block, the smaller the value is, the smaller the difference between the element values in the target matrix block is, the closer the element values are, the lower the stability of the target matrix block is, and the logic of the method is as follows: when the results obtained by acupuncture are consistent, such as acupuncture point positions, using technology, physiological reactions of patients and the like, the element values obtained by encoding are close to each other, so that the smaller the element value difference value in a target matrix block is, the lower the stability of the target matrix block is, the stability of the target matrix block is different from the stability of general sequence data, when the element values of certain two rows or columns in the target matrix block are similar or consistent, the closer the target matrix block is to a singular matrix, the closer the characteristic value is to 0, the less the element values are provided with the characteristics of the original matrix, and the lower the stability is the closer the element values are; /(I)Representing the difference between the element values and 0 elements in the target matrix block, the more the number of local area zero values in the target matrix block is intended to quantify the covariance matrix, the higher the matrix instability, when the difference result is closer to 0, the more zero elements or elements close to 0 in the target matrix block are, the lower the stability is, and the corresponding to the scene is: the more zero values in the matrix, the fewer features contained in the covariance matrix; will/>As a weight and/>And multiplying to obtain the stability of the target matrix block.
Further, according to the stability of the matrix blocks, a plurality of unstable matrix blocks and a plurality of stable matrix blocks of the covariance matrix of each patient are obtained, specifically as follows:
The method comprises the steps of presetting a first threshold value, describing with the first threshold value being 0.3, obtaining the stability of each matrix block in a target covariance matrix, carrying out linear normalization processing on the stability of all matrix blocks in the target covariance matrix, taking the obtained result as the normalized stability of each matrix block in the target covariance matrix, taking the matrix blocks with the normalized stability of the matrix blocks in the target covariance matrix being smaller than the first threshold value as unstable matrix blocks of the target covariance matrix, and taking the matrix blocks with the normalized stability of the matrix blocks in the target covariance matrix being greater than or equal to the first threshold value as stable matrix blocks of the target covariance matrix.
Further, according to the stability of the matrix block and the element values in the matrix block, the stability of each element value in the covariance matrix of each patient is obtained, specifically as follows:
Obtaining a Z score of each element value in the target matrix block through a Z-score algorithm; it should be noted that, the Z score of each element value in the target matrix block obtained by the Z-score algorithm is an existing method, which is not described in detail in this embodiment.
；
In the method, in the process of the invention,For stability of target matrix block,/>For the/>, in the target matrix blockThe Z-score of the value of the individual element,As a linear normalization function, the normalized object is the Z score of all element values in the target matrix block,/>For the/>, in the target matrix blockStability of individual element values.
It should be noted that, when a certain element value in the target matrix block deviates from the overall value, the higher the stability is, the closer the logic reference element values are, the lower the stability is, the relationship is quantified by using Z-score, the higher the score, the greater the deviation degree is, the higher the stability is,The larger then the/>The higher the stability of the individual element values.
So far, the stability of each element value in the covariance matrix of each patient is obtained.
Step S003, according to the element values in the unstable matrix block, the element values in the stable matrix block and the stability of the element values, obtaining an estimated value of each element value in the covariance matrix of each patient; obtaining an updated value of each element value in the covariance matrix of each patient according to the estimated quantity of the element value; replacing elements in the covariance matrix according to the updated values of the element values to obtain a final covariance matrix of each patient; and performing dimension reduction on the final covariance matrix of each patient to obtain a dimension reduction matrix of each patient.
It should be noted that, the more unstable the element value in the matrix, the more the element missing value is required to be estimated, and the element missing value estimation amount can be obtained on the premise of ensuring that the matrix characteristics are not lost or changed. Therefore, in order not to lose or change matrix eigenvalues and eigenvectors, element eigenvalue distribution is selected to be obtained, the eigenvalue distribution comprises the change relation among local elements and corresponding scene meanings, and further, the missing value estimated quantity can be obtained by combining element stability, and the lower the element value stability is, the higher the missing value estimated quantity is.
Specifically, according to the element values in the unstable matrix block, the element values in the stable matrix block and the stability of the element values, the estimated value of each element value in the covariance matrix of each patient is obtained, specifically as follows:
And marking any unstable matrix block of the target covariance matrix as a target unstable matrix block.
；
In the method, in the process of the invention,For the/>, in the target unstable matrix blockStability of individual element values; /(I)The present embodiment uses/>, as an exponential function based on natural constantsFor presenting the inverse proportion relation, and other inverse proportion functions can be set in the specific embodiment; /(I)For the number of element values in the target unstable matrix block,/>For the/>, in the target unstable matrix blockIndividual element values,/>Is a linear normalization function; /(I)When in normalization, the normalized object is the stability of all element values in the target unstable matrix block; when normalizing, the normalized object is the/> of all element values in the target unstable matrix block ；/>For the/>, in the target unstable matrix blockAn estimate of the value of the individual element.
For a stable matrix block of the target covariance matrix, the estimator of each element value in the stable matrix block is 0.
For the estimated amount of the element value, since the lower the stability of the element value is, the higher the estimated amount of the missing value is, and thus the value is estimated byIs to present the inverse proportional relation, i.e. >The different dimension values of the acupuncture data of a certain patient in the corresponding scene, such as the positions of acupuncture points, the use technology, the physiological response of the patient and the like, are similar after being converted into covariance, and the acupuncture times are less than those of acupuncture history data which can be used for reference, so that the patient has less characteristics, at the moment, the patient needs to be estimated according to the existing data, and the less the existing data, namely the lower the stability, the larger the estimated value, and the aim of improving the matrix element data characteristics is achieved while the stability of a matrix block is improved; /(I)Represents a weighted element mean with a weight of/>Namely, the higher the stability is, the higher the weight value is, and the corresponding scene is: for a certain patient, the more the data of acupuncture features exist, the higher the weight value is usually possessed when the data participate in constructing the estimated quantity for obtaining other dimension non-valued or lower stability data, the more the features of key data can be highlighted and amplified, and finally the estimated quantity of element values is obtained by normalization.
Further, an updated value of each element value in the covariance matrix of each patient is obtained according to the estimated values of the element values, specifically as follows:
; in the/> For the/>, in the target covariance matrixEstimated amount of individual element values,/>For the/>, in the target covariance matrixIndividual element values,/>For the/>, in the target covariance matrixUpdated values of the individual element values.
Further, replacing elements in the covariance matrix according to updated values of the element values to obtain a final covariance matrix of each patient, which is specifically as follows:
and acquiring an updated value of each element value in the target covariance matrix, replacing each element value in the target covariance matrix with a corresponding updated value to obtain a matrix, and recording the matrix as a final covariance matrix of the target patient.
Further, the final covariance matrix of each patient is subjected to dimension reduction, and a dimension reduction matrix of each patient is obtained, which is specifically as follows:
And performing dimension reduction on the final covariance matrix of the target patient through principal component analysis (PCA, PRINCIPAL COMPONENT ANALYSIS) to obtain a dimension reduction matrix of the target patient. It should be noted that, the dimension reduction is performed on the matrix by the principal component analysis (PCA, PRINCIPAL COMPONENT ANALYSIS), and the dimension reduction matrix is obtained by the existing method, which is not described in detail in this embodiment.
Thus, a dimension-reducing matrix of each patient is obtained.
And step S004, training the neural network according to the dimension-reducing matrix, and obtaining a reference index of the rehabilitation date of the patient according to the trained neural network model.
It should be noted that the dimension-reducing matrix contains data of different dimensions of acupuncture and reduces the data volume on the basis of retaining basic characteristics, so as to facilitate neural network training and recognition.
Specifically, training a neural network according to the dimension-reduction matrix, and obtaining a reference index of the rehabilitation date of the patient according to the trained neural network model, wherein the reference index is specifically as follows:
collecting acupuncture data of a large number of patients, acquiring a dimension reduction matrix of each patient in the large number of patients by using the embodiment, marking the dimension reduction matrix as a first dimension reduction matrix, and taking all the first dimension reduction matrices as a data set of a neural network model; the neural network model in this embodiment is specifically ResNet to 50.
Each first dimension-reducing matrix in the data set is artificially marked, and the marking method comprises the following steps: when the recovery date of the patient is less than or equal to one month, marking a first dimension-reduction matrix of the patient with the recovery date less than or equal to one month as a higher recovery degree; when the recovery date of the patient is more than one month and less than six months, marking the first dimension-reduction matrix of the patient with the recovery date of more than one month and less than six months as a medium recovery degree; when the recovery date of the patient is greater than or equal to six months, the first dimension-reduction matrix of the patient with the recovery date greater than or equal to six months is marked as a lower recovery degree.
Inputting each first dimension reduction matrix in the data set into a neural network model, wherein a loss function of the neural network model is a cross entropy loss function, and outputting to obtain the probability that each first dimension reduction matrix is of a higher rehabilitation degree, the probability that each first dimension reduction matrix is of a medium rehabilitation degree and the probability that each first dimension reduction matrix is of a low rehabilitation degree; and when the loss function converges, obtaining a trained neural network model.
The dimension-reducing matrix of the target patient is marked as a target dimension-reducing matrix; the target dimension reduction matrix is input into a neural network model after training, the probability of obtaining the higher recovery degree, the probability of obtaining the middle recovery degree and the probability of obtaining the lower recovery degree of the target dimension reduction matrix are output, and the probability of obtaining the higher recovery degree, the probability of obtaining the middle recovery degree and the probability of obtaining the lower recovery degree of the target dimension reduction matrix are used as reference indexes of the recovery date of a target patient.
It should be noted that, the neural network model is trained to obtain a trained neural network model, specifically, the existing method, and this embodiment will not be described again.
The covariance matrix of the patient is subjected to dimension reduction, the neural network is trained through the dimension reduction matrix, and the reference index of the rehabilitation date of the patient is obtained according to the trained neural network model.
Another embodiment of the present invention provides a neural network-based intelligent AI acupuncture model training system, the system including a memory and a processor, the processor executing a computer program stored in the memory to perform the following operations:
Acquiring acupuncture data of a plurality of patients, wherein the acupuncture data comprises a plurality of data; encoding each data in the acupuncture data of each patient to obtain row vectors corresponding to each data in the acupuncture data of each patient; obtaining a numerical matrix of each patient according to the acupuncture data of each patient and row vectors corresponding to each data in the acupuncture data; obtaining a covariance matrix of each patient according to the numerical matrix of each patient; equally dividing the covariance matrix of each patient to obtain a plurality of matrix blocks of the covariance matrix of each patient; according to the difference between the element values in the matrix blocks, the stability of each matrix block in the covariance matrix of each patient is obtained; according to the stability of the matrix blocks, a plurality of unstable matrix blocks and a plurality of stable matrix blocks of the covariance matrix of each patient are obtained; according to the stability of the matrix block and the element values in the matrix block, the stability of each element value in the covariance matrix of each patient is obtained; according to the element values in the unstable matrix block, the element values in the stable matrix block and the stability of the element values, obtaining an estimated value of each element value in the covariance matrix of each patient; obtaining an updated value of each element value in the covariance matrix of each patient according to the estimated quantity of the element value; replacing elements in the covariance matrix according to the updated values of the element values to obtain a final covariance matrix of each patient; performing dimension reduction on the final covariance matrix of each patient to obtain a dimension reduction matrix of each patient; training the neural network according to the dimension-reducing matrix, and obtaining a reference index of the rehabilitation date of the patient according to the trained neural network model; referring to fig. 3, fig. 3 is a flowchart of the embodiment for obtaining a reference index of a rehabilitation date of a patient.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the invention, but any modifications, equivalent substitutions, improvements, etc. within the principles of the present invention should be included in the scope of the present invention.
Claims (8)
1. An intelligent AI acupuncture model training method based on a neural network is characterized by comprising the following steps:
acquiring acupuncture data of a plurality of patients, wherein the acupuncture data comprises a plurality of data; encoding each data in the acupuncture data of each patient to obtain row vectors corresponding to each data in the acupuncture data of each patient; obtaining a numerical matrix of each patient according to the acupuncture data of each patient and row vectors corresponding to each data in the acupuncture data; obtaining a covariance matrix of each patient according to the numerical matrix of each patient;
Equally dividing the covariance matrix of each patient to obtain a plurality of matrix blocks of the covariance matrix of each patient; according to the difference between the element values in the matrix blocks, the stability of each matrix block in the covariance matrix of each patient is obtained; according to the stability of the matrix blocks, a plurality of unstable matrix blocks and a plurality of stable matrix blocks of the covariance matrix of each patient are obtained; according to the stability of the matrix block and the element values in the matrix block, the stability of each element value in the covariance matrix of each patient is obtained;
According to the element values in the unstable matrix block, the element values in the stable matrix block and the stability of the element values, obtaining an estimated value of each element value in the covariance matrix of each patient; obtaining an updated value of each element value in the covariance matrix of each patient according to the estimated quantity of the element value; replacing elements in the covariance matrix according to the updated values of the element values to obtain a final covariance matrix of each patient; performing dimension reduction on the final covariance matrix of each patient to obtain a dimension reduction matrix of each patient;
training the neural network according to the dimension-reducing matrix, and obtaining a reference index of the rehabilitation date of the patient according to the trained neural network model;
the method for obtaining the stability of each matrix block in the covariance matrix of each patient according to the difference between element values in the matrix blocks comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; marking any matrix block of the target covariance matrix as a target matrix block;
；
In the method, in the process of the invention, For the/>, in the target matrix blockIndividual element values,/>For the/>, in the target matrix blockIndividual element values,/>To take absolute value,/>For the number of element values in the target matrix block,/>Stability for the target matrix block;
according to the element values in the unstable matrix block, the element values in the stable matrix block and the stability of the element values, the estimation of each element value in the covariance matrix of each patient is obtained, and the method comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; recording any unstable matrix block of the target covariance matrix as a target unstable matrix block;
；
In the method, in the process of the invention, For the/>, in the target unstable matrix blockStability of individual element values; /(I)Is an exponential function with a natural constant as a base; /(I)For the number of element values in the target unstable matrix block,/>In the unstable matrix block for the targetIndividual element values,/>Is a normalization function; /(I)For the/>, in the target unstable matrix blockAn estimate of the individual element values;
for a stable matrix block of the target covariance matrix, the estimator of each element value in the stable matrix block is 0.
2. The training method of intelligent AI acupuncture model based on neural network according to claim 1, wherein the obtaining of a plurality of unstable matrix blocks and a plurality of stable matrix blocks of covariance matrix of each patient according to the stability of matrix blocks comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix;
Presetting a first threshold value, acquiring stability of each matrix block in a target covariance matrix, performing linear normalization processing on stability of all matrix blocks in the target covariance matrix, taking the obtained result as normalized stability of each matrix block in the target covariance matrix, taking the matrix blocks with normalized stability smaller than the first threshold value in the target covariance matrix as unstable matrix blocks of the target covariance matrix, and taking the matrix blocks with normalized stability larger than or equal to the first threshold value in the target covariance matrix as stable matrix blocks of the target covariance matrix.
3. The training method of intelligent AI acupuncture model based on neural network according to claim 1, wherein the obtaining the stability of each element value in the covariance matrix of each patient according to the stability of the matrix block and the element values in the matrix block comprises the following specific steps:
marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; marking any matrix block of the target covariance matrix as a target matrix block; obtaining a Z score of each element value in the target matrix block through a Z-score algorithm; and carrying out linear normalization processing on the Z scores of all the element values in the target matrix block, and taking the product of the normalized Z score of each element value in the target matrix block and the stability of the target matrix block as the stability of each element value in the target matrix block.
4. The training method of intelligent AI acupuncture model based on neural network according to claim 1, wherein the obtaining the updated value of each element value in the covariance matrix of each patient according to the estimated value of each element value comprises the following specific steps:
Marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; 1 is combined with the first of the target covariance matrices Sum of the estimates of the individual element values, and the/>, in the target covariance matrixThe product of the individual element values as the/>, in the target covariance matrixUpdated values of the individual element values.
5. The training method of intelligent AI acupuncture model based on neural network according to claim 1, wherein the replacing the elements in the covariance matrix according to the updated values of the element values to obtain the final covariance matrix of each patient comprises the following specific steps:
Marking any patient as a target patient; marking the covariance matrix of the target patient as a target covariance matrix; and acquiring an updated value of each element value in the target covariance matrix, replacing each element value in the target covariance matrix with a corresponding updated value to obtain a matrix, and recording the matrix as a final covariance matrix of the target patient.
6. The training method of intelligent AI acupuncture model based on neural network according to claim 1, wherein the obtaining the numerical matrix of each patient according to the acupuncture data of each patient and the row vector corresponding to each data in the acupuncture data comprises the following specific steps:
Marking any patient as a target patient, acquiring row vectors corresponding to each data in the acupuncture data of the target patient, and longitudinally splicing the row vectors corresponding to all data in the acupuncture data of the target patient to obtain an initial matrix of the target patient; acquiring all numerical data in acupuncture data of a target patient; and marking the product result of all the numerical data in the acupuncture data of the target patient as the comprehensive product parameter of the target patient, multiplying the comprehensive product parameter of the target patient with the initial matrix of the target patient to obtain a matrix, and marking the matrix as the numerical matrix of the target patient.
7. The training method of intelligent AI acupuncture model based on neural network according to claim 1, wherein the encoding each data in the acupuncture data of each patient to obtain a row vector corresponding to each data in the acupuncture data of each patient comprises the following specific steps:
And marking any One patient as a target patient, marking any One data in the acupuncture data of the target patient as target data, and Encoding the target data by using an One-hot Encoding algorithm to obtain row vectors corresponding to the target data.
8. A neural network based intelligent AI acupuncture model training system, the system comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program when executed by the processor performs the steps of a neural network based intelligent AI acupuncture model training method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410332261.7A CN117936103B (en) | 2024-03-22 | 2024-03-22 | Intelligent AI acupuncture model training system and method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410332261.7A CN117936103B (en) | 2024-03-22 | 2024-03-22 | Intelligent AI acupuncture model training system and method based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117936103A CN117936103A (en) | 2024-04-26 |
| CN117936103B true CN117936103B (en) | 2024-05-28 |
Family
ID=90770664
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410332261.7A Active CN117936103B (en) | 2024-03-22 | 2024-03-22 | Intelligent AI acupuncture model training system and method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117936103B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2006113747A2 (en) * | 2005-04-19 | 2006-10-26 | Prediction Sciences Llc | Diagnostic markers of breast cancer treatment and progression and methods of use thereof |
| CN107767946A (en) * | 2017-09-26 | 2018-03-06 | 浙江工业大学 | Breast cancer diagnosis system based on PCA (principal component analysis) and PSO-KE (particle swarm optimization-Key) L M (model-based regression) models |
| CN108310633A (en) * | 2018-02-01 | 2018-07-24 | 核工业四六医院 | A kind of intelligent paralytic patient recovering aid system |
| CN112370066A (en) * | 2020-09-30 | 2021-02-19 | 北京工业大学 | Brain-computer interface method of stroke rehabilitation system based on generation of countermeasure network |
| CN112466468A (en) * | 2020-11-27 | 2021-03-09 | 深圳市锦欣医疗科技创新中心有限公司 | Immune level prediction system, device and storage medium based on sparse coding |
| CN114023395A (en) * | 2021-11-29 | 2022-02-08 | 上海交通大学医学院附属瑞金医院 | Method and device for predicting mycophenolic acid drug exposure of kidney transplantation patient |
| CN114169375A (en) * | 2021-12-11 | 2022-03-11 | 福州大学 | Myoelectric gesture recognition method based on strength-independent robust features |
| CN114724705A (en) * | 2022-04-06 | 2022-07-08 | 郑州轻工业大学 | Esophageal squamous carcinoma survival prediction method based on improved ant lion algorithm and BP neural network |
| CN114783611A (en) * | 2022-06-22 | 2022-07-22 | 新泰市中医医院 | Neural recovered action detecting system based on artificial intelligence |
| WO2023225037A1 (en) * | 2022-05-17 | 2023-11-23 | Pisner Derek | Connectome ensemble transfer learning |
-
2024
- 2024-03-22 CN CN202410332261.7A patent/CN117936103B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2006113747A2 (en) * | 2005-04-19 | 2006-10-26 | Prediction Sciences Llc | Diagnostic markers of breast cancer treatment and progression and methods of use thereof |
| CN107767946A (en) * | 2017-09-26 | 2018-03-06 | 浙江工业大学 | Breast cancer diagnosis system based on PCA (principal component analysis) and PSO-KE (particle swarm optimization-Key) L M (model-based regression) models |
| CN108310633A (en) * | 2018-02-01 | 2018-07-24 | 核工业四六医院 | A kind of intelligent paralytic patient recovering aid system |
| CN112370066A (en) * | 2020-09-30 | 2021-02-19 | 北京工业大学 | Brain-computer interface method of stroke rehabilitation system based on generation of countermeasure network |
| CN112466468A (en) * | 2020-11-27 | 2021-03-09 | 深圳市锦欣医疗科技创新中心有限公司 | Immune level prediction system, device and storage medium based on sparse coding |
| CN114023395A (en) * | 2021-11-29 | 2022-02-08 | 上海交通大学医学院附属瑞金医院 | Method and device for predicting mycophenolic acid drug exposure of kidney transplantation patient |
| CN114169375A (en) * | 2021-12-11 | 2022-03-11 | 福州大学 | Myoelectric gesture recognition method based on strength-independent robust features |
| CN114724705A (en) * | 2022-04-06 | 2022-07-08 | 郑州轻工业大学 | Esophageal squamous carcinoma survival prediction method based on improved ant lion algorithm and BP neural network |
| WO2023225037A1 (en) * | 2022-05-17 | 2023-11-23 | Pisner Derek | Connectome ensemble transfer learning |
| CN114783611A (en) * | 2022-06-22 | 2022-07-22 | 新泰市中医医院 | Neural recovered action detecting system based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117936103A (en) | 2024-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107730474B (en) | Image processing method, processing device and processing equipment | |
| CN109166130B (en) | Image processing method and image processing device | |
| CN110096968B (en) | Ultra-high-speed static gesture recognition method based on depth model optimization | |
| Dear et al. | An approach for assessing publication bias prior to performing a meta-analysis | |
| CN110659665B (en) | Model construction method of different-dimension characteristics and image recognition method and device | |
| Taylan et al. | A new approach to multivariate adaptive regression splines by using Tikhonov regularization and continuous optimization | |
| CN110175168A (en) | A kind of time series data complementing method and system based on generation confrontation network | |
| CN116051574A (en) | Semi-supervised segmentation model construction and image analysis method, device and system | |
| CN112000772B (en) | Sentence-to-semantic matching method based on semantic feature cube and oriented to intelligent question and answer | |
| CN112000770B (en) | Semantic feature graph-based sentence semantic matching method for intelligent question and answer | |
| CN110335276B (en) | Medical image segmentation model, method, storage medium and electronic device | |
| CN105426929A (en) | Object shape alignment device, object processing device and methods thereof | |
| CN103914527B (en) | Graphic image recognition and matching method based on genetic programming algorithms of novel coding modes | |
| CN109886281A (en) | One kind is transfinited learning machine color image recognition method based on quaternary number | |
| CN109754357B (en) | Image processing method, processing device and processing equipment | |
| CN115310782A (en) | Power consumer demand response potential evaluation method and device based on neural turing machine | |
| CN114049501B (en) | Image description generation method, system, medium and device for fusing bundle search | |
| CN111310787A (en) | Brain function network multi-core fuzzy clustering method based on stacked encoder | |
| CN117936103B (en) | Intelligent AI acupuncture model training system and method based on neural network | |
| CN112528077B (en) | Video face retrieval method and system based on video embedding | |
| CN113449672A (en) | Remote sensing scene classification method and device based on bilinear twin framework | |
| CN111160161B (en) | Self-learning face age estimation method based on noise elimination | |
| CN116311472A (en) | Micro-expression recognition method and device based on multi-level graph convolution network | |
| CN109636600A (en) | A kind of investment combination method about subtracted using dimension | |
| CN116612860A (en) | Cerebral apoplexy training method and training system based on TST deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |