CN114334031A - Improved processing method, device, equipment and storage medium for pressing parameters - Google Patents
Improved processing method, device, equipment and storage medium for pressing parameters Download PDFInfo
- Publication number
- CN114334031A CN114334031A CN202111575591.1A CN202111575591A CN114334031A CN 114334031 A CN114334031 A CN 114334031A CN 202111575591 A CN202111575591 A CN 202111575591A CN 114334031 A CN114334031 A CN 114334031A
- Authority
- CN
- China
- Prior art keywords
- input vector
- sample input
- sample
- vector
- predicted
- 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.)
- Granted
Links
- 238000003825 pressing Methods 0.000 title claims abstract description 111
- 238000003860 storage Methods 0.000 title claims abstract description 23
- 238000003672 processing method Methods 0.000 title claims abstract description 12
- 239000013598 vector Substances 0.000 claims abstract description 301
- 239000003814 drug Substances 0.000 claims abstract description 66
- 238000000034 method Methods 0.000 claims abstract description 53
- 238000005457 optimization Methods 0.000 claims abstract description 53
- 229940079593 drug Drugs 0.000 claims abstract description 49
- 238000004364 calculation method Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 14
- 238000013527 convolutional neural network Methods 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 abstract description 4
- 238000004422 calculation algorithm Methods 0.000 description 18
- 239000000463 material Substances 0.000 description 14
- 230000008569 process Effects 0.000 description 9
- 238000002922 simulated annealing Methods 0.000 description 9
- 238000007906 compression Methods 0.000 description 8
- 239000008186 active pharmaceutical agent Substances 0.000 description 7
- 230000006835 compression Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 239000000546 pharmaceutical excipient Substances 0.000 description 6
- 238000004590 computer program Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 239000002245 particle Substances 0.000 description 4
- 230000002085 persistent effect Effects 0.000 description 4
- XOFYZVNMUHMLCC-ZPOLXVRWSA-N prednisone Chemical compound O=C1C=C[C@]2(C)[C@H]3C(=O)C[C@](C)([C@@](CC4)(O)C(=O)CO)[C@@H]4[C@@H]3CCC2=C1 XOFYZVNMUHMLCC-ZPOLXVRWSA-N 0.000 description 4
- 238000005094 computer simulation Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 239000002671 adjuvant Substances 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- FBPFZTCFMRRESA-KVTDHHQDSA-N D-Mannitol Chemical compound OC[C@@H](O)[C@@H](O)[C@H](O)[C@H](O)CO FBPFZTCFMRRESA-KVTDHHQDSA-N 0.000 description 1
- 229930195725 Mannitol Natural products 0.000 description 1
- 238000000137 annealing Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000007907 direct compression Methods 0.000 description 1
- 238000009509 drug development Methods 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 235000010355 mannitol Nutrition 0.000 description 1
- 239000000594 mannitol Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The application relates to an improved processing method, a device, equipment and a storage medium for pressing parameters, wherein the method comprises the following steps: obtaining a sample input vector and a preset sample target output vector aiming at the active molecules of the drug to be pressed, performing optimization iterative computation on the sample input vector by using an optimizer and a predicted pressing computing device until the difference value between the sample predicted output vector corresponding to the current sample input vector and the sample target output vector is calculated to be smaller than a first preset threshold value, and taking the current sample input vector as the target pressing parameter of the active molecules of the drug to be pressed. The medicine pressing parameters can be automatically adjusted and optimized without adopting a manual experiment trial and error method for repeated adjustment, so that the adjusting and optimizing efficiency is greatly improved.
Description
Technical Field
The present application relates to the field of drug development technologies, and in particular, to an improved processing method, device, apparatus, and storage medium for drug compression parameters.
Background
The methods, equipment, excipients and Active Pharmaceutical Ingredients (API) used in the compression process all affect the properties of the target tablet. In the past, experimenters continuously adjust the pressing parameters to manufacture target tablets by an experimental trial and error method, but the experimental trial and error method has the defects of low efficiency, high cost, difficulty in control, dependence on expert experience and the like.
With the development of computer technology, experimenters have begun to attempt to simulate the pressing process using a computer, which has many advantages. First, computer simulation can reduce decision time; second, computer simulation can save costs; third, the results can be controlled and analyzed by computer simulation.
Currently, a medicine pressing simulation software system is available on the market, and the software system is characterized in that a computer is used for simulating a medicine pressing process. However, if the pressing parameters need to be adjusted to find an improved method for a target tablet, an experimental trial and error method still needs to be used, and only the experimental trial and error is performed in the pressing simulation system, so that the method still depends on the experience of experimenters to a great extent, and the pressing simulation software system is only used as an auxiliary tool for the experimenters.
Disclosure of Invention
In order to solve or partially solve the problems in the related art, the application provides an improved processing method, device, equipment and storage medium for pressing, which can automatically adjust and optimize pressing parameters and improve the optimization efficiency.
A first aspect of the present application provides an improved processing method for pressing parameters, including:
obtaining a sample input vector aiming at the active molecules of the drug to be pressed;
acquiring a preset sample target output vector aiming at the active molecules of the drug to be pressed;
performing optimization iterative computation on the sample input vector by using a constructed predicted pressure calculation device and an optimizer until the difference value between the sample predicted output vector corresponding to the current sample input vector calculated by the predicted pressure calculation device and the preset sample target output vector is smaller than a first preset threshold, taking the current sample input vector as a target pressure parameter of the active molecule of the drug to be pressed, and ending the optimization iterative computation;
and outputting the target pressing parameters.
Preferably, the sample input vector includes a fixed input vector and an adjustable input vector, and when the sample input vector is subjected to optimization iterative computation, the adjustable input vector is adjusted, and the adjusted adjustable input vector and the fixed input vector are combined to form the current sample input vector.
Preferably, the performing optimized iterative computation on the sample input vector by using the constructed predicted pressure calculation device and the optimizer until the difference value between the sample predicted output vector corresponding to the current sample input vector calculated by the predicted pressure calculation device and the preset sample target output vector is smaller than a first preset threshold, and ending the optimized iterative computation by using the current sample input vector as the target pressure parameter of the active molecule of the drug to be pressed includes:
setting a termination temperature and a plurality of rounds of iteration bodies, wherein each round of iteration body is provided with an initial temperature;
circularly executing the current iteration body, and circularly executing the next iteration body if the initial temperature of the current iteration body is greater than or equal to the termination temperature and the iteration times of the iteration body overflow;
the iteration body is as follows: and randomly adjusting the sample input vector by using the optimizer, receiving the sample input vector after random adjustment as the current sample input vector according to a preset rule, if the difference value between the sample predicted output vector corresponding to the current sample input vector and the preset sample target output vector calculated by the predicted pressing calculator is smaller than a first preset threshold value, taking the current sample input vector as a target pressing parameter of the active molecule of the drug to be pressed, and finishing the optimization iterative calculation.
Preferably, when the initial temperature of the iteration body of the round is the initial temperature of the iteration body of the previous round, the attenuation coefficient is in a range of 0-1.
Preferably, the randomly adjusting the sample input vector by the optimizer includes:
and randomly generating a random number by using the optimizer, and adding the current sample input vector and the random number to obtain the sample input vector after random adjustment.
Preferably, the receiving the sample input vector after random adjustment according to a preset rule as the current sample input vector includes:
calculating a difference between sample predicted output vectors corresponding to the sample input vectors before and after the random adjustment by using the predicted pressing calculator;
if the difference is smaller than a second preset threshold value, taking the sample input vector after random adjustment as the current sample input vector;
and if the difference is larger than or equal to the second preset threshold value, taking the sample input vector after random adjustment as the current sample input vector according to Metropolis criterion probability.
Preferably, the construction of the predicted compression calculator includes:
acquiring original data, wherein the original data comprises an original input vector and an original output vector, and the original input vector and the original output vector have a corresponding relation;
and inputting the original data into a deep convolution neural network to construct a predicted pressing calculator.
Preferably, the inputting the raw data into the deep convolutional neural network to construct a predicted pressure drug calculator includes:
dividing the original data into training original data and testing original data according to a preset division method;
inputting the training original data into the deep convolution network to construct an initial prediction pressing calculator;
and verifying the test original data by utilizing the initial predicted pressing calculator to construct the predicted pressing calculator.
A second aspect of the present application provides an improved processing apparatus for pressing parameters, comprising:
the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a sample input vector aiming at a medicinal active molecule to be pressed;
the second acquisition module is used for acquiring a preset sample target output vector aiming at the active molecules of the drug to be pressed;
the optimization module is used for performing optimization iterative computation on the sample input vector by using a constructed predicted pressure calculation device and an optimizer until the difference value between the sample predicted output vector corresponding to the current sample input vector calculated by the predicted pressure calculation device and the preset sample target output vector is smaller than a first preset threshold value, taking the current sample input vector as a target pressure parameter of the active molecule of the drug to be pressed, and ending the optimization iterative computation;
and the output module is used for outputting the target pressing parameters.
Preferably, the sample input vector includes a fixed input vector and an adjustable input vector, and the optimization module performs an optimization iterative computation on the sample input vector, including: and adjusting the adjustable input vector, and combining the adjusted adjustable input vector and the fixed input vector to form the current sample input vector.
Preferably, the optimization module includes:
the device comprises an initialization unit, a control unit and a control unit, wherein the initialization unit is used for setting a termination temperature and a plurality of rounds of iteration bodies, and each round of iteration bodies is provided with an initial temperature;
the iteration unit is used for circularly executing the current iteration body, and circularly executing the next iteration body if the initial temperature of the current iteration body is greater than or equal to the termination temperature and the iteration times of the iteration body overflow;
the iteration body is as follows: and randomly adjusting the sample input vector by using the optimizer, receiving the sample input vector after random adjustment as the current sample input vector according to a preset rule, if the difference value between the sample predicted output vector corresponding to the current sample input vector and the preset sample target output vector calculated by the predicted pressing calculator is smaller than a first preset threshold value, taking the current sample input vector as a target pressing parameter of the active molecule of the drug to be pressed, and finishing the optimization iterative calculation.
A third aspect of the present application provides an electronic device comprising:
a processor; and
a memory having executable code stored thereon, which when executed by the processor, causes the processor to perform the method as described above.
A fourth aspect of the present application provides a computer-readable storage medium having stored thereon executable code, which, when executed by a processor of an electronic device, causes the processor to perform the method as described above.
The technical scheme provided by the application can comprise the following beneficial effects:
according to the technical scheme, a sample input vector and a preset sample target output vector for active molecules of the drug to be pressed are obtained, the sample input vector is subjected to optimization iterative calculation by using an optimizer and a predicted pressing calculator until the difference value between the sample predicted output vector corresponding to the current sample input vector and the preset sample target output vector is calculated to be smaller than a first preset threshold value, and the current sample input vector is used as a target pressing parameter of the active molecules of the drug to be pressed. The medicine pressing parameters can be automatically adjusted and optimized without adopting a manual experiment trial and error method for repeated adjustment, so that the adjusting and optimizing efficiency is greatly improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The foregoing and other objects, features and advantages of the application will be apparent from the following more particular descriptions of exemplary embodiments of the application as illustrated in the accompanying drawings wherein like reference numbers generally represent like parts throughout the exemplary embodiments of the application.
FIG. 1 is a schematic flow chart of an improved processing method for pressing parameters according to an embodiment of the present application;
FIG. 2 is a schematic flow chart illustrating a method for improved handling of pressing parameters according to another embodiment of the present application;
FIG. 3 is a block flow diagram illustrating an optimization iterative computation according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of an apparatus for processing the pressing parameters according to an embodiment of the present disclosure;
FIG. 5 is a schematic structural diagram of an apparatus for processing the pressing parameters according to another embodiment of the present application;
fig. 6 is a schematic structural diagram of an electronic device shown in an embodiment of the present application.
Detailed Description
Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. While embodiments of the present application are illustrated in the accompanying drawings, it should be understood that the present application may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It should be understood that although the terms "first," "second," "third," etc. may be used herein to describe various information, these information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present application. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
In the prior art, experimenters continuously fine-tune the pressing parameters to manufacture target tablets by an experimental trial-and-error method, but the experimental trial-and-error method has the defects of low efficiency, high cost, difficulty in control, dependence on expert experience and the like. In recent years, a computer is used for simulating a medicine pressing process, but when an improved method for a target tablet needs to be found, an experimental trial and error method still needs to be used, only the experimental trial and error is carried out in a medicine pressing simulation system, the experience of experimenters is still greatly depended, and a medicine pressing simulation software system is only used as an auxiliary tool for the experimenters.
In view of the above problems, the present application provides an improved processing method, apparatus, device and storage medium for pressing parameters, which can automatically adjust and optimize the pressing parameters and improve the tuning efficiency. In order to facilitate understanding of the technical solutions of the present application, the technical solutions of the present application are described in detail below with reference to the accompanying drawings.
Fig. 1 is a schematic flow chart illustrating an improved processing method for pressing parameters according to an embodiment of the present application.
Referring to fig. 1, an improved processing method of pressing parameters includes the following steps:
and S11, acquiring a sample input vector aiming at the drug active molecules to be pressed.
In the process of medicine production, the raw materials are almost produced by adopting a combination mode of active molecules of medicines (API for short), auxiliary materials and medicine pressing equipment. Thus, the sample input vector may contain three types of information, a first type for pharmaceutically active molecules, a second type for excipient molecules and a third type for a compression device. That is, the sample input vector may contain the pharmaceutically active molecule property parameters to be compressed, the excipient molecule property parameters, and the parameters related to the compression device.
And S12, acquiring a preset sample target output vector aiming at the active molecules of the drug to be pressed.
Wherein the preset sample target output vector is a preset target tablet property vector. The user can preset a target tablet property vector for the active molecule of the drug to be pressed according to the actually required tablet property.
And S13, performing optimization iterative computation on the sample input vector by using the constructed predicted pressure drug calculator and optimizer until the difference value between the sample predicted output vector corresponding to the current sample input vector and the preset sample target output vector calculated by the predicted pressure drug calculator is smaller than a first preset threshold, taking the current sample input vector as a target pressure drug parameter of the active molecule of the drug to be pressed, and finishing the optimization iterative computation.
The purpose of constructing the optimizer is to perform optimization iterative computation, and replace an old solution with a new solution, so as to compute a global optimal solution.
And inputting the current sample input vector into the predicted pressing calculator to obtain a sample predicted output vector corresponding to the current sample input vector. The current sample input vector corresponds to the combination of the pharmaceutically active molecule + the adjuvant + the pressing device.
And judging whether the difference value between the sample prediction output vector corresponding to the sample input vector and the sample target output vector is smaller than a first preset threshold value (namely judging whether the prediction tablet is close to the target tablet). If the predicted tablet is smaller than the first preset threshold, the predicted tablet is judged to be equivalent to the target tablet, and the current sample input vector is used as a target pressing parameter of the active molecules of the drug to be pressed (namely, the predicted tablet prepared according to the combination of the current active molecules of the drug, auxiliary materials and pressing equipment is equivalent to the target tablet). If the number of the active pharmaceutical molecules is larger than or equal to the first preset threshold, the predicted tablet is judged to be different from the target tablet (namely, the predicted tablet and the target tablet are different according to the combination of the current active pharmaceutical molecules, auxiliary materials and medicine pressing equipment).
And changing the current sample input vector by using an optimizer, performing optimization iterative computation on the current sample input vector until the difference value between the sample prediction output vector corresponding to the current sample input vector and the preset sample target output vector is calculated to be smaller than a first preset threshold value, and taking the current sample input vector as the target pressing parameter of the active molecule of the drug to be pressed. The difference value may be an absolute difference value or a ratio, and is not limited herein.
It should be noted that the sample input vector includes a fixed input vector and an adjustable input vector. The fixed input vector corresponds to the pharmaceutically active molecule, which is fixed and invariant. The adjustable input vector corresponds to the parameters of the auxiliary materials and the pressing equipment, and the parameters of the auxiliary materials and the pressing equipment are adjustable. And when the sample input vector is subjected to optimization iterative computation, the adjustable input vector is adjusted, and the adjusted adjustable input vector and the fixed input vector are combined to form the current sample input vector.
And S14, outputting the target pressing parameters.
According to the technical scheme, a sample input vector and a preset sample target output vector for the active molecules of the drug to be pressed are obtained, optimization iterative calculation is carried out on the sample input vector by using an optimizer and a predicted pressing calculator until the difference value between the sample predicted output vector corresponding to the current sample input vector and the sample target output vector is calculated to be smaller than a first preset threshold value, and the current sample input vector is used as the target pressing parameter of the active molecules of the drug to be pressed. The medicine pressing parameters can be automatically adjusted and optimized without adopting a manual experiment trial and error method for repeated adjustment, so that the adjusting and optimizing efficiency is greatly improved.
Fig. 2 is a flow chart illustrating a method for improving direct compression according to another embodiment of the present application.
Referring to fig. 2, an improved processing method of pressing parameters includes the following steps:
before the method is executed, a predicted pressure calculator is constructed in advance, and the construction of the predicted pressure calculator can be constructed in the following way:
collecting original data, wherein the original data comprises an original input vector and an original output vector, and the original input vector and the original output vector have a corresponding relation.
The original data comprises an original input vector, and due to the drug production link, the raw materials are almost produced by adopting a combination mode of active drug molecules (API for short, hereinafter, the same), auxiliary materials and pressing equipment. Thus, the original input vector may contain three types of information, a first type for pharmaceutically active molecules, a second type for excipient molecules and a third type for a compression device. That is, the sample input vector may contain the pharmaceutically active molecule property parameters to be compressed, the excipient molecule property parameters, and the parameters related to the compression device.
Table 1 shows an information table of the original input vector — the pharmaceutically active molecule, and table 1 may contain information related to a descriptor of the pharmaceutically active molecule, a fingerprint, an API type, a powder particle shape, a powder particle deformation characteristic, a mechanical modulus, brittleness, hardness, and the like.
Table 1 original input vector-information table of pharmaceutically active molecules
Table 2 shows an information table of the original input vector-the auxiliary material, and table 2 may contain information related to descriptors, fingerprints, volume ratios (API: auxiliary material), and the like of the auxiliary material.
| Accessory descriptor | Fingerprint symbol of auxiliary material | Volume ratio (API: auxiliary material) |
| Mannitol | 2936778132 | 1:9 |
TABLE 2 information Table of original input vector-adjuvants
It is understood that the descriptor and fingerprint of the molecule may have other expressions besides the above-shown expression, and is not limited herein.
Table 3 shows an information table of the original input vector-pressing apparatus, and table 3 may contain information on the apparatus name, the die height, the die circumference, the punch shape, the punch diameter, the number of revolutions, the temperature, the relative humidity, and the like.
| Device name | Height of the mould | Punch shape | Diameter of punch | Number of revolutions | Temperature of |
| Stylcam | 22.22 | Circular shape | 10 | 2000 | 25℃ |
TABLE 3 original input vector-information sheet of pressing apparatus
The information in tables 1, 2 and 3 together form the original input vector, i.e. the pharmaceutically active molecules, excipients and the pressing equipment are equivalent to the pressing parameters, and the data in tables 1, 2 and 3 can be collected through previous experimental records and filled in tables 1, 2 and 3 for sorting. The descriptors and fingerprints of the molecules can be calculated using chem.
Meanwhile, the original data comprises an original output vector corresponding to the performance of the target tablet. Table 4 shows a table of information of the original output vector-target tablet, and table 4 may contain information on the target tablet hardness, the target tablet macro-friability, the target tablet disintegration time, the target tablet Hausner ratio, the target tablet surface area (cm2), the target tablet mass (mg), the target tablet thickness (mm), the target tablet density (mg/cm2), the target tablet tensile strength (N/m2 ═ MPa), and the like.
Table 4 original output vector-target tablet information table
The original input vectors and the original output vectors have corresponding relations, i.e. a set of original input vectors corresponds to a set of original output vectors. The original output vectors are also collected from past experimental records and filled into table 4 for sorting.
And inputting the original data into a deep convolution neural network to construct a predicted pressing calculator.
The method comprises the steps of collecting original data, sorting the original data according to a table mode, inputting the original data into a deep Convolutional Neural Network (CNN), constructing a predicted pressing medicine calculator after the deep Convolutional Neural Network (CNN) is trained and learned as the original data comprises an original input vector and an original output vector which have a corresponding relation, wherein the predicted pressing medicine calculator is actually a deep convolutional neural network model, and when a group of original input vectors are input, the predicted pressing medicine calculator can output a group of original output vectors corresponding to the current original input vectors.
In order to make the prediction capability of the predicted pressure-drug calculator more accurate, the original data is divided into training original data and testing original data according to a preset division method, for example, 80% to 90% of the original data can be divided into training original data. 10% -20% of the raw data was divided into test raw data. Training original data to be input into a deep Convolutional Neural Network (CNN) to construct an initial predicted pressure medicament calculator, wherein the initial predicted pressure medicament calculator has a prediction function, namely a function of inputting a group of original input vectors and outputting a group of original output vectors corresponding to the current original input vectors. And inputting the test original data into the predicted pressing calculator for verification, thereby constructing the predicted pressing calculator with superior prediction function.
It should be noted that, the original data can be divided into training original data and testing original data by using a K-fold cross-validation method (where K is greater than or equal to 5). The specific principle of the K-fold cross-validation method is as follows: 1. the total original data set S is divided into k disjoint subsets, and assuming that the number of training samples in S is m, each subset has m/k training samples, and the corresponding subset is called { S1, S2, …, sk }. 2. And taking out one from the divided subsets each time as a test original data set, and taking the other k-1 as a training original data set. 3. Inputting k-1 training original data sets into a deep Convolutional Neural Network (CNN), constructing an initial predicted pressing calculator, taking out a testing original data set to verify the initial predicted pressing calculator, and executing the steps for k times to obtain the predicted pressing calculator.
And S21, acquiring a sample input vector aiming at the drug active molecules to be pressed.
The step S21 can be referred to the description of the step S11, and is not described herein.
And S22, acquiring a preset sample target output vector aiming at the active molecules of the drug to be pressed.
Besides the construction of the predicted pressure calculation device, an optimizer needs to be constructed, and the optimizer and the predicted pressure calculation device perform optimization iterative calculation on the sample input vector. The optimizer can be constructed by using a simulated annealing algorithm, a particle swarm algorithm, an evolution algorithm and the like. The embodiment adopts a simulated annealing algorithm to construct the optimizer.
Simulated Annealing (SA), which is a random optimization algorithm based on Monte-Carlo iterative solution strategy, starts from the similarity between the Annealing process of solid matter in physics and general combinatorial optimization problems. The simulated annealing algorithm starts from a certain high initial temperature, and randomly searches a global optimal solution of the objective function in a solution space by combining with the probability jump characteristic along with the continuous decrease of the temperature parameter, namely, the global optimal solution can jump out probabilistically in a local optimal solution and finally tends to be global optimal. The simulated annealing algorithm is an optimization algorithm which can effectively avoid trapping in a serial structure which is locally minimum and finally tends to global optimum by endowing a search process with time-varying probability jump property and finally tends to zero.
S23, setting a termination temperature and a plurality of rounds of iteration bodies, wherein each round of iteration body is provided with an initial temperature.
And S24, circularly executing the current iteration body, and circularly executing the next iteration body if the initial temperature of the current iteration body is greater than or equal to the termination temperature and the iteration times of the iteration body overflow.
Iteration body: and randomly adjusting the sample input vector by using an optimizer, receiving the sample input vector after random adjustment as a current sample input vector according to a preset rule, if the difference value between the sample predicted output vector corresponding to the current sample input vector and a preset sample target output vector is smaller than a first preset threshold value, calculating by using the current sample input vector as a target pressing parameter of the active molecules of the drug to be pressed by the predicted pressing calculator, and finishing the optimization iterative calculation.
It should be noted that the sample input vector includes a fixed input vector and an adjustable input vector. The fixed input vector corresponds to the pharmaceutically active molecule, which is fixed and invariant. The adjustable input vector corresponds to the parameters of the auxiliary materials and the pressing equipment, and the parameters of the auxiliary materials and the pressing equipment are adjustable. When the sample input vector is optimized and iterated, the optimizer adjusts the adjustable input vector and combines the adjusted adjustable input vector and the fixed input vector to form the current sample input vector.
In order to better understand the principle of the technical solution of the present embodiment, the following describes in detail the process of solving the sample target input vector (i.e. obtaining the target pressing parameter) by the optimizer constructed by the simulation algorithm and the predicted pressing calculator constructed based on the deep convolutional neural network CNN.
Referring to fig. 3, fig. 3 is a block flow diagram illustrating an optimization iteration calculation.
S31: input sample input vector ViDefining an objective function E (x), and inputting the current sample into a vector ViInputting the input vector into a target function E (x) to obtain a current sample input vector ViCorresponding prediction sample output vector E (V)i)。
S32: initializing the optimizer, setting a preset sample target output vector E (V) of the optimizert) End temperature Tf(e.g. termination temperature T)fAt 10 ℃ and a termination temperature TfThe value of (c) may be flexibly set in practical situations, and is not specifically limited), an iteration body number threshold k (for example, the iteration body number threshold k may be a positive integer greater than or equal to 1000), an attenuation coefficient α (the attenuation coefficient α is a number between 0 and 1, for example, the attenuation coefficient α is 0.98), a first preset threshold G (the first preset threshold G is not specifically limited), and a second preset threshold P (for example, the second preset threshold P is 0).
Wherein, the iterative body has a plurality of rounds, and each round of iterative body is provided with an initial temperature T0When the initial temperature T of the iteration body is in turn0Initial temperature T of last iteration0Attenuation coefficient α, e.g. initial temperature T of last iteration0At 1000 deg.C, the initial temperature T of the round iteration body01000 x 0.98 x 980 ℃. And the initial value i of the current iteration number of each round of iteration body is 0.
Setting a preset sample target output vector E (V)t) Termination temperature Tf10 ℃, the iteration body time threshold k is 1000 times, the attenuation coefficient alpha is 0.98, and the initial iteration body of the first roundInitial temperature T0＝1000℃。
S33: performing an iterative calculation;
s34: judging the initial temperature T of the current iteration body0Whether or not it is greater than the termination temperature Tf；
If the initial temperature T of the iteration body is in turn0Above the termination temperature Tf(i.e. T)0＞Tf) Step S35 is executed;
if the initial temperature T of the iteration body is in turn0Less than or equal to the termination temperature Tf(i.e. T)0＜＝Tf) And ending the algorithm, outputting a sample target input vector (namely the target pressing parameter) which cannot be obtained, and ending the algorithm.
S35: judging whether the current iteration times i are smaller than an iteration body time threshold k;
if the current iteration frequency i is smaller than the iteration body frequency threshold value k (i.e. i is less than k), the current sample is input into the vector ViInputting the sample to a predicted pressure calculator to obtain a current sample predicted output vector E (V)i) Calculating the current sample prediction output vector E (V)i) And a sample target output vector E (V)t) The difference Δ E of;
and if the current iteration number i is larger than or equal to the iteration body number threshold k (i.e. i > ═ k), entering the next iteration body.
S36: judging whether the absolute value of the difference value delta E is smaller than a first preset threshold value G or not;
if the absolute value of the difference value Delta E is smaller than a first preset threshold value G (namely Delta E is smaller than G), ending the algorithm, and inputting the current sample into the vector ViAs a sample target input vector Vt；
If the absolute value of the difference Δ E is greater than or equal to a first predetermined threshold G (i.e., Δ E ═ G), the vector V is input for the current sampleiRandom adjustment by calculating sample input vector V before and after random adjustment in predicted pressure-drug calculatoriCorresponding sample prediction output vector E (V)i) The difference dE between E and E (V)i After adjustment)-E(Vi Before adjustment) Step S37 is executed.
S37: judging whether the difference dE is smaller than a second preset threshold value P or not;
if the difference dE is smaller than a second predetermined threshold P (i.e., dE < P), the randomly adjusted sample is input into the vector Vi After adjustmentAs the current sample input vector ViAnd increments the current iteration number i by one (i.e., i ═ i +1), and the process returns to step S33;
if the difference dE is greater than or equal to a second predetermined threshold P (i.e., dE > -) the perturbed sample input vector data V is received according to Metropolis criterion probabilityi After adjustmentAs the current sample input vector ViAnd increments the current iteration number i by one (i.e., i ═ i +1), returning to the execution of step S33.
It should be noted that ending the algorithm means ending the process of optimizing the iterative computation. Randomly adjusting, namely changing the current sample input vector, randomly generating a random number (a predetermined range such as a number between 0 and 1, the predetermined range can be flexibly set according to actual conditions without specific limitation) within a predetermined range, and adding the current sample input vector and the random number to obtain a randomly adjusted sample input vector (namely E (V) (V))i After adjustment)＝E(Vi Before adjustment) + random number).
It should also be noted that Metropolis criterion probability is exp (-dE/T)0) Metropolis criterion probability is exp (-dE/T)0) Is received from the sample input vector V after perturbationi After perturbationAs the current sample input vector ViTo (1-exp (-dE/T)0) ) maintains the sample input vector V before random adjustmentiAs the current sample input vector Vi(i.e. maintaining the current sample input vector ViInvariant), the current sample is input into vector ViAs a target compression parameter.
And S25, outputting the target pressing parameters.
The step S25 can be referred to the related description in the step S14, and is not described herein again.
By the method, optimization iteration is performed by combining the optimizer and the predicted pressing calculation device, manual repeated adjustment of pressing parameters (namely manual repeated adjustment of the current sample input vector) is omitted, and pressing performance adjustment efficiency is improved.
In addition, the auxiliary materials and the pressing equipment can be simultaneously adjusted and optimized in parallel (the parallel adjustment is that the current sample input vector is continuously adjusted and optimized through a simulated annealing algorithm to find a global optimal solution), and the flux is greatly improved.
Moreover, a possible tuning parameter space is expanded, manual trial-and-error tuning depends on human experience, only a few parameter spaces acquired by experience and trial-and-error can be tried, and by utilizing a global optimization scheme based on a simulated annealing algorithm, a parameter space which is difficult to be touched by experimenters depending on experience can be explored, and a missed more optimal space (namely, a global optimal solution cannot be obtained due to the fact that the experimental personnel fall into a local optimal solution) caused by experience limitation of the experimenters is found.
Corresponding to the embodiment of the application function implementation method, the application also provides an improved processing device of the pressing parameters and a corresponding embodiment.
Fig. 4 is a schematic structural diagram of an improved processing device 40 for pressing parameters in the embodiment of the present application, the device including: a first obtaining module 410, a second obtaining module 420, an optimizing module 430, and an output module 440.
The first obtaining module 410 is used for obtaining a sample input vector for the pharmaceutically active molecule to be pressed.
The second obtaining module 420 is configured to obtain a preset sample target output vector for the active molecule of the drug to be pressed.
The optimization module 430 is configured to perform optimization iterative computation on the sample input vector by using a constructed predicted pressure calculation device and an optimization device until a difference value between a sample predicted output vector corresponding to the current sample input vector calculated by the predicted pressure calculation device and the preset sample target output vector is smaller than a first preset threshold, use the current sample input vector as a target pressure parameter of the active molecule of the drug to be pressed, and end the optimization iterative computation.
The output module 440 is configured to output the target pressing parameter.
In the device of this embodiment, a first obtaining module 410 obtains a sample input vector for a pharmaceutically active molecule to be compressed, a second obtaining module 420 obtains a preset sample target output vector for the pharmaceutically active molecule to be compressed, an optimizing module 430 combines an optimizer with a constructed predicted drug pressing calculator to perform optimization iterative calculation on the sample input vector until the predicted drug pressing calculator calculates that a difference value between a sample predicted output vector corresponding to a current sample input vector and the sample target output vector is less than a threshold, the current sample input vector is used as a target drug pressing parameter of the pharmaceutically active molecule to be compressed, the optimization iterative calculation is finished, and an output module 440 outputs the target drug pressing parameter. The improved pressing parameters can be obtained without adopting a manual experiment trial and error method for repeated adjustment, so that the adjusting efficiency can be improved.
Fig. 5 shows a schematic structural diagram of an improved processing device 40 for pressing parameters in another embodiment of the present application, which includes: a first obtaining module 410, a second obtaining module 420, an optimizing module 430, and an output module 440. Further, the optimization module 430 comprises an initialization unit 431 and an iteration unit 432.
The first obtaining module 410, the second obtaining module 420, the optimizing module 430, and the output module 440 may refer to the description in fig. 4, and are not described herein again.
In addition, the sample input vector includes a fixed input vector and an adjustable input vector, and the optimization module 430 performs an optimization iterative computation on the sample input vector, including: and adjusting the adjustable input vector, and combining the adjusted adjustable input vector and the fixed input vector to form the current sample input vector.
The initialization unit 431 is configured to set an end temperature and a plurality of iterations, each iteration having an initial temperature.
The iteration unit 432 is configured to execute the current iteration body in a loop, and if the initial temperature of the current iteration body is greater than or equal to the end temperature and the iteration times of the iteration body overflow, execute the next iteration body in a loop;
iteration body: and randomly adjusting the sample input vector by using an optimizer, receiving the sample input vector after random adjustment as a current sample input vector according to a preset rule, if the difference value between the sample predicted output vector corresponding to the current sample input vector and a preset sample target output vector is smaller than a first preset threshold value, calculating by using the current sample input vector as a target pressing parameter of the active molecules of the drug to be pressed by the predicted pressing calculator, and finishing the optimization iterative calculation. The difference value may be an absolute difference value or a ratio, and is not limited herein.
And when the initial temperature of the round iteration body is equal to the initial temperature of the previous round iteration body, the attenuation coefficient ranges from 0 to 1.
When the iteration unit 432 randomly adjusts the sample input vector using the optimizer: and randomly generating a random number, and adding the current sample input vector and the random number to obtain a randomly adjusted sample input vector.
When the iteration unit 432 receives the sample input vector after random adjustment as the current sample input vector according to the preset rule: calculating the difference between sample predicted output vectors corresponding to sample input vectors before and after random adjustment by using a predicted pressure medicament calculator;
if the difference is smaller than a second preset threshold, taking the sample input vector after random adjustment as the current sample input vector;
and if the difference is larger than or equal to a second preset threshold value, taking the sample input vector after random adjustment as the current sample input vector according to the Metropolis criterion probability.
In addition, the predicted pressing calculator is constructed by adopting a supervised learning method. Dividing original data into training original data and testing original data according to a preset division method; inputting training original data into a deep convolution network to construct an initial predicted pressing calculator; and verifying the original test data by using an initial predicted pressing calculator to construct a predicted pressing calculator.
The optimizer can be constructed by using a simulated annealing algorithm, a particle swarm algorithm, an evolution algorithm and the like. With regard to the apparatus in the above embodiments, the specific manner in which each module and unit performs operations has been described in detail in relation to the method embodiment corresponding to the apparatus, and will not be described in detail herein.
Referring to fig. 6, an electronic device 600 includes a processor 610 and a memory 620.
The Processor 610 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 620 may include various types of storage units, such as system memory, Read Only Memory (ROM), and permanent storage. Wherein the ROM may store static data or instructions that are required by the processor 610 or other modules of the computer. The persistent storage device may be a read-write storage device. The persistent storage may be a non-volatile storage device that does not lose stored instructions and data even after the computer is powered off. In some embodiments, the persistent storage device employs a mass storage device (e.g., magnetic or optical disk, flash memory) as the persistent storage device. In other embodiments, the permanent storage may be a removable storage device (e.g., floppy disk, optical drive). The system memory may be a read-write memory device or a volatile read-write memory device, such as a dynamic random access memory. The system memory may store instructions and data that some or all of the processors require at runtime. In addition, the memory 620 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (e.g., DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), magnetic and/or optical disks, as well. The memory 620 has stored thereon executable code that, when processed by the processor 610, may cause the processor 610 to perform some or all of the methods described above.
Furthermore, the method according to the present application may also be implemented as a computer program or computer program product comprising computer program code instructions for performing some or all of the steps of the above-described method of the present application.
Alternatively, the present application may also be embodied as a computer-readable storage medium (or non-transitory machine-readable storage medium or machine-readable storage medium) having executable code (or a computer program or computer instruction code) stored thereon, which, when executed by a processor of an electronic device (or server, etc.), causes the processor to perform part or all of the various steps of the above-described method according to the present application.
Having described embodiments of the present application, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (13)
1. An improved processing method for pressing parameters is characterized by comprising the following steps:
obtaining a sample input vector aiming at the active molecules of the drug to be pressed;
acquiring a preset sample target output vector aiming at the active molecules of the drug to be pressed;
performing optimization iterative computation on the sample input vector by using a constructed predicted pressure calculation device and an optimizer until the difference value between the sample predicted output vector corresponding to the current sample input vector calculated by the predicted pressure calculation device and the preset sample target output vector is smaller than a first preset threshold, taking the current sample input vector as a target pressure parameter of the active molecule of the drug to be pressed, and ending the optimization iterative computation;
and outputting the target pressing parameters.
2. The method according to claim 1, wherein the sample input vector comprises a fixed input vector and an adjustable input vector, and when performing the optimization iterative computation on the sample input vector, the adjustable input vector is adjusted, and the adjusted adjustable input vector and the fixed input vector are combined to form the current sample input vector.
3. The method of claim 1, wherein the performing optimization iterative computation on the sample input vector by using the constructed predicted pressure calculation device and optimization device until the predicted pressure calculation device calculates that a difference value between a sample predicted output vector corresponding to a current sample input vector and the preset sample target output vector is smaller than a first preset threshold value, and ending the optimization iterative computation by using the current sample input vector as a target pressure parameter of the active molecule of the drug to be pressed comprises:
setting a termination temperature and a plurality of rounds of iteration bodies, wherein each round of iteration body is provided with an initial temperature;
circularly executing the current iteration body, and circularly executing the next iteration body if the initial temperature of the current iteration body is greater than or equal to the termination temperature and the iteration times of the iteration body overflow;
the iteration body is as follows: and randomly adjusting the sample input vector by using the optimizer, receiving the sample input vector after random adjustment as the current sample input vector according to a preset rule, if the difference value between the sample predicted output vector corresponding to the current sample input vector and the preset sample target output vector calculated by the predicted pressing calculator is smaller than a first preset threshold value, taking the current sample input vector as a target pressing parameter of the active molecule of the drug to be pressed, and finishing the optimization iterative calculation.
4. The method according to claim 3, wherein the initial temperature of the iteration body in a previous round is an attenuation coefficient, wherein the attenuation coefficient is in a range of 0 to 1.
5. The method of claim 3, wherein randomly adjusting the sample input vector using the optimizer comprises:
and randomly generating a random number by using the optimizer, and adding the current sample input vector and the random number to obtain the sample input vector after random adjustment.
6. The method of claim 5, wherein receiving the randomly adjusted sample input vector as the current sample input vector according to a preset rule comprises:
calculating a difference between sample predicted output vectors corresponding to the sample input vectors before and after the random adjustment by using the predicted pressing calculator;
if the difference is smaller than a second preset threshold value, taking the sample input vector after random adjustment as the current sample input vector;
and if the difference is larger than or equal to the second preset threshold value, taking the sample input vector after random adjustment as the current sample input vector according to Metropolis criterion probability.
7. The method according to any one of claims 1 to 6, wherein the construction of the predicted pressure calculator comprises:
acquiring original data, wherein the original data comprises an original input vector and an original output vector, and the original input vector and the original output vector have a corresponding relation;
and inputting the original data into a deep convolution neural network to construct a predicted pressing calculator.
8. The method of claim 7, wherein inputting the raw data into a deep convolutional neural network constructs a predicted pressure calculator comprising:
dividing the original data into training original data and testing original data according to a preset division method;
inputting the training original data into the deep convolution network to construct an initial prediction pressing calculator;
and verifying the test original data by utilizing the initial predicted pressing calculator to construct the predicted pressing calculator.
9. An improved processing device for pressing parameters is characterized by comprising:
the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a sample input vector aiming at a medicinal active molecule to be pressed;
the second acquisition module is used for acquiring a preset sample target output vector aiming at the active molecules of the drug to be pressed;
the optimization module is used for performing optimization iterative computation on the sample input vector by using a constructed predicted pressure calculation device and an optimizer until the difference value between the sample predicted output vector corresponding to the current sample input vector calculated by the predicted pressure calculation device and the preset sample target output vector is smaller than a first preset threshold value, taking the current sample input vector as a target pressure parameter of the active molecule of the drug to be pressed, and ending the optimization iterative computation;
and the output module is used for outputting the target pressing parameters.
10. The apparatus of claim 9, wherein the sample input vector comprises a fixed input vector and an adjustable input vector, and wherein the optimization module performs an optimization iteration on the sample input vector, comprising: and adjusting the adjustable input vector, and combining the adjusted adjustable input vector and the fixed input vector to form the current sample input vector.
11. The apparatus of claim 9, wherein the optimization module comprises:
the device comprises an initialization unit, a control unit and a control unit, wherein the initialization unit is used for setting a termination temperature and a plurality of rounds of iteration bodies, and each round of iteration bodies is provided with an initial temperature;
the iteration unit is used for circularly executing the current iteration body, and circularly executing the next iteration body if the initial temperature of the current iteration body is greater than or equal to the termination temperature and the iteration times of the iteration body overflow;
the iteration body is as follows: and randomly adjusting the sample input vector by using the optimizer, receiving the sample input vector after random adjustment as the current sample input vector according to a preset rule, if the difference value between the sample predicted output vector corresponding to the current sample input vector and the preset sample target output vector is smaller than a second first preset threshold value, calculated by the predicted pressing calculator, taking the current sample input vector as a target pressing parameter of the active molecule of the drug to be pressed, and finishing the optimization iterative calculation.
12. An electronic device, comprising:
a processor; and
a memory having executable code stored thereon, which when executed by the processor, causes the processor to perform the method of any one of claims 1 to 8.
13. A computer readable storage medium having stored thereon executable code which, when executed by a processor of an electronic device, causes the processor to perform the method of any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111575591.1A CN114334031B (en) | 2021-12-21 | 2021-12-21 | Improved processing method, device, equipment and storage medium for drug pressing parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111575591.1A CN114334031B (en) | 2021-12-21 | 2021-12-21 | Improved processing method, device, equipment and storage medium for drug pressing parameters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114334031A true CN114334031A (en) | 2022-04-12 |
| CN114334031B CN114334031B (en) | 2024-10-25 |
Family
ID=81054329
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111575591.1A Active CN114334031B (en) | 2021-12-21 | 2021-12-21 | Improved processing method, device, equipment and storage medium for drug pressing parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114334031B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101432310A (en) * | 2006-04-26 | 2009-05-13 | B·布朗·梅尔松根有限公司 | Manufacture and use of modified polysaccharide chitosan bonds and a process to improve the preparation of HES-medicinal substance compounds |
| WO2020006571A1 (en) * | 2018-06-29 | 2020-01-02 | pulseData Inc. | Machine learning systems and methods for predicting risk of renal function decline |
| US20200050717A1 (en) * | 2018-08-09 | 2020-02-13 | Kabushiki Kaisha Toshiba | Optimization apparatus, simulation system and optimization method |
| WO2021180246A1 (en) * | 2020-10-31 | 2021-09-16 | 平安科技（深圳）有限公司 | Drug molecule generation method and apparatus, terminal device, and storage medium |
| CN113707234A (en) * | 2021-08-27 | 2021-11-26 | 中南大学 | Lead compound pharmacy optimization method based on machine translation model |
-
2021
- 2021-12-21 CN CN202111575591.1A patent/CN114334031B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101432310A (en) * | 2006-04-26 | 2009-05-13 | B·布朗·梅尔松根有限公司 | Manufacture and use of modified polysaccharide chitosan bonds and a process to improve the preparation of HES-medicinal substance compounds |
| WO2020006571A1 (en) * | 2018-06-29 | 2020-01-02 | pulseData Inc. | Machine learning systems and methods for predicting risk of renal function decline |
| US20200050717A1 (en) * | 2018-08-09 | 2020-02-13 | Kabushiki Kaisha Toshiba | Optimization apparatus, simulation system and optimization method |
| WO2021180246A1 (en) * | 2020-10-31 | 2021-09-16 | 平安科技（深圳）有限公司 | Drug molecule generation method and apparatus, terminal device, and storage medium |
| CN113707234A (en) * | 2021-08-27 | 2021-11-26 | 中南大学 | Lead compound pharmacy optimization method based on machine translation model |
Non-Patent Citations (1)
| Title |
|---|
| 白茹;滕奇志;杨晓敏;杨胜勇;向明礼;: "基于SVM和GA的药物与人血清白蛋白结合的预测", 计算机工程与应用, no. 12, 21 April 2009 (2009-04-21) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114334031B (en) | 2024-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Linderman et al. | Efficient algorithms for t-distributed stochastic neighborhood embedding | |
| US11334819B2 (en) | Method and system for distributed machine learning | |
| CN109522435B (en) | Image retrieval method and device | |
| WO2021232789A1 (en) | Mirna-disease association prediction method, system, terminal, and storage medium | |
| Heimel et al. | Self-tuning, gpu-accelerated kernel density models for multidimensional selectivity estimation | |
| CN115485695A (en) | Method and system for hierarchical weighted sparse convolution processing | |
| Wang et al. | Retrieval-based controllable molecule generation | |
| CN106777388B (en) | Double-compensation multi-table Hash image retrieval method | |
| TWI804041B (en) | Method and system for convolution with workload-balanced activation sparsity | |
| CN113066525A (en) | Multi-target drug screening method based on ensemble learning and hybrid neural network | |
| CN104573738B (en) | Signal processing method and its device | |
| CN114334031A (en) | Improved processing method, device, equipment and storage medium for pressing parameters | |
| Comin et al. | Fast entropic profiler: An information theoretic approach for the discovery of patterns in genomes | |
| Romberg et al. | Bundle min-Hashing: Speeded-up object retrieval | |
| Misztal | Some remarks on the data imputation using “missForest” method | |
| Lu et al. | From Comparing Clusterings to Combining Clusterings. | |
| WO2023115338A1 (en) | Improvement processing method and apparatus for drug pressing parameters, device, and storage medium | |
| Firouzabadi et al. | An optimized framework for cancer prediction using immunosignature | |
| Vreven et al. | Exploring angular distance in protein-protein docking algorithms | |
| CN112906634B (en) | Video segment sequence prediction model establishment and sequence prediction method and system based on VSS | |
| Chua et al. | Hyperparameter-Tuned CNN for the Classification of Ten Tomato Plant Diseases | |
| US20220207892A1 (en) | Method and device for classifing densities of cells, electronic device using method, and storage medium | |
| CN116384471A (en) | Model pruning method, device, computer equipment, storage medium and program product | |
| TWI858581B (en) | Method and system for convolution with workload-balanced activation sparsity | |
| US20240086408A1 (en) | Data query apparatus, method, and storage medium |
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 |