Training method of deep learning model, method and device for determining data category

Abstract

The present disclosure provides a method for training a deep learning model, which relates to the field of artificial intelligence technology, and in particular to deep learning technology. The specific implementation scheme is: according to a parameter set, a classification model is trained to obtain a target parameter set; according to the parameter set and the target parameter set, a first loss value is obtained; and according to the first loss value, a deep learning model for determining the parameter set is trained. The present disclosure also provides a method for training a target classification model, a method for determining a data category, a device, an electronic device, and a storage medium.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to deep learning technology. More specifically, the present disclosure provides a training method for a deep learning model, a training method for a target classification model, a method for determining a data category, a device, an electronic device, and a storage medium. Background Art

[0002] After completing a training task using a deep learning model, the trained deep learning model can be used to perform another training task based on transfer learning technology. Summary of the invention

[0003] The present disclosure provides a deep learning model training method, a target classification model training method, a data category determination method, an apparatus, a device, and a storage medium.

[0004] According to a first aspect, a deep learning model training method is provided, the method comprising: training a classification model according to a parameter set to obtain a target parameter set; obtaining a first loss value according to the parameter set and the target parameter set; and training a deep learning model for determining the parameter set according to the first loss value.

[0005] According to a second aspect, a training method for a target classification model is provided, the method comprising: inputting second sample data with an artificial label into the target classification model to obtain a second output result; obtaining a second loss value based on the artificial label and the second output result; and training the target classification model based on the third loss value; wherein the initialization parameters of the target classification model are determined by a deep learning model, and the deep learning model is trained according to the method provided in the present disclosure.

[0006] According to a third aspect, a method for determining data categories is provided, the method comprising: inputting target data into a target classification model to obtain the category of the target data; wherein the target classification model is trained according to the method provided in the present disclosure.

[0007] According to a fourth aspect, a training device for a deep learning model is provided, which includes: a first training module, used to train a classification model according to a parameter set to obtain a target parameter set; a first acquisition module, used to obtain a first loss value according to the parameter set and the target parameter set; and a second training module, used to train a deep learning model for determining the parameter set according to the first loss value.

[0008] According to a fifth aspect, a training device for a target classification model is provided, the device comprising: a second acquisition module, used to input second sample data with an artificial label into the target classification model to obtain a second output result; a third acquisition module, used to obtain a third loss value based on the artificial label and the second output result; and a third training module, used to train the target classification model based on the third loss value; wherein the initialization parameters of the target classification model are determined by a deep learning model, and the deep learning model is trained according to the device provided by the present disclosure.

[0009] According to a sixth aspect, a device for determining data categories is provided, the device comprising: a fourth acquisition module, used to input target data into a target classification model to obtain the category of the target data; wherein the target classification model is trained according to the device provided by the present disclosure.

[0010] According to the seventh aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method provided according to the present disclosure.

[0011] According to an eighth aspect, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method provided according to the present disclosure.

[0012] According to a ninth aspect, a computer program product is provided, comprising a computer program, which implements the method provided according to the present disclosure when executed by a processor.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0015] Figure 1 is a flowchart of a method for training a deep learning model according to an embodiment of the present disclosure;

[0016] Figure 2 is a flowchart of a method for training a deep learning model according to another embodiment of the present disclosure;

[0017] Figure 3A is a schematic diagram of original image sample data according to an embodiment of the present disclosure;

[0018] Figure 3B is a schematic diagram of first sample data according to an embodiment of the present disclosure;

[0019] Figure 3C is a schematic diagram of a training method for a deep learning model according to an embodiment of the present disclosure;

[0020] Figure 4 is a flowchart of a method for training a target classification model according to an embodiment of the present disclosure;

[0021] Figure 5A is a schematic diagram of second sample data according to an embodiment of the present disclosure;

[0022] Figure 5B is a schematic diagram of another second sample data according to an embodiment of the present disclosure;

[0023] Figure 5C is a schematic diagram of a training method for a target classification model according to an embodiment of the present disclosure;

[0024] Figure 6 is a rendering of a method for training a target classification model according to an embodiment of the present disclosure;

[0025] Figure 7 is a flow chart of a device for determining a data category according to an embodiment of the present disclosure;

[0026] Figure 8 is a block diagram of a training apparatus for a deep learning model according to an embodiment of the present disclosure;

[0027] Fig. 9 is a block diagram of a training device for a target classification model according to an embodiment of the present disclosure;

[0028] Fig.10 is a block diagram of an apparatus for determining a data category according to an embodiment of the present disclosure; and

[0029] Fig.11 It is a block diagram of an electronic device to which a training method for a deep learning model, a training method for a target classification model and / or a method for determining a data category can be applied according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0031] Based on the transfer learning technology, the backbone structure of the classification model can remain consistent when performing different training tasks. For example, a training task Task_A is: based on a natural image dataset, training classification model Model_1. The backbone structure of Model_1 can be ResNet (Residual Network). Another training task Task_B is: based on a medical image dataset, training classification model Model_1. After completing the training task Task_A, based on the transfer learning technology, the trained classification model Model_1 can be used for another training task Task_B. Compared with the untrained classification model Model_1, the trained classification model Model_1 can complete the training task Task_B in fewer training cycles.

[0032] However, if the training task Task_B' is: based on the medical image dataset, train the classification model Model_0. The backbone structure of the classification model Model_0 is the LSTM (Long and Short Term Memory) network. After completing the training task Task_A, based on the transfer learning technology, the parameters of the trained classification model Model_1 cannot be used as the initialization parameters of the classification model Model_0.

[0033] Figure 1 It is a flowchart of a method for training a deep learning model according to an embodiment of the present disclosure.

[0034] like Figure 1 As shown, the method 100 may include operations S110 to S130.

[0035] In operation S110, a classification model is trained according to the parameter set to obtain a target parameter set.

[0036] For example, the classification model is a model from a classification model set. The classification model set includes multiple classification models. In an example, the classification model set may include multiple classification models such as classification model Model_1 and classification model Model_2. The backbone structure of the classification model Model_1 is as described above. The backbone structure of the classification model Model_2 is DenseNet (Dense convolution network).

[0037] For example, a parameter set includes multiple parameters, each of which corresponds to a node on the backbone structure of the classification model.

[0038] For example, the classification model may be trained for at least one sub-cycle to obtain a target parameter set. In one example, each node in the backbone structure of the trained classification model corresponds to a parameter, and the target parameter set may be obtained by acquiring these parameters.

[0039] In operation S120, a first loss value is obtained according to the parameter set and the target parameter set.

[0040] For example, according to the parameter set and the target parameter set, any loss function may be used to obtain the first loss value. In one example, the loss function may be, for example, a cross entropy loss function.

[0041] In operation S130, a deep learning model for determining a parameter set is trained according to the first loss value.

[0042] For example, according to the first loss value, the parameters of the deep learning model are adjusted so that the adjusted deep learning model can determine different parameter sets according to the calculation graph of the classification model.

[0043] Through the embodiments of the present disclosure, the trained deep learning model can determine different parameter sets according to the backbone structures of different classification models, thereby saving the time cost of pre-training the classification model.

[0044] Through the embodiments of the present disclosure, unlike related transfer learning technologies, the backbone structure of the classification model can be inconsistent, thereby being able to adapt to various application scenarios.

[0045] It can be understood that once the main structure of the model is determined, the calculation graph of the model can be determined.

[0046] In some embodiments, training the classification model according to the parameter set to obtain the target parameter set includes: training the classification model for at least one sub-cycle to the target parameter set. For example, the initialized classification model Model_1 may be trained for multiple sub-cycles until the obtained loss value is less than a preset loss threshold. The parameters of each node in the backbone structure of the classification model Model_1 after multiple sub-cycles of training may be obtained to obtain the target parameter set.

[0047] In some embodiments, the deep learning model determines a set of parameters based on a computational graph of a classification model.

[0048] For example, a computational graph can represent the backbone structure of a classification model.

[0049] For example, the deep learning model can generate a parameter set based on a computational graph. In one example, the deep learning model can be a GHNs (Graph HyperNetworks) model. In one example, the deep learning model can generate a parameter set Para_1 based on a computational graph of a classification model Model_1.

[0050] Figure 2 It is a flowchart of a method for training a deep learning model according to another embodiment of the present disclosure.

[0051] like Figure 2 As shown, method 210 can train a classification model according to a parameter set to obtain a target parameter set, which will be described in detail below in conjunction with operations S211 to S214.

[0052] In operation S211, the parameters of the classification model are adjusted according to the parameter set to obtain the adjusted classification model.

[0053] For example, for the classification model Model_1, the parameters of each node on its backbone structure can be replaced with the parameters in the parameter set Para_1 to adjust the parameters of the classification model Model_1 and obtain the adjusted classification model.

[0054] In operation S212, the first sample data is input into the classification model after parameter adjustment to obtain a first output result.

[0055] For example, the first sample data has a sample label.

[0056] In the embodiment of the present disclosure, the first sample data is obtained based on unlabeled original sample data.

[0057] For example, the original sample data includes original image sample data. In one example, the original image sample data may include original medical image sample data, such as an original fundus image.

[0058] For example, image enhancement is performed on the original image sample data to obtain first sample data and a sample label of the first sample data. For another example, the sample label is used to characterize the image enhancement method.

[0059] In one example, image enhancement may include various image enhancement methods such as image rotation and image flipping. Taking image rotation as an example, the original fundus image may be rotated 90° clockwise to obtain a first sample data. The sample label Lable_1 of the first sample data may be "rotated 90°".

[0060] In operation S213, a second loss value is obtained according to the first output result and the sample label.

[0061] For another example, as described above, the first sample image is input into the parameter-adjusted classification model Model_1 to obtain the category of the fundus image, which is the first output result Output_1. According to the first output result Output_1 and the sample label Lable_1, a second loss value can be obtained by using any loss function. In one example, the loss function can be, for example, the cross entropy loss function described above.

[0062] In operation S214, the parameters of the adjusted classification model are adjusted according to the second loss value to obtain a target parameter set.

[0063] For example, based on the second loss value described above, the parameters of the adjusted classification model Model_1 can be adjusted to complete a sub-cycle of training. Afterwards, the parameters of each node in the trunk structure of the adjusted classification model Model_1 can be obtained to obtain the target parameter set.

[0064] Through the embodiments of the present disclosure, the classification model can be trained based on a self-supervised training method. For training tasks based on medical image data sets, labels can be generated automatically based on the medical image data, which can reduce the labeling cost of sample data.

[0065] It can be understood that the first sample data can be data in various modalities such as text data, image data, audio data, etc., and the present disclosure does not limit this.

[0066] It can be understood that the original sample data may also include data in various modalities such as original text sample data and original audio sample data, and the present disclosure does not limit this.

[0067] In some embodiments, according to the first loss value, training the deep learning model includes: for the current training cycle, according to the first loss value, adjusting the parameters of the deep learning model to obtain an adjusted deep learning model. For example, the adjusted deep learning model can generate a parameter set Para_1' different from the parameter set Para_1 above according to the calculation graph of the classification model Model_1.

[0068] In some embodiments, according to the first loss value, training a deep learning model for determining a parameter set further includes: for the next training cycle, inputting the computation graph of the classification model corresponding to the next training cycle into the adjusted deep learning model, obtaining a parameter set of the classification model corresponding to the next training cycle, to train the adjusted deep learning model. For example, the classification model corresponding to the next training cycle can still be the classification model Model_1, or it can be another classification model Model_2. In this embodiment, the classification model Model_2 is taken as an example. The adjusted deep learning model described above can generate a parameter set Para_2 based on the computation graph of the classification model Model_2.

[0069] Figure 3A is a schematic diagram of original image sample data according to an embodiment of the present disclosure.

[0070] like Figure 3A As shown, the original image sample data 301 may be a fundus image.

[0071] Figure 3B is a schematic diagram of first sample data according to an embodiment of the present disclosure.

[0072] like Figure 3B As shown, the first sample data 302 may be obtained by clockwise rotating the original image sample data 301. The rotation angle may be, for example, 90°. The label of the first sample data 302 may be "rotated 90°".

[0073] Figure 3C It is a schematic diagram of a training method for a deep learning model according to an embodiment of the present disclosure.

[0074] like Figure 3C As shown, the deep learning model 310 outputs a parameter set according to the calculation graph of the classification model. The classification model adjusts the parameters according to the parameter set and converts it into a classification model 320 after parameter adjustment.

[0075] Next, the parameter-adjusted classification model 320 takes the first sample image 302 as input and outputs a first output result 304. The first sample image 302 may have a sample label 303. The sample label 303 may be the "rotated 90°" described above. According to the first output result 304 and the sample label 303, a second loss value 305 may be obtained. According to the second loss value 305, the parameters of the parameter-adjusted classification model 320 may be adjusted to obtain a target parameter set. According to the parameter set and the target parameter set, a first loss value may be obtained. Next, according to the first loss value, the parameters of the deep learning model 310 may be adjusted to train the deep learning model 310.

[0076] Figure 4 It is a flowchart of a method for training a target classification model according to an embodiment of the present disclosure.

[0077] like Figure 4 As shown, the method 400 includes operations S410 to S430.

[0078] In operation S410, second sample data with manual labels is input into a target classification model to obtain a second output result.

[0079] For example, the target classification model may be one of the classification model sets described above, or may be other classification models, and the present disclosure does not impose any limitation on this.

[0080] For example, some medical image sample data may be manually annotated to obtain second sample data. In one example, the second sample image may be a fundus image, and the manual label of the second sample image may be "Negative". The manual label "Negative" may indicate that there is no lesion in the eye.

[0081] For another example, there may be multiple second sample data.

[0082] In operation S420, a third loss value is obtained according to the manual label and the second output result.

[0083] For example, according to the manual label and the second output result, a third loss value can be obtained by using any loss function. In one example, the loss function can be, for example, the cross entropy loss function described above.

[0084] In operation S430, a target classification model is trained according to the third loss value.

[0085] For example, the initialization parameters of the target classification model are determined by the deep learning model. In one example, the initialization parameters of the target classification model are obtained by inputting the calculation graph of the target classification model into the deep learning model.

[0086] For example, the deep learning model is trained according to the method provided by the present disclosure. In one example, the deep learning model is trained according to the method 100 provided by the present disclosure.

[0087] Through the embodiments of the present disclosure, the initialization parameters of the target classification model can be quickly determined based on the trained deep learning model, and the target classification model can be trained efficiently.

[0088] It can be understood that the second sample data can be data in various modalities such as text data, image data, audio data, etc., and the present disclosure does not limit this.

[0089] Figure 5A is a schematic diagram of second sample data according to an embodiment of the present disclosure.

[0090] like Figure 5A As shown, the second sample data 501 may be the original image sample data 301 mentioned above. The original image sample data 301 is manually labeled to obtain the second sample data 501. The manual label of the second sample data 501 may be "Negative", which may indicate that there is no eye lesion.

[0091] Figure 5B is a schematic diagram of another second sample data according to an embodiment of the present disclosure.

[0092] like Figure 5B As shown, the artificial label of the second sample data 502 may be “Positive”, which may indicate that there is a lesion in the eye.

[0093] Figure 5C It is a schematic diagram of a training method for a target classification model according to an embodiment of the present disclosure.

[0094] like Figure 5C As shown, the deep learning model 510 can output a parameter set according to the calculation graph of the target classification model 530 to determine the initialization parameters of the target classification model 530. After the deep learning model 310 described above is trained, it can be used as the deep learning model 510 in this embodiment.

[0095] Next, the target classification model 530 takes the second sample image 501 as input and outputs a second output result 504. The second sample image 502 may have an artificial label 503. The artificial label 503 may be the "Negative" described above, which may indicate that the eye has no lesions. According to the second output result 504 and the artificial label 503, a third loss value 505 may be obtained. According to the third loss value 505, the parameters of the target classification model 530 are adjusted to train the target classification model 530.

[0096] Figure 6 It is a diagram showing the effect of a training method for a target classification model according to an embodiment of the present disclosure.

[0097] For example, in this embodiment, the training task M_1 is: based on the fundus image dataset EyePACS, train the target classification model. The fundus image dataset EyePACS contains multiple fundus images, and the label of each fundus image can represent the grade of diabetic retinopathy. There are 5 levels of diabetic retinopathy, which is equivalent to 5 categories corresponding to multiple fundus images. The target classification model can be classified according to a fundus image. In the training task M_1, the initialization parameters of the target classification model Model_3 are obtained by inputting the calculation graph of the target classification model Model_3 into the deep learning model, and the deep learning model is obtained after the training of, for example, method 100 is completed.

[0098] For another example, in order to illustrate the effect of the present disclosure, training task M_2 and training task M_3 are set as a comparison of training task M_1.

[0099] The difference between training task M_2 and training task M_1 is that the initialization parameters of the target classification model Model_4 are obtained after pre-training the target classification model Model_4 based on the ImageNet dataset.

[0100] The difference between training task M_3 and training task M_1 is that the initialization parameters of the target classification model Model_5 are randomly generated.

[0101] The three training tasks described above are performed separately, and the distribution consistency index Kappa is used for evaluation to obtain Figure 6 For example, by executing the training task M_1 described above, we can obtain Figure 6 The curve C601 is shown. By executing the training task M_2 described above, we can obtain Figure 6 The curve C602 is shown. By executing the training task M_3 described above, we can obtain Figure 6 Curve C603 is shown.

[0102] like Figure 6 As shown in the figure, as the training progresses, the difference between curve C601 and curve C602 gradually decreases, and finally obtains a close Kappa index value. It can be understood that the training results obtained after initialization using the parameter set generated by the deep learning model are close to the training results obtained by pre-training using the ImageNet dataset. However, for target classification models with different backbone structures, the trained deep learning model can quickly obtain different initialization parameters according to different calculation graphs, which is more efficient.

[0103] like Figure 6 As shown, as the training progresses, curve C601 is always above curve C603. It can be understood that the training results obtained after initialization using the parameter set generated by the deep learning model are significantly better than the training results obtained by random initialization.

[0104] Figure 7 The present invention is a method for determining a data category according to an embodiment of the present disclosure.

[0105] like Figure 7 As shown, method 700 may include operation S710.

[0106] In operation S710, target data is input into a target classification model to obtain a category of the target data.

[0107] For example, the target classification model is trained according to the method provided in the present disclosure.

[0108] For example, the target classification model is trained according to the method 400 provided in the present disclosure.

[0109] It can be understood that the target data can be data in various modalities such as text data, image data, audio data, etc., and the present disclosure does not limit this.

[0110] Figure 8 is a block diagram of a training device for a deep learning model according to an embodiment of the present disclosure.

[0111] like Figure 8 As shown, the apparatus 800 may include a first training module 810 , a first obtaining module 820 , and a second training module 830 .

[0112] The first training module 810 is used to train a classification model according to a parameter set to obtain a target parameter set.

[0113] The first obtaining module 820 is used to obtain a first loss value according to the parameter set and the target parameter set.

[0114] The second training module 830 is used to train a deep learning model for determining the parameter set according to the first loss value.

[0115] In some embodiments, the first training module includes: an adjustment unit, used to adjust the parameters of the classification model according to the parameter set to obtain the classification model after parameter adjustment; a first acquisition unit, used to input the first sample data into the classification model after parameter adjustment to obtain a first output result, wherein the first sample data has a sample label; a second acquisition unit, used to obtain a second loss value according to the first output result and the sample label; and a second adjustment unit, used to adjust the parameters of the classification model after parameter adjustment according to the second loss value to obtain a target parameter set.

[0116] In some embodiments, the first sample data is obtained based on unlabeled original sample data.

[0117] In some embodiments, the original sample data includes original image sample data, and the first sample data is obtained by the following units performing relevant operations based on the original image sample data: an enhancement unit, used to perform image enhancement on the original image sample data to obtain the first sample data and a sample label of the first sample data, wherein the sample label is used to characterize the image enhancement method.

[0118] In some embodiments, the original image sample data includes medical image sample data.

[0119] In some embodiments, the second training module includes: a third adjustment unit, used to adjust the parameters of the deep learning model according to the first loss value for the current training cycle to obtain an adjusted deep learning model; a fourth adjustment unit, used to input the calculation graph of the classification model corresponding to the next training cycle into the adjusted deep learning model for the next training cycle to obtain a parameter set of the classification model corresponding to the next training cycle to train the adjusted deep learning model.

[0120] In some embodiments, the deep learning model determines the parameter set based on a computational graph of the classification model.

[0121] Fig. 9 is a block diagram of a training device for a target classification model according to an embodiment of the present disclosure.

[0122] like Fig. 9 As shown, the device 900 may include a second obtaining module 910 , a third obtaining module 920 and a third training module 930 .

[0123] The second obtaining module 910 is used to input the second sample data with artificial labels into the target classification model to obtain a second output result.

[0124] The third obtaining module 920 is used to obtain a third loss value according to the manual label and the second output result.

[0125] The third training module 930 is used to train the target classification model according to the third loss value.

[0126] For example, the initialization parameters of the target classification model are determined by a deep learning model, and the deep learning model is trained according to the device provided by the present disclosure.

[0127] Fig.10 It is a block diagram of an apparatus for determining a data category according to an embodiment of the present disclosure.

[0128] like Fig.10 As shown, the device 1000 may include a fourth obtaining module 1010 .

[0129] The fourth obtaining module 1010 is used to input the target data into the target classification model to obtain the category of the target data.

[0130] For example, the target classification model is trained according to the apparatus of claim 15.

[0131] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0132] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0133] Fig.11 A schematic block diagram of an example electronic device 1100 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0134] like Fig.11As shown, the device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the device 1100 can also be stored. The computing unit 1101, the ROM 1102, and the RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0135] A number of components in the device 1100 are connected to the I / O interface 1105, including: an input unit 1106, such as a keyboard, a mouse, etc.; an output unit 1107, such as various types of displays, speakers, etc.; a storage unit 1108, such as a disk, an optical disk, etc.; and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1109 allows the device 1100 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0136] The computing unit 1101 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1101 performs the various methods and processes described above, such as a training method for a deep learning model, a training method for a target classification model, and / or a method for determining a data category. For example, in some embodiments, a training method for a deep learning model, a training method for a target classification model, and / or a method for determining a data category may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded into RAM 1103 and executed by the computing unit 1101, one or more steps of the deep learning model training method, the target classification model training method, and / or the method for determining the data category described above may be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to execute the deep learning model training method, the target classification model training method, and / or the method for determining the data category in any other appropriate manner (e.g., by means of firmware).

[0137] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0138] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0139] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0140] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0141] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0142] A computer system may include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship to each other.

[0143] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0144] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A training method for a deep learning model, comprising: According to the parameter set, adjust the parameters of the classification model to obtain the classification model after parameter adjustment; Inputting a first sample image into the parameter-adjusted classification model to obtain a first output result, wherein the first sample image has a sample label; Obtaining a second loss value according to the first output result and the sample label; According to the second loss value, adjusting the parameters of the adjusted classification model to obtain a target parameter set; Obtaining a first loss value according to the parameter set and the target parameter set; and Training a deep learning model for determining the parameter set according to the first loss value includes: For the current training cycle, According to the first loss value, adjusting the parameters of the deep learning model to obtain an adjusted deep learning model; For the next training cycle, The calculation graph of the classification model corresponding to the next training cycle is input into the adjusted deep learning model to obtain a parameter set of the classification model corresponding to the next training cycle to train the adjusted deep learning model.

2. The method according to claim 1, wherein: The first sample image is obtained based on unlabeled original image sample data.

3. The method according to claim 2, wherein: The first sample image is obtained by performing the following operations according to the original image sample data: Performing image enhancement on the original image sample data to obtain the first sample image and a sample label of the first sample image, The sample label is used to characterize the image enhancement method.

4. The method according to claim 3, wherein: The original image sample data includes medical image sample data.

5. The method according to claim 1, wherein: The deep learning model determines the parameter set according to the calculation graph of the classification model.

6. A method for training a target classification model, comprising: Inputting the second sample data with the artificial label into the target classification model to obtain a second output result; Obtaining a third loss value according to the manual label and the second output result; as well as Training the target classification model according to the third loss value; Wherein, the initialization parameters of the target classification model are determined by a deep learning model, and the deep learning model is trained according to the method according to any one of claims 1 to 5.

7. A method for determining a data category, comprising: Inputting the target data into a target classification model to obtain a category of the target data; Wherein, the target classification model is trained according to the method according to claim 6.

8. A training device for a deep learning model, comprising: An adjustment unit, used to adjust the parameters of the classification model according to the parameter set to obtain the classification model after parameter adjustment; A first obtaining unit, configured to input a first sample image into the parameter-adjusted classification model to obtain a first output result, wherein the first sample image has a sample label; a second obtaining unit, configured to obtain a second loss value according to the first output result and the sample label; and a second adjustment unit, configured to adjust the parameters of the adjusted classification model according to the second loss value to obtain a target parameter set; a first acquisition module, configured to obtain a first loss value according to the parameter set and the target parameter set; and A second training module is used to train a deep learning model for determining the parameter set according to the first loss value, Wherein, the second training module includes: A third adjustment unit, configured to adjust the parameters of the deep learning model according to the first loss value for a current training cycle to obtain an adjusted deep learning model; The fourth adjustment unit is used to input the calculation graph of the classification model corresponding to the next training cycle into the adjusted deep learning model for the next training cycle, obtain the parameter set of the classification model corresponding to the next training cycle, and train the adjusted deep learning model.

9. The device according to claim 8, wherein: The first sample image is obtained based on unlabeled original image sample data.

10. The device according to claim 9, wherein: The first sample image is obtained by the following units performing relevant operations according to the original image sample data: an enhancement unit, configured to perform image enhancement on the original image sample data to obtain the first sample image and a sample label of the first sample image, The sample label is used to characterize the image enhancement method.

11. The device according to claim 10, wherein: The original image sample data includes medical image sample data.

12. The device according to claim 8, wherein: The deep learning model determines the parameter set according to the calculation graph of the classification model.

13. A training device for a target classification model, comprising: A second acquisition module, used for inputting second sample data with artificial labels into the target classification model to obtain a second output result; A third obtaining module, used for obtaining a third loss value according to the manual label and the second output result; as well as A third training module, used for training the target classification model according to the third loss value; Wherein, the initialization parameters of the target classification model are determined by a deep learning model, and the deep learning model is trained according to the device according to any one of claims 8 to 12.

14. A device for determining a data category, comprising: A fourth acquisition module is used to input the target data into the target classification model to obtain the category of the target data; Wherein, the target classification model is trained according to the device according to claim 13.

15. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

16. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 7.

17. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images