CN116384506A - Model training method and device, storage medium and electronic equipment - Google Patents
Model training method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN116384506A CN116384506A CN202310315439.2A CN202310315439A CN116384506A CN 116384506 A CN116384506 A CN 116384506A CN 202310315439 A CN202310315439 A CN 202310315439A CN 116384506 A CN116384506 A CN 116384506A
- Authority
- CN
- China
- Prior art keywords
- data
- global
- detection model
- service data
- client
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000003860 storage Methods 0.000 title claims abstract description 21
- 238000001514 detection method Methods 0.000 claims abstract description 235
- 230000005856 abnormality Effects 0.000 claims abstract description 42
- 230000002159 abnormal effect Effects 0.000 claims description 98
- 238000005457 optimization Methods 0.000 claims description 20
- 230000004931 aggregating effect Effects 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 15
- 230000002776 aggregation Effects 0.000 claims description 11
- 238000004220 aggregation Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 230000006870 function Effects 0.000 description 18
- 238000010586 diagram Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 9
- 230000006872 improvement Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 229920001296 polysiloxane Polymers 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000010979 ruby Substances 0.000 description 1
- 229910001750 ruby Inorganic materials 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The specification discloses a model training method, device, storage medium and electronic equipment, which acquire local business data and train a preset abnormality detection model. And transmitting model data of an anomaly detection model obtained by executing a model training task based on each service data to a server, so that the server aggregates the received model data transmitted by each client to obtain a global anomaly detection model, and transmitting the global anomaly detection model to each client. And inputting each service data into the global anomaly detection model to obtain the data characteristic of each service data, and determining the characteristic center of the data characteristic of each service data. And sending the feature centers to a server so that the server aggregates the received feature centers to obtain global feature centers, and sending the global feature centers to each client. And training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center.
Description
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a method and apparatus for model training, a storage medium, and an electronic device.
Background
With the rapid development of internet technology, protection of private data and business wind control are increasingly attracting attention. In order to better protect the information security of the user, it is generally required to perform anomaly detection on the service data of the user, and in practical application, the anomaly detection is generally performed on the service data by using a locally deployed anomaly detection model, and wind control is performed in time.
In federal learning, a client can train a locally deployed anomaly detection model by using local data, then upload model data of the locally trained anomaly detection model to a server, and the server can aggregate the received model data of a plurality of anomaly detection models to obtain a global anomaly detection model and then issue the global anomaly detection model to each client.
However, because the sample data used by different clients in training the anomaly detection model are different, it may result in that the global anomaly detection model obtained by aggregating the model data based on a plurality of anomaly detection models cannot accurately detect the anomaly data after being deployed locally. For example, the locally trained model can accurately identify abnormal data in the local service data, but the aggregated global abnormal detection model may not accurately identify the local service data, and even may misidentify the abnormal data as normal data.
Therefore, how to make the aggregated global anomaly detection model accurately detect local anomalies is a problem to be solved.
Disclosure of Invention
The specification provides a method, a device, a storage medium and electronic equipment for model training, so as to solve the problem that an aggregated global abnormality detection model in the prior art cannot accurately detect local abnormality.
The technical scheme adopted in the specification is as follows:
the specification provides a method of model training, the method is applied to a local client, and comprises the following steps:
obtaining local service data, wherein the service data comprises the following steps: normal service data and abnormal service data;
model data of an anomaly detection model obtained by executing a model training task based on the service data is sent to a server, so that the server aggregates the received model data sent by each client to obtain a global anomaly detection model, and the global anomaly detection model is sent to each client;
inputting the business data into the global anomaly detection model issued by the server to obtain the data characteristics of each business data, and determining the characteristic center of the data characteristics of the business data;
The feature centers are sent to the server, so that the server aggregates the received feature centers after receiving the feature centers sent by each client, obtains global feature centers and sends the global feature centers to each client;
and training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center.
Optionally, the service data are input into the global anomaly detection model issued by the server, so as to obtain the data feature of each service data, and the feature center of the data feature of each service data is determined, which specifically includes:
inputting the normal business data in the business data into the global abnormal detection model to obtain the data characteristics of the normal business data;
and determining the characteristic center of the data characteristic of each normal service data.
Optionally, training the global anomaly detection model according to the deviation between the data feature of each service data and the global feature center, specifically including:
and training the global anomaly detection model by taking the deviation between the data features of the normal business data and the global feature center as an optimization target and the deviation between the data features of the abnormal business data and the global feature center as the maximum.
Optionally, the service data are input into the global anomaly detection model issued by the server, so as to obtain the data feature of each service data, and the feature center of the data feature of each service data is determined, which specifically includes:
inputting abnormal business data in the business data into the global abnormal detection model to obtain data characteristics of the abnormal business data;
and determining a characteristic center of the data characteristic of each abnormal service data.
Optionally, training the global anomaly detection model according to the deviation between the data feature of each service data and the global feature center, specifically including:
and training the global anomaly detection model by taking the deviation between the data features of the normal business data and the global feature center as an optimization target and the deviation between the data features of the abnormal business data and the global feature center as a minimum.
Optionally, before training the global anomaly detection model according to the deviation between the data feature of each service data and the global feature center, the method further includes:
Inputting the business data into the global anomaly detection model to obtain an output result; wherein, the probability that each service data belongs to the normal service data or the abnormal service data is given in the output result;
training the global anomaly detection model according to the deviation between the data features of the business data and the global feature center, wherein the training comprises the following steps:
and training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center and by taking the minimized difference between the output result corresponding to each service data and the preset label corresponding to each service data as an optimization target.
The specification also provides a method of model training, the method being applied to a server and comprising:
receiving model data of a trained anomaly detection model sent by each client, wherein the trained anomaly detection model corresponding to each client is obtained by training the anomaly detection model deployed locally at the client through local service data of the client, and the service data comprise: normal service data and abnormal service data;
Aggregating the received model data of each trained anomaly detection model to obtain a global anomaly detection model;
issuing the global anomaly detection model to each client so that each client inputs local business data into the global anomaly detection model to obtain the data characteristics of each business data, and determining the characteristic center of the data characteristics of each business data;
receiving a feature center sent by each client, and aggregating the received feature centers to obtain a global feature center;
and sending the global feature center to each client so that each client trains a local global abnormality detection model according to the deviation between the data features of local business data and the global feature center.
The present specification provides an apparatus for model training, the apparatus comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring local business data, and the business data comprise: normal service data and abnormal service data;
the first sending module is used for sending the model data of the anomaly detection model obtained by executing the model training task based on the service data to a server so that the server aggregates the received model data sent by each client to obtain a global anomaly detection model, and sends the global anomaly detection model to each client;
The first input module is used for inputting the business data into the global anomaly detection model issued by the server, obtaining the data characteristics of each business data and determining the characteristic center of the data characteristics of the business data;
the second sending module is used for sending the feature centers to the server so that the server can aggregate the received feature centers after receiving the feature centers sent by each client to obtain global feature centers and send the global feature centers to each client;
and the training module is used for training the global anomaly detection model according to the deviation between the data characteristics of the business data and the global characteristic center.
Optionally, the first input module is specifically configured to input normal service data in the service data into the global anomaly detection model, so as to obtain data features of the normal service data; and determining the characteristic center of the data characteristic of each normal service data.
Optionally, the training module is specifically configured to train the global anomaly detection model with a minimum deviation between the data feature of each normal service data and the global feature center and a maximum deviation between the data feature of each abnormal service data and the global feature center as an optimization target.
Optionally, the first input module is further configured to input abnormal service data in the service data into the global anomaly detection model, so as to obtain data features of the abnormal service data; and determining a characteristic center of the data characteristic of each abnormal service data.
Optionally, the training module is further configured to train the global anomaly detection model with an optimization objective that maximizes a deviation between the data feature of each normal service data and the global feature center and minimizes a deviation between the data feature of each abnormal service data and the global feature center.
Optionally, the apparatus further comprises:
the second input module is used for inputting the business data into the global abnormality detection model before training the global abnormality detection model according to the deviation between the data characteristics of the business data and the global characteristic center to obtain an output result; wherein, the probability that each service data belongs to the normal service data or the abnormal service data is given in the output result;
the training module is used for training the global anomaly detection model according to the deviation between the data characteristics of the business data and the global characteristic center and by taking the difference between the output result corresponding to the minimized business data and the preset label corresponding to the business data as an optimization target.
The present specification also provides an apparatus for model training, the apparatus comprising:
the first receiving module is configured to receive model data of a trained anomaly detection model sent by each client, where, for each client, the trained anomaly detection model corresponding to the client is obtained by training an anomaly detection model deployed locally at the client through local service data of the client, where each service data includes: normal service data and abnormal service data;
the aggregation module is used for aggregating the received model data of each trained abnormality detection model to obtain a global abnormality detection model;
the issuing module is used for issuing the global abnormality detection model to each client so that each client inputs local business data into the global abnormality detection model to obtain the data characteristics of each business data and determine the characteristic center of the data characteristics of each business data;
the second receiving module is used for receiving the feature centers sent by each client and aggregating the received feature centers to obtain a global feature center;
and the sending module is used for sending the global feature center to each client so that each client trains the local global abnormality detection model according to the deviation between the data features of the local business data and the global feature center.
The present specification provides a computer readable storage medium storing a computer program which when executed by a processor implements the method of model training described above.
The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing a method of model training as described above when executing the program.
The above-mentioned at least one technical scheme that this specification adopted can reach following beneficial effect:
in the model training method provided by the specification, a client acquires local business data to train a preset abnormality detection model, model data of the trained abnormality detection model is sent to a server, so that the server aggregates the model data after receiving the model data sent by each client, a global abnormality detection model is obtained, and the global abnormality detection model is sent to each client. And inputting each service data into the global anomaly detection model to obtain the data characteristic of each service data, and determining the characteristic center of the data characteristic of each service data. And sending the feature centers to a server so that the server aggregates the received feature centers to obtain global feature centers, and sending the global feature centers to each client. And training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center.
According to the method, the client can obtain the data characteristics of each service data and determine the characteristic centers of the data characteristics of each service data based on the global anomaly detection model, and after the server aggregates the characteristic centers sent by each client to obtain the global characteristic centers, the local client can use the local service data to train the global anomaly detection model based on the global characteristic centers and the global anomaly detection model. That is, the client does not directly use the global anomaly detection model aggregated by the server to perform anomaly detection, and after receiving the global anomaly detection model and the global feature center, the client also needs to train the global anomaly detection model according to the deviation between the data features of each service data and the global feature center, so that the trained global anomaly detection model is beneficial to more accurately identify the anomaly service data.
Drawings
The accompanying drawings, which are included to provide a further understanding of the specification, illustrate and explain the exemplary embodiments of the present specification and their description, are not intended to limit the specification unduly. Attached at
In the figure:
FIG. 1 is a flow chart of a method of model training provided in the present specification;
FIG. 2 is a flow chart of another method of model training provided in the present specification;
FIG. 3 is a schematic diagram of an apparatus for a method of model training provided in the present specification;
FIG. 4 is a schematic diagram of an apparatus of another method of model training provided in the present specification;
fig. 5 is a schematic diagram of an electronic device corresponding to fig. 1 and 2 provided in the present specification.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the present specification more apparent, the technical solutions of the present specification will be clearly and completely described below with reference to specific embodiments of the present specification and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present specification. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.
The following describes in detail the technical solutions provided by the embodiments of the present specification with reference to the accompanying drawings.
Fig. 1 is a flow chart of a method for model training provided in the present specification, including the following steps:
S100: obtaining local service data, wherein the service data comprises the following steps: normal traffic data and abnormal traffic data.
S102: and transmitting model data of an anomaly detection model obtained by executing a model training task based on the service data to a server, so that the server aggregates the received model data transmitted by each client to obtain a global anomaly detection model, and transmitting the global anomaly detection model to each client.
In federal learning, a client sends model data (such as gradient data) of a locally trained anomaly detection model to a server, and the server can aggregate the received model data of a plurality of anomaly detection models to obtain a global anomaly detection model, and then send the global anomaly detection model to each client. However, because different clients use different sample data in training the anomaly detection model, the global anomaly detection model may not accurately identify local anomaly data after being deployed locally. Based on the above, the specification provides a model training method to solve the problem that the aggregated global anomaly detection model cannot accurately identify anomaly data.
The execution body of the present specification is a terminal device, for example, a mobile device such as a mobile phone or a tablet computer, or may be a client installed in the terminal device, and for convenience of description, a mode of model training will be described below taking the client as an execution body.
The client can acquire local business data and train a preset abnormality detection model based on the business data. Wherein, each business data comprises normal business data and abnormal business data.
Specifically, the client may acquire local service data and a label corresponding to each preset service data, and input each service data into a preset anomaly detection model to obtain an output result, where the output result is a probability that the service data belongs to normal service data or anomaly data.
Further, the client can determine the loss between the output result corresponding to each service data and the preset label corresponding to each service data. The client may perform model training using various loss functions, for example, may perform training tasks using cross entropy loss, which is specifically as follows:
L 1 ＝―[y log(y 1 )]+(1-y 1 )log(1-y 1 )
wherein y is the output result corresponding to each service data, y 1 And the label is a label corresponding to each preset service data. The client can train the preset abnormality detection model by taking the difference between the output result corresponding to the minimized business data and the label corresponding to the preset business data as an optimization target.
It should be noted that the above cross entropy function is merely an example to illustrate the training process of the local anomaly detection model, and of course, other types of loss functions may be adopted in practical applications, and the present specification is not limited to the specific loss function used.
In addition, the client may further input each service data into a preset anomaly detection model to obtain data features of each service data, determine feature centers of the data features of each service data, and preset training the preset anomaly detection model according to deviation between the data features of each service data and the feature centers of the data features of each service data.
For example, if the number of normal service data in the local service data of the client is greater than the number of abnormal service data, the client may train the preset abnormal detection model with the objective of minimizing the deviation between the data features of each normal service data and the feature centers of the data features of each service data and maximizing the deviation between the data features of each abnormal service data and the feature centers of the data features of each service data.
In addition, the client may further input each service data into a preset anomaly detection model to obtain data features of each service data, determine feature centers of the data features of each service data, and train the preset anomaly detection model according to deviation between the data features of each service data and the feature centers of the data features of each service data.
For example, if the number of normal service data in the local service data of the client is greater than the number of abnormal service data, the client may optimize the preset abnormality detection model with a view to minimizing the deviation between the data feature of each normal service data and the feature center of the data feature of each service data and maximizing the deviation between the data feature of each abnormal service data and the feature center of the data feature of each service data.
If the number of normal service data in the local service data of the client is smaller than the number of abnormal service data, the client can optimize a preset abnormal detection model by taking the deviation between the maximum data characteristics of each normal service data and the characteristic centers of the data characteristics of each service data and the minimum deviation between the data characteristics of each abnormal service data and the characteristic centers of the data characteristics of each service data as an optimization target.
It is worth to say that the abnormality detection model is trained by the client side by using local service data, and can accurately identify normal service data and abnormal service data in the local service data.
The client can send the model data of the trained abnormality detection model to the server, so that the server aggregates the received model data after receiving the model data sent by each client to obtain a global abnormality detection model, and sends the global abnormality detection model to each client.
S104: and inputting the business data into the global anomaly detection model issued by the server to obtain the data characteristics of each business data, and determining the characteristic center of the data characteristics of the business data.
After receiving the global anomaly detection model issued by the server, the client can obtain the data characteristic of each service data based on each local service data and the global anomaly detection model, and determine the characteristic center of the data characteristic of each service data.
In this specification, feature centers of data features of respective business data can be classified into three cases.
Since the number of normal service data in each service data is larger than the number of abnormal service data under normal conditions, the client can determine the feature center of the data features of each service data according to the data features of all service data. Specifically, the client may input all the service data into the global anomaly detection model to obtain all the data features of the service data, thereby determining feature centers of the data features of the service data. For example, the client may use an average value of the data characteristics of all the service data as a feature center of the data characteristics of each service data, based on the data characteristics of all the service data.
The number of the normal service data is larger, so that the feature center of the data feature of each service data determined by the client is closer to the data feature of each normal service data, and is relatively far away from the data feature of each abnormal service data.
Therefore, after the feature center of the data features of each service data is determined in this way, the subsequent client sends the feature center to the server, so that when the server aggregates and determines the global feature center, the obtained global feature center is also the feature center which is closer to the data features of each normal service data, but is relatively far away from the feature center of the data features of each abnormal service data.
Of course, the feature center of the data feature of each service data may be the feature center of the data feature of each normal service data. The client can input the normal service data in the service data into the global anomaly detection model to obtain the data characteristics of the normal service data, and further, the client can determine the characteristic center of the data characteristics of the normal service data. For example, according to the data characteristics of each normal service data, the client may use an average value of the data characteristics of each normal service data as a feature center of the data characteristics of each normal service data.
In addition, the feature center of the data feature of each service data may be the feature center of the data feature of each abnormal service data. The client can input abnormal business data in each business data into the global abnormal detection model to obtain the data characteristics of each abnormal business data, and further, the client can determine the characteristic center of the data characteristics of each abnormal business data. For example, according to the data characteristics of the different service data, the client may use the average value of the data characteristics of the different service data as the characteristic center of the data characteristics of the different service data.
It should be noted that, after determining the feature center of the data feature of each service data in this way, the subsequent client sends the feature center to the server, so that when the server aggregates and determines the global feature center, the obtained global feature center is closer to the feature center of the data feature of each abnormal service data, and is relatively far away from the feature center of the data feature of each normal service data.
S106: and sending the feature center to the server, so that the server aggregates the received feature centers after receiving the feature centers sent by each client to obtain a global feature center, and sending the global feature center to each client.
S108: and training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center.
The client can send the feature center for determining the data features of the service data to the server, so that the server aggregates the received feature centers after receiving the feature centers sent by each client to obtain global feature centers, and sends the global feature centers to each client.
The global feature center refers to the center of each service data feature in the feature space where the data feature of each service data is located.
After receiving the global feature center, the client can train the global abnormality detection model continuously based on the global abnormality detection model and the global feature center according to the deviation between the data features of each service data and the global feature center.
Specifically, if the feature center of the data features of each service data sent to the server by the client is determined based on the feature centers of the data features of all the service data, the global feature center aggregated by the server is determined according to the feature centers of the data features of all the service data of each client.
Therefore, after receiving the global feature center, the client can determine the deviation between the data feature of each normal service data and the global feature center, and the deviation between the data feature of each abnormal service data and the global feature center, and determine the loss value according to the determined deviation, as shown in the following formula:
L 2 ＝max(D(F pos ,C)―D(F neg ,C)+margin，0)
wherein C is a global feature center, D is a distance function, F pos Is the data characteristic of each normal service data, F neg Is the data characteristic of each abnormal business data, and margin is used for controlling the data characteristic of each normal business data and the distance between the data characteristic of each abnormal business data and the global characteristic center.
As can be seen from the above formula, in the case that the data features of each normal service data are close to the global feature center and the data features of each abnormal service data are far from the global feature center, the distance between the data features of each normal service data and the global feature center is smaller than the distance between the data features of each abnormal service data and the global feature center, that is, D (F) pos ,C)―D(F neg The value of C) is less than 0, in which case L 2 =0. In contrast, in the case that the distance between the data feature of each normal service data and the global feature center is greater than the distance between the data feature of each different service data and the global feature center, L 2 The value of (2) is a positive number.
Therefore, if the global feature center is determined by the server aggregating the feature centers of the data features of all the service data sent by each client, the client may train the global anomaly detection model with the minimized deviation between the data features of each normal service data and the global feature center, and the maximized deviation between the data features of each abnormal service data and the global feature center as an optimization objective (i.e., with the minimized loss value as an optimization objective).
Under normal conditions, the number of normal business data in each business data is larger than the number of abnormal business data, and by the training mode, the data characteristics of each normal business data determined by the trained global abnormal detection model are closer to the global characteristic center, and meanwhile, the data characteristics of each determined abnormal business data are farther away from the global characteristic center.
Of course, the global feature center may also be determined by the server aggregating the feature centers of the data features of the normal service data sent by each client, and after receiving the global feature center, the client may determine the deviation between the data features of the normal service data and the global feature center, and the deviation between the data features of the abnormal service data and the global feature center.
The client can train the global anomaly detection model by taking the minimized deviation between the data features of each normal business data and the global feature center and the maximized deviation between the data features of each abnormal business data and the global feature center as optimization targets, and can refer to the formula L 2 . Through the training mode, the client can enable the data characteristics of each piece of normal service data to be closer to the global characteristic center, and enable the data characteristics of each piece of abnormal service data to be farther away from the global characteristic center.
Similarly, if the global feature center is determined by aggregating the feature centers of the data features of the abnormal service data sent by each client by the server, at this time, the client may determine a deviation between the data features of the normal service data and the feature centers of the data features of the service data, and determine a loss value according to the determined deviation, where the deviation between the data features of the abnormal service data and the feature centers of the data features of the service data is as shown in the following formula:
L 3 ＝max(D(F neg ,C)-D(F pos ,C)+margin，0)
from the above formula, it can be seen thatIn the case that the data characteristics of the abnormal business data are close to the global characteristic center and the data characteristics of the normal business data are far from the global characteristic center, L 3 =0. Conversely, in the case that the data features of the abnormal traffic data are far from the global feature center and the data features of the normal traffic data are close to the global feature center, L 3 The value of (2) is a positive number.
Therefore, the client can train the global anomaly detection model by taking the maximized deviation between the data features of the normal business data and the global feature center and the minimized deviation between the data features of the abnormal business data and the global feature center as optimization targets, and the client can enable the data features of the abnormal business data to be closer to the global feature center and enable the data features of the normal business data to be further away from the global feature center through the training method.
Of course, in addition to the foregoing way of training the global anomaly detection model, in the training process, the client may also add the loss function employed when pre-training the anomaly detection model to training.
That is, the client may input each service data into the global anomaly detection model to obtain an output result. Wherein, the probability that each business data belongs to normal business data or abnormal data is outputted in the result. The client can train the global anomaly detection model by taking the difference between the output result corresponding to the minimized business data and the preset label corresponding to the business data as an optimization target and according to the deviation between the data characteristic of the business data and the global characteristic center.
Along the above example of training the local anomaly detection model by using the cross entropy function, when training the global anomaly detection model, if the global feature center is determined by the server aggregating the feature centers of the data features of the normal service data sent by each client, the client can determine that the loss function in the training process of the global anomaly detection model is: l (L) 4 ＝L 1 +L 2 ＝―[y log(y 1 )]+(1-y 1 )log(1-y 1 )+max(D(F pos ,C)―D(F neg C) +margin, 0), and to minimize L 4 To optimize the objective, a global anomaly detection model is trained.
If the global feature center is determined by the server aggregating the feature centers of the data features of the abnormal service data sent by each client, the client can determine that the loss function in the training process of the global abnormal detection model is: l (L) 5 ＝L 1 +L 3 ＝―[y log(y 1 )]+(1-y 1 )log(1-y 1 )+max(D(F neg ,C)―D(F pos C) +margin, 0), and to minimize L 5 To optimize the objective, a global anomaly detection model is trained.
It is worth to say that, after the client trains the global anomaly detection model, a trained global anomaly detection model is obtained, model data of the trained global anomaly detection model can be uploaded to the server again, so that the server aggregates the trained global anomaly detection model sent by each client again, and the aggregated global anomaly detection model is sent to each client.
At this time, the client receives the global anomaly detection model after the second aggregation of the server, and may further input each service data into the global anomaly detection model after the second aggregation in the above-mentioned manner of training the global anomaly detection model, so as to obtain the data features of each service data and determine the feature center of the data features of each service data. The client can send the feature center of the data feature of each service data to the server, so that the server aggregates the feature centers of the data feature of each received service data again, and obtains the global feature center again and sends the global feature center to each client.
The client can train the global anomaly detection model locally by utilizing the service data based on the global feature center and the global anomaly detection model after the second aggregation, and the client can upload the training data to the server for aggregation to obtain the global anomaly detection model after the third aggregation. Through the continuous iteration mode, the client can continuously train the global anomaly detection model issued by the server repeatedly until the training of the global anomaly detection model is finished.
According to the method, the client can obtain the data characteristics of each service data and determine the characteristic centers of the data characteristics of each service data based on the global anomaly detection model, and after the server aggregates the characteristic centers sent by each client to obtain the global characteristic centers, the local client can use the local service data to train the global anomaly detection model based on the global characteristic centers and the global anomaly detection model.
In this way, the client can accurately determine the feature center for distinguishing the normal service data from the abnormal service data, and train the global abnormal detection model by controlling the distances between the data features of the normal service data and the abnormal service data and the global features, so that the global abnormal detection model can accurately identify the abnormal service data. In addition, the specification only divides the business data into normal and abnormal business data, and as long as the global abnormality detection model detects that the distance between the abnormal business data and the global feature center is far, the abnormal business data can be rapidly identified, and the abnormality detection efficiency is further improved.
The present specification also provides a method for model training, as shown in fig. 2.
FIG. 2 is a flow chart of another model training method in the present specification, specifically comprising the following steps:
s200: receiving model data of a trained anomaly detection model sent by each client, wherein the trained anomaly detection model corresponding to each client is obtained by training the anomaly detection model deployed locally at the client through local service data of the client, and the service data comprise: normal traffic data and abnormal traffic data.
After each client transmits the model data of the trained abnormality detection model to the server, the server may receive the model data of the abnormality detection model transmitted by each client in response to the transmission operation of each client. For each client, the trained anomaly detection model corresponding to the client is obtained by training the anomaly detection model deployed locally on the client through each piece of local service data of the client, wherein each piece of service data comprises: normal traffic data and abnormal traffic data.
S202: and aggregating the received model data of each trained abnormality detection model to obtain a global abnormality detection model.
After the server receives the model data of the anomaly detection model sent by each client, the server can aggregate the model data of each trained anomaly detection model to obtain a global anomaly detection model.
Specifically, the server may aggregate model data of each trained anomaly detection model by using a model aggregation method used in federal learning, and the present specification does not limit the model data aggregation method.
S204: and issuing the global anomaly detection model to each client so that each client inputs local business data into the global anomaly detection model to obtain the data characteristics of each business data, and determining the characteristic center of the data characteristics of each business data.
S206: and receiving the feature centers sent by each client, and aggregating the received feature centers to obtain a global feature center.
After each client transmits the determined feature centers of the data features of the service data to the server, the server can receive the feature centers transmitted by each client in response to the transmitting operation of each client, and aggregate the received feature centers to obtain a global feature center.
Specifically, the server may use the average value of the received feature centers as the global feature center, and of course, the server may also determine the global feature center by adopting other aggregation manners, which is not limited in the present specification.
S208: and sending the global feature center to each client so that each client trains a local global abnormality detection model according to the deviation between the data features of local business data and the global feature center.
According to the method, the server aggregates the received model data of each trained abnormality detection model sent by each client and the feature centers of the data features of each service data to determine a global abnormality detection model and a global feature center, and then sends the global abnormality detection model and the global feature center to each client.
After the client trains the global anomaly detection model based on the global anomaly detection model and the global feature center issued by the server by utilizing the local business data, the server can continuously gather and issue the trained global anomaly detection model uploaded by each client and the feature center of each business data feature re-determined by each client again, so that the client can repeatedly and iteratively train the global anomaly detection model re-gathered by the server until the training of the gathered global anomaly detection model is finished.
The above method for model training provided for one or more embodiments of the present disclosure further provides a corresponding model training apparatus based on the same concept, as shown in fig. 3.
Fig. 3 is a schematic diagram of a model training apparatus provided in the present specification, the apparatus including:
the obtaining module 300 is configured to obtain local service data, where each service data includes: normal service data and abnormal service data;
the first sending module 302 is configured to send model data of an anomaly detection model obtained by performing a model training task based on the service data to a server, so that the server aggregates the received model data sent by each client to obtain a global anomaly detection model, and sends the global anomaly detection model to each client;
the first input module 304 is configured to input the service data into the global anomaly detection model issued by the server, obtain a data feature of each service data, and determine a feature center of the data feature of each service data;
the second sending module 306 is configured to send the feature center to the server, so that after the server receives the feature center sent by each client, the server aggregates the received feature centers to obtain a global feature center, and sends the global feature center to each client;
The training module 308 is configured to train the global anomaly detection model according to a deviation between the data feature of each service data and the global feature center.
Optionally, the first input module 304 is specifically configured to input normal service data in the service data into the global anomaly detection model, so as to obtain data features of the normal service data; and determining the characteristic center of the data characteristic of each normal service data.
Optionally, the training module 308 is specifically configured to train the global anomaly detection model with a view to minimizing a deviation between the data feature of each normal service data and the global feature center and maximizing a deviation between the data feature of each abnormal service data and the global feature center as an optimization objective.
Optionally, the first input module 304 is further configured to input abnormal service data in the service data into the global abnormal detection model, so as to obtain data features of the abnormal service data; and determining a characteristic center of the data characteristic of each abnormal service data.
Optionally, the training module 308 is further configured to train the global anomaly detection model with a goal of optimizing to maximize a deviation between the data feature of each normal service data and the global feature center and minimize a deviation between the data feature of each abnormal service data and the global feature center.
Optionally, the apparatus further comprises:
a second input module 310, configured to input each service data into the global anomaly detection model before training the global anomaly detection model according to a deviation between a data feature of each service data and the global feature center, so as to obtain an output result; wherein, the probability that each service data belongs to the normal service data or the abnormal service data is given in the output result;
the training module 308 is configured to train the global anomaly detection model according to a deviation between the data feature of each service data and the global feature center, and by using a difference between an output result corresponding to each service data and a preset label corresponding to each service data as an optimization target.
The present specification also provides another model training apparatus, as shown in fig. 4.
FIG. 4 is a schematic diagram of another model training apparatus provided in the present specification, the apparatus comprising:
the first receiving module 400 is configured to receive model data of a trained anomaly detection model sent by each client, where, for each client, the trained anomaly detection model corresponding to the client is obtained by training, by using service data local to the client, the service data including: normal service data and abnormal service data;
The aggregation module 402 is configured to aggregate the received model data of each trained anomaly detection model to obtain a global anomaly detection model;
the issuing module 404 is configured to issue the global anomaly detection model to each client, so that each client inputs local service data into the global anomaly detection model, obtains a data feature of each service data, and determines a feature center of the data feature of each service data;
a second receiving module 406, configured to receive the feature centers sent by each client, and aggregate the received feature centers to obtain a global feature center;
and the sending module 408 is configured to send the global feature center to each client, so that each client trains the local global anomaly detection model according to a deviation between the data feature of each local service data and the global feature center.
The present specification also provides a computer readable storage medium having stored thereon a computer program operable to perform a method of model training as provided in fig. 1 above.
The present specification also provides a schematic structural diagram of an electronic device corresponding to fig. 1 and 2 shown in fig. 5. At the hardware level, as in fig. 5, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile storage, although it may include hardware required for other services. The processor reads the corresponding computer program from the nonvolatile memory into the memory and then runs to implement the model training method of fig. 1 and 2. Of course, other implementations, such as logic devices or combinations of hardware and software, are not excluded from the present description, that is, the execution subject of the following processing flows is not limited to each logic unit, but may be hardware or logic devices.
In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable Gate Array, FPGA)) is an integrated circuit whose logic function is determined by the programming of the device by a user. A designer programs to "integrate" a digital system onto a PLD without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented by using "logic compiler" software, which is similar to the software compiler used in program development and writing, and the original code before the compiling is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but not just one of the hdds, but a plurality of kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), lava, lola, myHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.
The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers, and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller may thus be regarded as a kind of hardware component, and means for performing various functions included therein may also be regarded as structures within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.
The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functions of each element may be implemented in one or more software and/or hardware elements when implemented in the present specification.
It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The present description is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.
It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The foregoing is merely an example of the present specification and is not intended to limit the present specification. Various modifications and alterations to this specification will become apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, or the like, which are within the spirit and principles of the present description, are intended to be included within the scope of the claims of the present description.
Claims (16)
1. A method of model training, the method applied to a local client, comprising:
obtaining local service data, wherein the service data comprises the following steps: normal service data and abnormal service data;
model data of an anomaly detection model obtained by executing a model training task based on the service data is sent to a server, so that the server aggregates the received model data sent by each client to obtain a global anomaly detection model, and the global anomaly detection model is sent to each client;
Inputting the business data into the global anomaly detection model issued by the server to obtain the data characteristics of each business data, and determining the characteristic center of the data characteristics of the business data;
the feature centers are sent to the server, so that the server aggregates the received feature centers after receiving the feature centers sent by each client, obtains global feature centers and sends the global feature centers to each client;
and training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center.
2. The method of claim 1, wherein the inputting the service data into the global anomaly detection model issued by the server obtains a data feature of each service data, and determines a feature center of the data feature of each service data, specifically including:
inputting the normal business data in the business data into the global abnormal detection model to obtain the data characteristics of the normal business data;
and determining the characteristic center of the data characteristic of each normal service data.
3. The method of claim 2, training the global anomaly detection model according to a deviation between the data feature of each service data and the global feature center, specifically comprising:
and training the global anomaly detection model by taking the deviation between the data features of the normal business data and the global feature center as an optimization target and the deviation between the data features of the abnormal business data and the global feature center as the maximum.
4. The method of claim 1, wherein the inputting the service data into the global anomaly detection model issued by the server obtains a data feature of each service data, and determines a feature center of the data feature of each service data, specifically including:
inputting abnormal business data in the business data into the global abnormal detection model to obtain data characteristics of the abnormal business data;
and determining a characteristic center of the data characteristic of each abnormal service data.
5. The method of claim 4, training the global anomaly detection model according to a deviation between the data feature of each service data and the global feature center, specifically comprising:
And training the global anomaly detection model by taking the deviation between the data features of the normal business data and the global feature center as an optimization target and the deviation between the data features of the abnormal business data and the global feature center as a minimum.
6. The method of claim 3 or 5, further comprising, prior to training the global anomaly detection model based on deviations between data features of the respective business data and the global feature center:
inputting the business data into the global anomaly detection model to obtain an output result; wherein, the probability that each service data belongs to the normal service data or the abnormal service data is given in the output result;
training the global anomaly detection model according to the deviation between the data features of the business data and the global feature center, wherein the training comprises the following steps:
and training the global anomaly detection model according to the deviation between the data characteristics of each service data and the global characteristic center and by taking the minimized difference between the output result corresponding to each service data and the preset label corresponding to each service data as an optimization target.
7. A method of model training, the method applied to a server, comprising:
receiving model data of a trained anomaly detection model sent by each client, wherein the trained anomaly detection model corresponding to each client is obtained by training the anomaly detection model deployed locally at the client through local service data of the client, and the service data comprise: normal service data and abnormal service data;
aggregating the received model data of each trained anomaly detection model to obtain a global anomaly detection model;
issuing the global anomaly detection model to each client so that each client inputs local business data into the global anomaly detection model to obtain the data characteristics of each business data, and determining the characteristic center of the data characteristics of each business data;
receiving a feature center sent by each client, and aggregating the received feature centers to obtain a global feature center;
and sending the global feature center to each client so that each client trains a local global abnormality detection model according to the deviation between the data features of local business data and the global feature center.
8. An apparatus for model training, the apparatus comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring local business data, and the business data comprise: normal service data and abnormal service data;
the first sending module is used for sending the model data of the anomaly detection model obtained by executing the model training task based on the service data to a server so that the server aggregates the received model data sent by each client to obtain a global anomaly detection model, and sends the global anomaly detection model to each client;
the first input module is used for inputting the business data into the global anomaly detection model issued by the server, obtaining the data characteristics of each business data and determining the characteristic center of the data characteristics of the business data;
the second sending module is used for sending the feature centers to the server so that the server can aggregate the received feature centers after receiving the feature centers sent by each client to obtain global feature centers and send the global feature centers to each client;
And the training module is used for training the global anomaly detection model according to the deviation between the data characteristics of the business data and the global characteristic center.
9. The apparatus of claim 8, wherein the first input module is specifically configured to input normal service data in the service data into the global anomaly detection model to obtain data features of the normal service data; and determining the characteristic center of the data characteristic of each normal service data.
10. The apparatus of claim 9, the training module is specifically configured to train the global anomaly detection model with a goal of optimizing to minimize a deviation between the data features of the normal traffic data and the global feature center and to maximize a deviation between the data features of the abnormal traffic data and the global feature center.
11. The apparatus of claim 8, wherein the first input module is further configured to input abnormal service data in the service data into the global anomaly detection model to obtain data features of the abnormal service data; and determining a characteristic center of the data characteristic of each abnormal service data.
12. The apparatus of claim 9, the training module further to train the global anomaly detection model with optimization objectives that maximize a deviation between the data features of the respective normal traffic data and the global feature center and minimize a deviation between the data features of the respective anomaly traffic data and the global feature center.
13. The apparatus of claim 10 or 12, the apparatus further comprising:
the second input module is used for inputting the business data into the global abnormality detection model before training the global abnormality detection model according to the deviation between the data characteristics of the business data and the global characteristic center to obtain an output result; wherein, the probability that each service data belongs to the normal service data or the abnormal service data is given in the output result;
the training module is used for training the global anomaly detection model according to the deviation between the data characteristics of the business data and the global characteristic center and by taking the difference between the output result corresponding to the minimized business data and the preset label corresponding to the business data as an optimization target.
14. An apparatus for model training, the apparatus comprising:
the first receiving module is configured to receive model data of a trained anomaly detection model sent by each client, where, for each client, the trained anomaly detection model corresponding to the client is obtained by training an anomaly detection model deployed locally at the client through local service data of the client, where each service data includes: normal service data and abnormal service data;
the aggregation module is used for aggregating the received model data of each trained abnormality detection model to obtain a global abnormality detection model;
the issuing module is used for issuing the global abnormality detection model to each client so that each client inputs local business data into the global abnormality detection model to obtain the data characteristics of each business data and determine the characteristic center of the data characteristics of each business data;
the second receiving module is used for receiving the feature centers sent by each client and aggregating the received feature centers to obtain a global feature center;
and the sending module is used for sending the global feature center to each client so that each client trains the local global abnormality detection model according to the deviation between the data features of the local business data and the global feature center.
15. A computer readable storage medium storing a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-7.
16. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any of the preceding claims 1-7 when the program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310315439.2A CN116384506A (en) | 2023-03-28 | 2023-03-28 | Model training method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310315439.2A CN116384506A (en) | 2023-03-28 | 2023-03-28 | Model training method and device, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116384506A true CN116384506A (en) | 2023-07-04 |
Family
ID=86964051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310315439.2A Pending CN116384506A (en) | 2023-03-28 | 2023-03-28 | Model training method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116384506A (en) |
-
2023
- 2023-03-28 CN CN202310315439.2A patent/CN116384506A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115712866B (en) | Data processing method, device and equipment | |
| CN116304720B (en) | Cost model training method and device, storage medium and electronic equipment | |
| CN115618748B (en) | Model optimization method, device, equipment and storage medium | |
| CN116011815A (en) | Model training method and device, electronic equipment and storage medium | |
| CN116757278B (en) | Training method and device of prediction model, storage medium and electronic equipment | |
| CN116896512B (en) | Cloud edge cooperative system evaluation method and device, storage medium and electronic equipment | |
| CN116578877B (en) | Method and device for model training and risk identification of secondary optimization marking | |
| CN116822606A (en) | Training method, device, equipment and storage medium of anomaly detection model | |
| CN116821647A (en) | Optimization method, device and equipment for data annotation based on sample deviation evaluation | |
| CN116861976A (en) | Training method, device, equipment and storage medium of anomaly detection model | |
| CN116384506A (en) | Model training method and device, storage medium and electronic equipment | |
| CN111242195B (en) | Model, insurance wind control model training method and device and electronic equipment | |
| CN116340852B (en) | Model training and business wind control method and device | |
| CN109325127A (en) | A kind of Risk Identification Method and device | |
| CN117009729B (en) | Data processing method and device based on softmax | |
| CN117348999B (en) | Service execution system and service execution method | |
| CN116109008B (en) | Method and device for executing service, storage medium and electronic equipment | |
| CN111461352B (en) | Model training method, service node identification device and electronic equipment | |
| CN116401541A (en) | Model training method and device, storage medium and electronic equipment | |
| CN116957330A (en) | Model training method, service wind control method, device, equipment and storage medium | |
| CN117743824A (en) | Model training and service execution method and device, storage medium and equipment | |
| CN117313739A (en) | Training method, device, equipment and storage medium of language model | |
| CN116227604A (en) | Bayesian network training method and device based on distribution | |
| CN117494052A (en) | Prediction method and device based on automatic generation of space-time static information | |
| CN117592998A (en) | Wind control method and device, storage medium and electronic equipment |
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 |