CN109493216A - Model training method, device, system and storage medium - Google Patents
Model training method, device, system and storage medium Download PDFInfo
- Publication number
- CN109493216A CN109493216A CN201811160666.8A CN201811160666A CN109493216A CN 109493216 A CN109493216 A CN 109493216A CN 201811160666 A CN201811160666 A CN 201811160666A CN 109493216 A CN109493216 A CN 109493216A
- Authority
- CN
- China
- Prior art keywords
- model
- block chain
- node
- data analysis
- analysis model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The disclosure is directed to a kind of model training method, device, system and storage mediums, this method is applied to model optimization node, this method comprises: calling the intelligent contract stored in block chain, obtain the first block chain address of Data Analysis Model, wherein, the first block chain address is that model possesses node and obtains and be written in intelligent contract after block chain is written in Data Analysis Model;The block chain is inquired according to the first block chain address, obtains Data Analysis Model;Using the first preset training dataset training data analysis model, model optimization information is obtained;Model optimization information is written in block chain, the second block chain address is obtained;Second block chain address is written in intelligent contract, the second block chain address possesses node for model and obtains model optimization information.Pass through the technical solution of the embodiment of the present disclosure, it is ensured that privacy of user overcomes the problems, such as in the related technology because training obtained model accuracy not high caused by training dataset lazy weight.
Description
Technical field
This disclosure relates to field of communication technology more particularly to a kind of model training method, device, system and storage medium.
Background technique
Artificial intelligence is that data analyze indispensable a part, when carrying out data analysis using artificial intelligence technology,
It is generally necessary to a large amount of suitable training datasets is selected to carry out training pattern, and the quantity of training dataset often determines a mould
The key factor of type performance.
However, partially can be used as the data of training dataset because being related to privacy of user for some specific application scenarios
And use can not be shared, lead to the lazy weight of training dataset, and then cause the precision of the obtained model of training not high.
Summary of the invention
To overcome the problems in correlation technique, the disclosure provides a kind of model training method, device, system and storage
Medium.
According to the first aspect of the embodiments of the present disclosure, a kind of model training systems are provided, including model possesses node and mould
Type optimizes node, and the model possesses node and the model optimization node is the node in block chain network;
The model possesses node and is used for, and Data Analysis Model is written in block chain, the first block chain address is obtained, will
The first block chain address is written in the intelligent contract stored in the block chain;
The model optimization node is used for, and is called the intelligent contract, is obtained the first block chain address;According to described
First block chain address inquires the block chain, obtains the Data Analysis Model;It is assembled for training using the first preset training data
Practice the Data Analysis Model, obtains model optimization information;The model optimization information is written in the block chain, obtains the
Two block chain addresses;It will be in the second block chain address write-in intelligent contract;
The model possesses node and is also used to, and calls the intelligent contract, the second block chain address is obtained, according to institute
It states the second block chain address and inquires the block chain, obtain the model optimization information;Utilize the model optimization information update
The Data Analysis Model obtains target data analysis model.
Optionally, the model possesses node and is used for:
Based on preset homomorphic encryption algorithm, the Data Analysis Model is encrypted using public key;
Encrypted Data Analysis Model is written in the block chain;
The model optimization node is used for:
Based on the homomorphic encryption algorithm, mould is analyzed using the first training dataset training encrypted data
Type obtains the model optimization information.
Optionally, the model possesses node and is used for:
The model optimization information is decrypted using private key;
Using Data Analysis Model described in the model optimization information update after decryption, the target data analysis mould is obtained
Type.
Optionally, the model training systems further include model using node, and the model is block link network using node
Node in network;
The model possesses node and is also used to:
The target data analysis model is written in the block chain, third block chain address is obtained;
It will be in the third block chain address write-in intelligent contract;
The model is used for using node:
The intelligent contract is called, the third block chain address is obtained;
The block chain is inquired according to the third block chain address, obtains the target data analysis model;
It is analysed to data and inputs the target data analysis model, obtain analysis result.
Optionally, the model possesses node and is used for:
Based on preset homomorphic encryption algorithm, encrypted using Data Analysis Model of the public key to the target;
The block chain is written into encrypted target data analysis model;
The model is used for using node:
Based on the homomorphic encryption algorithm, it is analysed to data and inputs the encrypted Data Analysis Model, obtain institute
State analysis result.
Optionally, the model is also used to using node:
The analysis result is written in the block chain, the 4th block chain address is obtained;
It will be in the 4th block chain address write-in intelligent contract;
The model possesses node and is also used to:
The intelligent contract is called, the 4th block chain address is obtained;
The block chain is inquired according to the 4th block chain address, obtains the analysis result;
The analysis result is decrypted using private key, and the 4th block chain is written into the analysis result after decryption
In the block of the corresponding block chain in address;
The model is also used to using node:
The block chain is inquired according to the 4th block chain address, the analysis result after obtaining the decryption.
Optionally, the model optimization information includes the data analysis mould after the Data Analysis Model or optimization after optimization
The gradient value of type.
According to the second aspect of an embodiment of the present disclosure, a kind of model training method is provided, model optimization node, institute are applied to
Stating model optimization node is the node in block chain network, which comprises
The intelligent contract stored in block chain is called, obtains the first block chain address of Data Analysis Model, wherein described
First block chain address is that model possesses node and obtains after the block chain is written in the Data Analysis Model and institute is written
It states in intelligent contract, it is the node in the block chain network that the model, which possesses node,;
The block chain is inquired according to the first block chain address, obtains the Data Analysis Model;
Using preset the first training dataset training Data Analysis Model, model optimization information is obtained;
The model optimization information is written in the block chain, the second block chain address is obtained;
By in the second block chain address write-in intelligent contract, the second block chain address is used for the model
Possess node and obtains the model optimization information.
Optionally, the Data Analysis Model is that the model possesses node based on preset homomorphic encryption algorithm, is utilized
Public key carries out encryption and is written in the block chain;
It is described to train the Data Analysis Model using the first preset training dataset, model optimization information is obtained, is wrapped
It includes:
Based on the homomorphic encryption algorithm, mould is analyzed using the first training dataset training encrypted data
Type obtains the model optimization information.
Optionally, the model optimization information includes the data analysis mould after the Data Analysis Model or optimization after optimization
The gradient value of type.
According to the third aspect of an embodiment of the present disclosure, a kind of model training method is provided, possesses node applied to model, institute
Stating model to possess node is the node in block chain network, which comprises
Data Analysis Model is written in block chain, the first block chain address is obtained;
The first block chain address is written in the intelligent contract stored in the block chain, the first block chain
Location obtains the Data Analysis Model for model optimization node and is written in the intelligent contract for the data point
Analyse the second block chain address of the model optimization information of model, wherein the model optimization node is in the block chain network
Node;
The intelligent contract is called, the second block chain address is obtained;
The block chain is inquired according to the second block chain address, obtains the model optimization information;
Using Data Analysis Model described in the model optimization information update, target data analysis model is obtained.
It is optionally, described that Data Analysis Model is written in block chain, comprising:
Based on preset homomorphic encryption algorithm, the Data Analysis Model is encrypted using public key;
Encrypted Data Analysis Model is written in the block chain.
It is optionally, described to utilize Data Analysis Model described in the model optimization information update, comprising:
The model optimization information is decrypted using private key;
Utilize Data Analysis Model described in the model optimization information update after decryption.
Optionally, the method also includes:
The target data analysis model is written in the block chain, third block chain address is obtained;
By in the third block chain address write-in intelligent contract, the third block chain address is used for model
Node obtains the target data analysis model, wherein the model uses the node that node is in block chain network.
It is optionally, described that the target data analysis model is written in the block chain, comprising:
Based on preset homomorphic encryption algorithm, the target data analysis model is encrypted using public key;
Encrypted target data analysis model is written in the block chain.
According to a fourth aspect of embodiments of the present disclosure, a kind of model training apparatus is provided, model optimization node, institute are applied to
Stating model optimization node is the node in block chain network, and described device includes:
First calling module is configured as calling the intelligent contract that stores in block chain, obtains the of Data Analysis Model
One block chain address, wherein the first block chain address be model possess node by the Data Analysis Model be written institute
It obtains and is written in the intelligent contract after stating block chain, it is the section in the block chain network that the model, which possesses node,
Point;
First enquiry module is configured as inquiring the block chain according to the first block chain address, obtains the number
According to analysis model;
First training module is configured as obtaining using preset the first training dataset training Data Analysis Model
To model optimization information;
First writing module is configured as the model optimization information being written in the block chain, obtains the second block
Chain address;
Second writing module is configured as the second block chain address being written in the intelligent contract, described second
Block chain address possesses node for the model and obtains the model optimization information.
Optionally, the Data Analysis Model is that the model possesses node based on preset homomorphic encryption algorithm, is utilized
Public key carries out encryption and is written in the block chain, and the first training module includes:
First training submodule is configured as assembling for training based on the homomorphic encryption algorithm using first training data
Practice the encrypted Data Analysis Model, obtains the model optimization information.
Optionally, the model optimization information includes the data analysis mould after the Data Analysis Model or optimization after optimization
The gradient value of type.
According to a fifth aspect of the embodiments of the present disclosure, a kind of model training apparatus is provided, possesses node applied to model, institute
Stating model to possess node is the node in block chain network, and described device includes:
Third writing module is configured as Data Analysis Model being written in block chain, obtains the first block chain address;
4th writing module is configured as that the intelligence conjunction stored in the block chain is written the first block chain address into
In about, the first block chain address is for the model optimization node acquisition Data Analysis Model and in the intelligent contract
Second block chain address of the middle write-in for the model optimization information of the Data Analysis Model, wherein the model optimization section
Point is the node in the block chain network;
Second calling module is configured as calling the intelligent contract, obtains the second block chain address;
Second enquiry module is configured as inquiring the block chain according to the second block chain address, obtains the mould
Type optimizes information;
Update module is configured as obtaining number of targets using Data Analysis Model described in the model optimization information update
According to analysis model.
Optionally, the third writing module includes:
First encryption submodule is configured as analyzing the data using public key based on preset homomorphic encryption algorithm
Model is encrypted;
First write-in submodule, is configured as encrypted Data Analysis Model being written in the block chain.
Optionally, the update module includes:
First decryption submodule, is configured as that the model optimization information is decrypted using private key;
Submodule is updated, is configured as utilizing Data Analysis Model described in the model optimization information update after decryption.
Optionally, described device further include:
5th writing module is configured as the target data analysis model being written in the block chain, obtains third
Block chain address;
6th writing module is configured as the third block chain address being written in the intelligent contract, the third
Block chain address obtains the target data analysis model using node for model, wherein the model is area using node
Node in block chain network.
Optionally, the 5th writing module includes:
Second encryption submodule, is configured as based on preset homomorphic encryption algorithm, using public key to the target data
Analysis model is encrypted;
Second write-in submodule, is configured as encrypted target data analysis model being written in the block chain.
According to a sixth aspect of an embodiment of the present disclosure, a kind of model training apparatus is provided, model optimization node, institute are applied to
Stating model optimization node is the node in block chain network, and described device includes: processor;For the executable finger of storage processor
The memory of order;Wherein, the processor is configured to: call the intelligent contract that stores in block chain, obtain data analysis mould
First block chain address of type, wherein the first block chain address is that model possesses node by the Data Analysis Model
It obtains and is written in the intelligent contract after the block chain is written, it is in the block chain network that the model, which possesses node,
Node;Using preset the first training dataset training Data Analysis Model, model optimization information is obtained;By the mould
Type optimizes information and is written in the block chain, obtains the second block chain address;The intelligence is written into the second block chain address
In energy contract, the second block chain address possesses node for the model and obtains the model optimization information.
According to the 7th of the embodiment of the present disclosure the aspect, a kind of model training apparatus is provided, possesses node applied to model, institute
Stating model to possess node is the node in block chain network, and described device includes: processor;For the executable finger of storage processor
The memory of order;Wherein, the processor is configured to: by Data Analysis Model be written block chain in, obtain the first block chain
Address;The first block chain address is written in the intelligent contract stored in the block chain, the first block chain address
It obtains the Data Analysis Model for model optimization node and is written in the intelligent contract and analyzed for the data
Second block chain address of the model optimization information of model, wherein the model optimization node is in the block chain network
Node;The intelligent contract is called, the second block chain address is obtained;The area is inquired according to the second block chain address
Block chain obtains the model optimization information;Using Data Analysis Model described in the model optimization information update, number of targets is obtained
According to analysis model.
According to the eighth aspect of the embodiment of the present disclosure, a kind of computer readable storage medium is provided, is stored thereon with calculating
Machine program instruction realizes the step of model training method provided by the disclosure second aspect when program instruction is executed by processor
Suddenly.
According to the 9th of the embodiment of the present disclosure the aspect, a kind of computer readable storage medium is provided, calculating is stored thereon with
Machine program instruction realizes the step of model training method provided by the disclosure third aspect when program instruction is executed by processor
Suddenly.
The technical scheme provided by this disclosed embodiment can include the following benefits:
The model for possessing Data Analysis Model is possessed into node by block chain and possesses the model of a large amount of training datasets
Optimization node connects, and model is possessed the Data Analysis Model in node write-in block chain by way of intelligent contract and is used
With model training, i.e. model optimization node can call the intelligent contract stored in block chain to obtain Data Analysis Model, utilize
Preset training dataset is trained Data Analysis Model, and block chain is written in the model optimization information that training is obtained
In, so that model possesses the available model optimization information of node.In this way, model optimization node is without sharing its training possessed
Data set, but locally carrying out model training and uploading model optimization information, it ensure that privacy of user, overcome the relevant technologies
The not high problem of the obtained model accuracy of training caused by the middle lazy weight because of training dataset.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The disclosure can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.
Fig. 1 is a kind of flow chart of model training method shown according to an exemplary embodiment, wherein this method application
In model optimization node;
Fig. 2 is a kind of flow chart of model training method shown according to an exemplary embodiment, wherein this method application
Possess node in model;
Fig. 3 is a kind of schematic diagram of implementation environment shown according to an exemplary embodiment;
Fig. 4 is a kind of Signalling exchange schematic diagram of model training method shown according to an exemplary embodiment;
Fig. 5 is a kind of block diagram of model training systems shown according to an exemplary embodiment;
Fig. 6 is a kind of block diagram of the model training systems shown according to another exemplary embodiment;
Fig. 7 is a kind of block diagram of model training apparatus shown according to an exemplary embodiment, wherein the device is applied to
Model optimization node;
Fig. 8 is a kind of block diagram of the model training apparatus shown according to another exemplary embodiment, wherein the device application
In model optimization node;
Fig. 9 is a kind of block diagram of model training apparatus shown according to an exemplary embodiment, wherein the device is applied to
Model possesses node;
Figure 10 is a kind of block diagram of the model training apparatus shown according to another exemplary embodiment, wherein the device is answered
Possess node for model;
Figure 11 is a kind of block diagram of model training apparatus shown according to an exemplary embodiment.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.
It should be noted that the specification and claims of the disclosure and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, without being interpreted as specific sequence or precedence.
In order to make those skilled in the art be easier to understand the technical solution of embodiment of the present disclosure offer, first below to this
The open the relevant technologies being related to simply are introduced.
Block chain is the decentralization distributed data base system for participating in maintenance jointly by nodes all in block chain network,
It is by a series of data chunks generated based on cryptography method at each data block is a block in block chain.
According to the sequencing of generation time, block is linked together in an orderly manner, forms a data chain, is visually known as area
Block chain.Block chain is generated by its special block and transaction, indentification protocol can not forge with that can not change, is completely traceable
Security feature.
The related notion explanation being related in block chain technology:
Block chain node: block chain network is based on P2P (Peer to Peer, peer-to-peer network) network, each participates in business
With the node that block storage, verifying, the P2P network node forwarded are all in a block chain network.
The write-in of block chain data: block chain node to block chain network publication " transaction " (Transaction) by realizing
Data are written to block chain.Use oneself private key to the signature of transaction comprising user in transaction, to prove the identity of user.Transaction
The new block of generation is recorded by " miner " (the block chain node for executing block chain common recognition competition mechanism), is then published to block
Chain network, and after being passed through and received by other block chain node verifications, transaction data is written into block chain.
Intelligent contract: technically, intelligent contract is considered as network server, and two writing servers only are simultaneously
It is not set up on the internet, is erected on block chain, so as to run specific conjunction above it using IP address
About program.But unlike network server, intelligent contract does not depend on some specific hardware device, in fact, intelligence
The code of contract is executed by all equipment for participating in digging mine.
Intelligent contract is the assembler language being programmed on block chain, such as similar with Solidity or Javascript
Specific language pre-establishes intelligent contract when creating block chain.These bytecodes provide finger to the functionality of block chain really
Draw, therefore code can be easy to interact with it, such as transfer pin currency and record event.
The characteristic that block chain has:
Decentralization: entire block catenary system does not have the hardware of centralization or management organization, the power between arbitrary node
Benefit and obligation are all impartial, and the damage of any node or lose the running that can not all influence whole system.It therefore can also
To think that block chain has fabulous robustness.
It goes trustization: participating in carrying out data exchange between each node in entire block catenary system being without trusting each other
, the operation regulation of whole system is that open and clear, all data content is also rule that are disclosed, therefore specifying in system
It is that can not cannot also cheat other nodes then within the scope of scope and time, between node.
Based on block chain technology, the embodiment of the present disclosure provides a kind of model training method, and it is excellent that the method is applied to model
Change node, wherein model optimization node is the node in block chain network, as shown in Figure 1, method includes the following steps:
In step s 11, the intelligent contract stored in block chain is called, with obtaining the first block chain of Data Analysis Model
Location.
Referring to the above-mentioned introduction to intelligent contract, intelligent contract described in the embodiment of the present disclosure can be the block chain network
Each block chain node consensus formulation when creation.
In embodiment of the disclosure, the first block chain address is that model possesses node by Data Analysis Model write area
It obtains and is written in intelligent contract after block chain, wherein it is also the node in block chain network that model, which possesses node,.
In step s 12, block chain is inquired according to the first block chain address, obtains Data Analysis Model.
When model optimization node has model training demand, intelligence can be called by way of to intelligent contract write request
Contract triggers intelligent contract and executes code with to the first block chain address of model optimization node returned data analysis model, this
Sample, model optimization node, which inquires block chain by the first block chain address, can obtain Data Analysis Model.
In step s 13, using the first preset training dataset training data analysis model, model optimization letter is obtained
Breath.
Optionally, Data Analysis Model can be possesses node encrytion and writes direct in block chain without model,
It can be model and possess node based on homomorphic encryption algorithm, block is written after encrypting using public key to the Data Analysis Model
In chain.
For the former, model optimization node can directly utilize the first training dataset training Data Analysis Model.
For the latter, in this process, since the private key for only having model to possess node could analyze encrypted data
Model is decrypted and the private key only possesses node by model and grasps, thus model optimization node can not be to encrypted data point
Analysis model is decrypted, and can possess the identical homomorphic encryption algorithm of node based on model, utilizes the first preset training number
According to collect training the encrypted Data Analysis Model, correspondingly, training obtain model optimization information be also it is encrypted after information
And it is consistent with the result that the Data Analysis Model to unencryption is trained after model optimization information decryption.Due to entire
Data Analysis Model all keeps encrypted state in the process, and Model Fusion and iteration under encrypted state may be implemented, avoid data
Analysis model is stolen, and has ensured that model possesses the interests of node.
In step S14, model optimization information is written in block chain, the second block chain address is obtained.
In step S15, the second block chain address is written in intelligent contract.
In one possible implementation, model optimization information can be the Data Analysis Model after optimization, such
In the case of, directly the Data Analysis Model after optimization is written in block chain for model optimization node, and the second block that will be obtained
Chain address is written in intelligent contract, can get the data point after optimization by the intelligent contract of calling so that model possesses node
Analyse model.
In alternatively possible embodiment, model optimization information can also be the ladder of the Data Analysis Model after optimization
Only the gradient value obtained after optimization is written in block chain for angle value, in this case, model optimization node, and the will obtained
Two block chain addresses are written in intelligent contract, can obtain gradient value and benefit by the intelligent contract of calling so that model possesses node
Data Analysis Model is updated with gradient value.Relative to the first implementation, the storage resource of the block chain of which occupancy is more
Small, writing speed faster, and can be stolen to avoid the Data Analysis Model after optimization, ensure model optimization node
Interests.In addition, which is easier to assess the contribution margin of model optimization node.
It is worth noting that block chain described in the embodiment of the present disclosure can be alliance's chain, authorization node is allowed to be added
Block chain network, thus in the block chain network with permission control, the model optimization node and model, which possess node, is
The node for having information write-in and access limit.
Secondly, can have multiple model optimization nodes in block chain network described in the embodiment of the present disclosure, each model is excellent
Change node can be used above-mentioned model training method and optimize to Data Analysis Model.
In addition, model training described in the disclosure refers to the use artificial intelligence that can be trained in a distributed manner, machine learning, system
Meter is learned and the calculating process of the cross method of database discovery mode in relatively large-scale data set.For example, establishing depth
Neural network model, AI (Artificial Intelligence, artificial intelligence) training, Decision-Tree Classifier Model of study etc.,
The disclosure does not limit this.
Using the above method, the model for possessing Data Analysis Model is possessed by node by block chain and possesses a large amount of training
The model optimization node of data set connects, and model is possessed the number in node write-in block chain by way of intelligent contract
According to analysis model to model training, i.e. model optimization node can call the intelligent contract stored in block chain to obtain data point
The model optimization information analysed model, Data Analysis Model is trained using preset training dataset, and training is obtained
It is written in block chain, so that model possesses the available model optimization information of node.In this way, model optimization node is without sharing it
The training dataset possessed, but locally carrying out model training and uploading model optimization information, it ensure that privacy of user, overcome
The problem not high because of the obtained model accuracy of training caused by training dataset lazy weight in the related technology.
The embodiment of the present disclosure also provides a kind of model training method, and the method is applied to model and possesses node, wherein mould
It is the node in block chain network that type, which possesses node, as shown in Fig. 2, method includes the following steps:
In the step s 21, Data Analysis Model is written in block chain, obtains the first block chain address.
In one possible implementation, model, which possesses node, block chain directly is written by Data Analysis Model
In, in this case, other nodes (such as model optimization node) in block chain network can obtain the number from block chain
According to analysis model.
In alternatively possible implementation, model, which possesses node, can utilize its public key based on homomorphic encryption algorithm
It is written in block chain after being encrypted to Data Analysis Model, in this case, other nodes in block chain network (such as
Model optimization node) encrypted Data Analysis Model can be obtained from block chain.Since only model possesses the private key of node
Encrypted Data Analysis Model could be decrypted and the private key only possesses node by model and grasps, thus block link network
Other any nodes in network can not all be decrypted encrypted Data Analysis Model, and then can analyze mould to avoid data
Type is stolen, and has ensured that model possesses the interests of node.
In step S22, the first block chain address is written in the intelligent contract stored in block chain, the block chain
Location obtains Data Analysis Model and mould of the write-in for the Data Analysis Model in intelligent contract for model optimization node
Second block chain address of type optimization information.
Model, which possesses node, the first block chain address can be written in intelligent contract, in this way, intelligent contract is by model
When optimizing node calling, the first block chain address can be returned into the model optimization node.
Intelligent contract and model optimization node call the process of the intelligence contract referring to above-mentioned saying to method shown in Fig. 1
Bright, details are not described herein again.
In step S23, intelligent contract is called, the second block chain address is obtained.
In step s 24, block chain is inquired according to the second block chain address, obtains model optimization information.
It, can be by way of to intelligent contract write request when model possesses node and has Data Analysis Model to upgrade demand
Intelligent contract is called, intelligent contract is triggered and executes code to possess the second block chain that node returns to model optimization information to model
Address, in this way, model, which possesses node, can obtain model optimization information by the second block chain address inquiry block chain.
In step s 25, using model optimization information update Data Analysis Model, target data analysis model is obtained.
In a kind of possible embodiment, model optimization information can be the Data Analysis Model after optimization, such
In the case of, model possesses the essence that node can use the Data Analysis Model that preset predictive data set is assessed respectively after the optimization
The precision of degree and Data Analysis Model, and using the higher one of precision as target data analysis model.
In alternatively possible embodiment, model optimization information can be the gradient of the Data Analysis Model after optimization
Value, in this case, model possess node can use gradient value and Data Analysis Model optimize after data analyze
Model, and the precision and Data Analysis Model of the Data Analysis Model after the optimization are assessed using preset predictive data set respectively
Precision, and using the higher one of precision as target data analysis model.
It is worth noting that can have multiple model optimization nodes in block chain network described in the embodiment of the present disclosure, often
One model optimization node can obtain Data Analysis Model from block chain, utilize the respective training dataset training data
Simultaneously the model optimization information respectively obtained is written in block chain for analysis model.In this case, model possesses node and passes through
The block chain address for calling the available multiple model optimization information of intelligent contract inquires block chain by each block chain address,
Available all model optimization information analyzes mould using the data after the available multiple optimizations of these model optimization information
Type assesses the precision of the Data Analysis Model after each optimization and the essence of Data Analysis Model by preset predictive data set respectively
Degree, and it regard one of precision highest as target data analysis model.
In addition, possessing node based on preset homomorphic encryption algorithm, using its public key to Data Analysis Model for model
The implementation in block chain is written after being encrypted, since model optimization node can not be to encrypted Data Analysis Model solution
It is close and can only be based on identical homomorphic encryption algorithm, keep Data Analysis Model encryption in the case where carry out model training, phase
Ying Di, model optimization node obtain and the model optimization information that is written in block chain be also it is encrypted after and the model optimization
It is consistent with the result that the Data Analysis Model to unencryption is trained after information decryption.Therefore, in this case, mould
Type, which possesses node, can be decrypted model optimization information first with its private key, and the model optimization information after recycling decryption is more
New data analysis model obtains target data analysis model.
Further, in another embodiment, model possess node can also be by target data analysis model write area
In block chain, third block chain address is obtained, and third block chain address is written in intelligent contract, so as in block chain network
Model is obtained and is used the target data analysis model using node.
It is worth noting that model, which possesses node, can be divided target data using its public key based on homomorphic encryption algorithm
Analyse model write-in block chain in, in this case, model using node can not to encrypted target data analysis model into
Row decryption can be based on identical homomorphic encryption algorithm, be analysed to data and input the encrypted target data analysis mould
Type is to obtain analysis result.Since target data analysis model, and then can be to avoid mesh all in encrypted state in whole process
Mark Data Analysis Model is stolen, and assurance model possesses the interests of node.
Further, the analysis result that model is obtained using node the is also in encrypted state and analysis result is decrypted
Consistent with the target data analysis model progress result of operation using unencryption afterwards, in this case, model uses node
It can will be written in intelligent contract in analysis result write-in block chain and by the block chain address of obtained analysis result, by model
Possess node and obtains analysis by way of calling intelligent contract and inquiry block chain as a result, and using its private key to analysis result
It is decrypted, then block chain is written into the analysis result after decryption, and then model can be obtained using node by inquiring block chain
Analysis result after getting decryption.
Using the above method, the model for possessing Data Analysis Model is possessed by node by block chain and possesses a large amount of training
The model optimization node of data set connects, and model is possessed the number in node write-in block chain by way of intelligent contract
According to analysis model to model training, i.e. model optimization node can call the intelligent contract stored in block chain to obtain data point
The model optimization information analysed model, Data Analysis Model is trained using preset training dataset, and training is obtained
It is written in block chain.And model possesses node and can obtain model optimization information by the intelligent contract of calling and utilize model optimization
Information update Data Analysis Model.In this way, the training dataset that it possesses without model optimization nodes sharing, only needs model optimization
Node uploads model optimization information, and model possesses node using model optimization information update Data Analysis Model, ensure that user
Privacy overcomes problem not high because of the obtained model accuracy of training caused by training dataset lazy weight in the related technology.
For the technical solution for making those skilled in the art more understand that the embodiment of the present disclosure provides, said below with reference to Fig. 3
The model training method that the bright embodiment of the present disclosure provides.
Fig. 3 is a kind of schematic diagram of implementation environment shown according to an exemplary embodiment, as shown in figure 3, the implementation ring
Border includes possessing the model of Data Analysis Model to possess node 301, model optimization node 302, model using node 303 and area
Block chain network 304, wherein it is block chain that model, which possesses node 301, model optimization node 302 and model using node 303,
Node in network 304.In conjunction with Fig. 3, a kind of flow chart of model training method shown in one exemplary embodiment of the disclosure is such as
Shown in Fig. 4, method includes the following steps:
In step S41, model possesses node 301 and Data Analysis Model is written in block chain, obtains the first block chain
Address.
Optionally, model possesses node 301 and directly Data Analysis Model can be written in block chain, can also be based on same
State Encryption Algorithm is written in block chain after being encrypted using its public key to Data Analysis Model.
In step S42, model possesses node 301 and first block chain address is written to the intelligent contract stored in block chain
In.
In step S43, model optimization node 302 calls intelligent contract, obtains the first block chain address.
In step S44, model optimization node 302 inquires block chain according to the first block chain address, obtains data analysis
Model.
In step S45, model optimization node 302 utilizes the first preset training dataset training data analysis model,
Obtain model optimization information.
Optionally, if Data Analysis Model is to possess the encryption of node 301 without model, model optimization node 302 can
To utilize the first training dataset training data analysis model.
Optionally, if Data Analysis Model is to possess node 301 through model to be based on preset homomorphic encryption algorithm, utilize public affairs
What key was encrypted, model optimization node 302 can then be based on identical homomorphic encryption algorithm, straight using the first training dataset
Connect the training encrypted Data Analysis Model, correspondingly, the obtained model optimization information of training be also it is encrypted after information
And it is consistent with the result that the Data Analysis Model to unencryption is trained after model optimization information decryption.In entire mistake
Cheng Zhong, Data Analysis Model all keep encrypted state.
In step S46, model optimization information is written in block chain model optimization node 302, obtains the second block chain
Address.
Optionally, model optimization information can be the Data Analysis Model after optimization, the data after being also possible to optimization point
Analyse the gradient value of model.
In step S47, the second block chain address is written in intelligent contract model optimization node 302.
In step S48, model possesses node 301 and calls intelligent contract, obtains the second block chain address.
In step S49, model possesses node 301 and inquires block chain according to the second block chain address, obtains model optimization
Information.
In step s 50, model possesses node 301 using model optimization information update Data Analysis Model, obtains target
Data Analysis Model.
Optionally, if model optimization information is the Data Analysis Model after optimization, model, which possesses node 301, be can use
Preset predictive data set assesses the precision of the Data Analysis Model after the optimization and the precision of Data Analysis Model respectively, and will
The higher one of precision is as target data analysis model.
Optionally, if model optimization information is the gradient value of the Data Analysis Model after optimization, model possesses node 301
It can use gradient value and Data Analysis Model determine the Data Analysis Model after optimization, and utilize preset predictive data set point
The precision of Data Analysis Model after Ping Gu not optimizing and the precision of Data Analysis Model, and using the higher one of precision as
Target data analysis model.
In step s 51, model possesses node 301 and target data analysis model is written in block chain, obtains third area
Block chain address.
Optionally, model possesses node 301 and directly target data analysis model can be written in block chain, can also be with base
In homomorphic encryption algorithm, be written in block chain after being encrypted using its public key to target data analysis model.
In step S52, model possesses node 301 and third block chain address is written in intelligent contract.
In step S53, model calls intelligent contract using node 303, obtains third block chain address.
In step S54, model inquires block chain according to third block chain address using node 303, obtains target data
Analysis model.
In step S55, model is analysed to data input target data analysis model using node 303, is analyzed
As a result.
Optionally, if target data analysis model is to possess the encryption of node 301 without model, model uses node 303
It can be then analysed to data input target data analysis model, obtain analysis result.
Optionally, if target data analysis model is to possess node 301 through model to be based on preset homomorphic encryption algorithm, benefit
Carried out with public key it is encrypted, model using node 303 then can be based on identical homomorphic encryption algorithm, it is straight to be analysed to data
Connect input the encrypted target data analysis model, correspondingly, the analysis result of output be also it is encrypted after result and this
It analyzes consistent with the result of target data analysis model output to unencryption after result is decrypted.In the whole process, number of targets
Encrypted state is all kept according to analysis model.
Further, model can will analyze the analysis result that in result write-in block chain and will be obtained using node 303
Block chain address be written in intelligent contract, node 301 is possessed by calling the side of intelligent contract and inquiry block chain by model
Formula obtain analysis as a result, and using its private key to analysis result be decrypted, then by after decryption analysis result be written block chain,
And then model can get the analysis result after decrypting by inquiring block chain using node 303.
Using the above method, the model for possessing Data Analysis Model is possessed by node by block chain, possesses a large amount of training
It the model optimization node of data set and is connected using the model of Data Analysis Model using node, passes through intelligent contract
Model is possessed the Data Analysis Model in node write-in block chain and analyzed to model training and data by mode, i.e. model optimization
Node can call the intelligent contract stored in block chain to obtain Data Analysis Model, using preset training dataset to data
Analysis model is trained, and in the model optimization information write-in block chain that training is obtained, model, which possesses node, to be passed through
It calls intelligent contract to obtain model optimization information and using model optimization information update Data Analysis Model and uploads update gained
Target data analysis model, model is using node by calling intelligent contract to obtain target data analysis model and using the mesh
It marks Data Analysis Model and carries out data analysis.In this way, the training dataset that it possesses without model optimization nodes sharing, only needs mould
Type optimizes node and uploads model optimization information, and model possesses node using model optimization information update Data Analysis Model, guarantees
Privacy of user, overcome in the related technology because caused by training dataset lazy weight the obtained model accuracy of training it is not high
Problem.
It is worth noting that for simple description, therefore, it is stated as a series of dynamic for above method embodiment
It combines, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described.Secondly, this
Field technical staff also should be aware of, and the embodiments described in the specification are all preferred embodiments, and related movement is simultaneously
It is not necessarily necessary to the present invention.
Fig. 5 is a kind of block diagram of model training systems shown according to an exemplary embodiment, as shown in figure 5, the system
500 include that model possesses node 501 and model optimization node 502.
The model possesses node 501 and is used for, and Data Analysis Model is written in block chain, the first block chain address is obtained,
The first block chain address is written in the intelligent contract stored in the block chain;
The model optimization node 502 is used for, and is called the intelligent contract, is obtained the first block chain address;According to institute
It states the first block chain address and inquires the block chain, obtain the Data Analysis Model;Utilize the first preset training dataset
The training Data Analysis Model, obtains model optimization information;The model optimization information is written in the block chain, is obtained
Second block chain address;It will be in the second block chain address write-in intelligent contract;
The model possesses node 501 and is also used to, and calls the intelligent contract, obtains the second block chain address, according to
The second block chain address inquires the block chain, obtains the model optimization information;More using the model optimization information
The new Data Analysis Model, obtains target data analysis model.
Optionally, the model possesses node 501 and is used for:
Based on preset homomorphic encryption algorithm, the Data Analysis Model is encrypted using public key;
Encrypted Data Analysis Model is written in the block chain;
The model optimization node 502 is used for:
Based on the homomorphic encryption algorithm, mould is analyzed using the first training dataset training encrypted data
Type obtains the model optimization information.
Optionally, the model possesses node 501 and is used for:
The model optimization information is decrypted using private key;
Using Data Analysis Model described in the model optimization information update after decryption, the target data analysis mould is obtained
Type.
In another embodiment, as shown in fig. 6, the model training systems 500 further include that model uses node 503,
The model uses the node that node 503 is in block chain network;
The model possesses node 501 and is also used to:
The target data analysis model is written in the block chain, third block chain address is obtained;
It will be in the third block chain address write-in intelligent contract;
The model is used for using node 503:
The intelligent contract is called, the third block chain address is obtained;
The block chain is inquired according to the third block chain address, obtains the target data analysis model;
It is analysed to data and inputs the target data analysis model, obtain analysis result.
Optionally, the model possesses node 501 and is used for:
Based on preset homomorphic encryption algorithm, encrypted using Data Analysis Model of the public key to the target;
The block chain is written into encrypted target data analysis model;
The model is used for using node 503:
Based on the homomorphic encryption algorithm, it is analysed to data and inputs the encrypted Data Analysis Model, obtain institute
State analysis result.
Optionally, the model is also used to using node 503:
The analysis result is written in the block chain, the 4th block chain address is obtained;
It will be in the 4th block chain address write-in intelligent contract;
The model possesses node 501 and is also used to:
The intelligent contract is called, the 4th block chain address is obtained;
The block chain is inquired according to the 4th block chain address, obtains the analysis result;
The analysis result is decrypted using private key, and the 4th block chain is written into the analysis result after decryption
In the block of the corresponding block chain in address;
The model is also used to using node 503:
The block chain is inquired according to the 4th block chain address, the analysis result after obtaining the decryption.
Optionally, the model optimization information includes the data analysis mould after the Data Analysis Model or optimization after optimization
The gradient value of type.
Those skilled in the art should be well understood, for convenience of description and succinctly, the mould of foregoing description
The specific work process of each node in type training system, can be with reference to corresponding process in preceding method embodiment, herein not
It repeats again.
Using above-mentioned model training systems, the model for possessing Data Analysis Model is possessed by node by block chain, is possessed
It the model optimization node of a large amount of training datasets and is connected using the model of Data Analysis Model using node, passes through intelligence
Model is possessed the Data Analysis Model in node write-in block chain and analyzed to model training and data by the mode of energy contract, i.e.,
Model optimization node can call the intelligent contract stored in block chain to obtain Data Analysis Model, utilize preset training data
Collection is trained Data Analysis Model, and in the model optimization information write-in block chain that training is obtained, model possesses node
And by the intelligent contract acquisition model optimization information of calling and using model optimization information update Data Analysis Model can upload
Resulting target data analysis model is updated, model is using node by calling intelligent contract to obtain target data analysis model simultaneously
Data analysis is carried out using the target data analysis model.In this way, the training data that it possesses without model optimization nodes sharing
Collection, only needs model optimization node to upload model optimization information, and model is possessed node and analyzed using model optimization information update data
Model ensure that privacy of user, overcome in the related technology because of the obtained model of training caused by training dataset lazy weight
The not high problem of precision.In addition, the model training systems relative to centralization deployment, the decentralization of block chain can reduce mould
The O&M cost of type training system.
It is worth noting that any of the above-described embodiment of the disclosure can be applied to different scenes.Illustratively, with medical treatment neck
For domain, model, which possesses node, can be machine learning or image recognition company, scientific research institution etc., possess data analysis mould
Type needs data to help the performance of lift scheme;Model optimization node can be large hospital, medical information company, section
Mechanism etc. is ground, a large amount of patient medical datas are possessed;Model can be medical APP company, medical robot, hospital using node
Deng possessing patient medical data to be analyzed.
Under the application scenarios, model possesses node without sharing its patient medical data possessed, but local into
Row model training simultaneously uploads model optimization information to block chain, possesses node by model and optimizes information from block chain download model,
It is uploaded to block chain using model optimization information update Data Analysis Model and by obtained target data analysis model is updated, and
The patient medical data being analysed to inputs target data analysis model, obtains analysis result.In this way, without making model optimization section
The shared patient medical data of point can just make model possess node updates Data Analysis Model, ensure that privacy of user, while can
Patient medical data to be analyzed is analyzed so that model can use model with better accuracy using node, is obtained more quasi-
True analysis result.
In addition, patient medical data can be medical imaging data (such as X-ray etc.) under the application scenarios, due to
Medical imaging data have versatility, can occur to avoid the incompatible problem of data format.
Fig. 7 is a kind of block diagram of model training apparatus shown according to an exemplary embodiment, which is applied to mould
Type optimizes node, wherein model optimization node is the node in block chain network, for implementing Fig. 1 in above method embodiment
Shown in method and step, as shown in fig. 7, the device 700 includes:
First calling module 701 is configured as calling the intelligent contract stored in block chain, obtains Data Analysis Model
First block chain address, wherein the first block chain address is that model possesses node and is written by the Data Analysis Model
It obtains and is written in the intelligent contract after the block chain, it is the section in the block chain network that the model, which possesses node,
Point;
First enquiry module 702 is configured as inquiring the block chain according to the first block chain address, obtains described
Data Analysis Model;
First training module 703 is configured as using preset the first training dataset training Data Analysis Model,
Obtain model optimization information;
First writing module 704 is configured as the model optimization information being written in the block chain, obtains the secondth area
Block chain address;
Second writing module 705 is configured as the second block chain address being written in the intelligent contract, and described the
Two block chain addresses possess node for the model and obtain the model optimization information.
Optionally, as shown in figure 8, the Data Analysis Model is that the model possesses node based on preset homomorphic cryptography
Algorithm carries out encryption using public key and is written in the block chain;
First training module 703 includes:
First training submodule 731 is configured as utilizing first training dataset based on the homomorphic encryption algorithm
The training encrypted Data Analysis Model, obtains the model optimization information.
Optionally, the model optimization information includes the data analysis mould after the Data Analysis Model or optimization after optimization
The gradient value of type.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
In addition, the above-mentioned division that the model training apparatus comprising modules for being applied to model optimization node are carried out, only one
Kind logical function partition, there may be another division manner in actual implementation.Also, the physics realization of modules can also have
Various ways, which is not limited by the present invention.
Using above-mentioned model training apparatus, the model for possessing Data Analysis Model is possessed by node by block chain and is possessed
The model optimization node of a large amount of training datasets connects, and model is possessed node write-in block by way of intelligent contract
For Data Analysis Model in chain to model training, i.e. model optimization node can call the intelligent contract stored in block chain to obtain
The model for taking Data Analysis Model, Data Analysis Model being trained using preset training dataset, and training is obtained
Optimize in information write-in block chain, so that model possesses the available model optimization information of node.In this way, model optimization node without
Its training dataset possessed need to be shared, but is locally carrying out model training and is uploading model optimization information, ensure that user
Privacy overcomes problem not high because of the obtained model accuracy of training caused by training dataset lazy weight in the related technology.
Fig. 9 is a kind of block diagram of model training apparatus shown according to an exemplary embodiment, which is applied to mould
Type possesses node, wherein it is the node in block chain network that model, which possesses node, for implementing Fig. 2 in above method embodiment
Shown in method and step, as shown in figure 9, the device 900 includes:
Third writing module 901 is configured as Data Analysis Model being written in block chain, with obtaining the first block chain
Location;
4th writing module 902 is configured as the first block chain address intelligence stored in the block chain is written
In energy contract, the first block chain address is for the model optimization node acquisition Data Analysis Model and in the intelligence
Second block chain address of the write-in for the model optimization information of the Data Analysis Model in contract, wherein the model is excellent
Changing node is the node in the block chain network;
Second calling module 903 is configured as calling the intelligent contract, obtains the second block chain address;
Second enquiry module 904 is configured as inquiring the block chain according to the second block chain address, obtains described
Model optimization information;
Update module 905 is configured as obtaining target using Data Analysis Model described in the model optimization information update
Data Analysis Model.
Optionally, as shown in Figure 10, the third writing module 901 includes:
First encryption submodule 911, is configured as based on preset homomorphic encryption algorithm, using public key to the data point
Analysis model is encrypted;
First write-in submodule 912, is configured as encrypted Data Analysis Model being written in the block chain.
Optionally, as shown in Figure 10, the update module 905 includes:
First decryption submodule 951, is configured as that the model optimization information is decrypted using private key;
Submodule 952 is updated, is configured as utilizing Data Analysis Model described in the model optimization information update after decryption.
Optionally, as shown in Figure 10, described device 900 further include:
5th writing module 906 is configured as the target data analysis model being written in the block chain, obtains the
Three block chain addresses;
6th writing module 907 is configured as the third block chain address being written in the intelligent contract, and described the
Three block chain addresses using node obtain the target data analysis model for model, wherein the model is using node
Node in block chain network.
Optionally, as shown in Figure 10, the 5th writing module 906 includes:
Second encryption submodule 961, is configured as based on preset homomorphic encryption algorithm, using public key to the number of targets
It is encrypted according to analysis model;
Second write-in submodule 962, is configured as encrypted target data analysis model being written in the block chain.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
In addition, the above-mentioned division that the model training apparatus comprising modules for possessing node applied to model are carried out, only one
Kind logical function partition, there may be another division manner in actual implementation.Also, the physics realization of modules can also have
Various ways, which is not limited by the present invention.
Using above-mentioned model training apparatus, the model for possessing Data Analysis Model is possessed by node by block chain and is possessed
The model optimization node of a large amount of training datasets connects, and model is possessed node write-in block by way of intelligent contract
For Data Analysis Model in chain to model training, i.e. model optimization node can call the intelligent contract stored in block chain to obtain
The model for taking Data Analysis Model, Data Analysis Model being trained using preset training dataset, and training is obtained
Optimize in information write-in block chain.And model possesses node and can obtain model optimization information by the intelligent contract of calling and utilize
Model optimization information update Data Analysis Model.In this way, the training dataset that it possesses without model optimization nodes sharing, only needs
Model optimization node uploads model optimization information, and model possesses node using model optimization information update Data Analysis Model, protects
Privacy of user has been demonstrate,proved, has been overcome in the related technology because training obtained model accuracy not high caused by training dataset lazy weight
The problem of.
The embodiment of the present disclosure also provides a kind of computer readable storage medium, is stored thereon with computer program instructions, should
The step of above-mentioned model training method shown in FIG. 1 is realized when program instruction is executed by processor.
Correspondingly, the embodiment of the present disclosure also provides a kind of model training apparatus, which is applied to model optimization node, institute
Stating model optimization node is the node in block chain network, and described device includes: processor;For the executable finger of storage processor
The memory of order;Wherein, the processor is configured to: execute above-mentioned model training method shown in FIG. 1.
The embodiment of the present disclosure also provides a kind of computer readable storage medium, is stored thereon with computer program instructions, should
The step of above-mentioned model training method shown in Fig. 2 is realized when program instruction is executed by processor.
Correspondingly, the embodiment of the present disclosure also provides a kind of model training apparatus, which is applied to model and possesses node, institute
Stating model to possess node is the node in block chain network, and described device includes: processor;For the executable finger of storage processor
The memory of order;Wherein, the processor is configured to: execute above-mentioned model training method shown in Fig. 2.
Figure 11 is a kind of block diagram of model training apparatus 1100 shown according to an exemplary embodiment, which can
Possess node applied to model optimization node or model, wherein it is block chain that model optimization node and model, which possess node,
Node in network.For example, device 1100 can be mobile phone, computer, digital broadcasting terminal, messaging device is put down
Panel device, personal digital assistant etc..
Referring to Fig.1 1, device 1100 may include following one or more components: processing component 1102, memory 1104,
Electric power assembly 1106, multimedia component 1108, audio component 1110, the interface 1112 of input/output (I/O), sensor module
1114 and communication component 1116.
The integrated operation of the usual control device 1100 of processing component 1102, such as with display, telephone call, data communication,
Camera operation and record operate associated operation.Processing component 1102 may include one or more processors 1120 to execute
Instruction, wherein when the device 1100 is applied to model optimization node, processing component 1102 can be completed above-mentioned shown in FIG. 1
The all or part of the steps of model training method;When the device 1100, which is applied to model, possesses node, processing component 1102 can
To complete all or part of the steps of above-mentioned model training method shown in Fig. 2.In addition, processing component 1102 may include one
Or multiple modules, convenient for the interaction between processing component 1102 and other assemblies.For example, processing component 1102 may include more matchmakers
Module, to facilitate the interaction between multimedia component 1108 and processing component 1102.
Memory 1104 is configured as storing various types of data to support the operation in device 1100.These data
Example includes the instruction of any application or method for operating on device 1100, contact data, telephone book data,
Message, picture, video etc..Memory 1104 can by any kind of volatibility or non-volatile memory device or they
Combination is realized, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), it is erasable can
Program read-only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash memory
Reservoir, disk or CD.
Electric power assembly 1106 provides electric power for the various assemblies of device 1100.Electric power assembly 1106 may include power management
System, one or more power supplys and other with for device 1100 generate, manage, and distribute the associated component of electric power.
Multimedia component 1108 includes the screen of one output interface of offer between described device 1100 and user.?
In some embodiments, screen may include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel,
Screen may be implemented as touch screen, to receive input signal from the user.Touch panel includes that one or more touch passes
Sensor is to sense the gesture on touch, slide, and touch panel.The touch sensor can not only sense touch or sliding is dynamic
The boundary of work, but also detect duration and pressure associated with the touch or slide operation.In some embodiments, more
Media component 1108 includes a front camera and/or rear camera.When device 1100 is in operation mode, as shot mould
When formula or video mode, front camera and/or rear camera can receive external multi-medium data.Each preposition camera shooting
Head and rear camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 1110 is configured as output and/or input audio signal.For example, audio component 1110 includes a wheat
Gram wind (MIC), when device 1100 is in operation mode, when such as call mode, recording mode, and voice recognition mode, microphone quilt
It is configured to receive external audio signal.The received audio signal can be further stored in memory 1104 or via communication
Component 1116 is sent.In some embodiments, audio component 1110 further includes a loudspeaker, is used for output audio signal.
I/O interface 1112 provides interface, above-mentioned peripheral interface module between processing component 1102 and peripheral interface module
It can be keyboard, click wheel, button etc..These buttons may include, but are not limited to: home button, volume button, start button and
Locking press button.
Sensor module 1114 includes one or more sensors, and the state for providing various aspects for device 1100 is commented
Estimate.For example, sensor module 1114 can detecte the state that opens/closes of device 1100, the relative positioning of component, such as institute
The display and keypad that component is device 1100 are stated, sensor module 1114 can be with detection device 1100 or device 1,100 1
The position change of a component, the existence or non-existence that user contacts with device 1100,1100 orientation of device or acceleration/deceleration and dress
Set 1100 temperature change.Sensor module 1114 may include proximity sensor, be configured in not any physics
It is detected the presence of nearby objects when contact.Sensor module 1114 can also include optical sensor, as CMOS or ccd image are sensed
Device, for being used in imaging applications.In some embodiments, which can also include acceleration sensing
Device, gyro sensor, Magnetic Sensor, pressure sensor or temperature sensor.
Communication component 1116 is configured to facilitate the communication of wired or wireless way between device 1100 and other equipment.Dress
The wireless network based on communication standard, such as WiFi can be accessed by setting 1100,2G or 3G or their combination.It is exemplary at one
In embodiment, communication component 1116 receives broadcast singal or broadcast correlation from external broadcasting management system via broadcast channel
Information.In one exemplary embodiment, the communication component 1116 further includes near-field communication (NFC) module, to promote short distance
Communication.For example, radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band can be based in NFC module
(UWB) technology, bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, device 1100 can be by one or more application specific integrated circuit (ASIC), number
Signal processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, for executing above-mentioned Fig. 1 (when the device
1100 be applied to model optimization node when) or Fig. 2 (when the device 1100 be applied to model possess node when) shown in model training
Method.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instruction, example are additionally provided
It such as include the memory 1104 of instruction, above-metioned instruction can be executed by the processor 1120 of device 1100 to complete above-mentioned Fig. 1 (when this
When device 1100 is applied to model optimization node) or Fig. 2 (when the device 1100, which is applied to model, possesses node) institute's representation model
Training method.For example, the non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-
ROM, tape, floppy disk and optical data storage devices etc..
Those skilled in the art will readily occur to other embodiment party of the disclosure after considering specification and practicing the disclosure
Case.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or adaptability
Variation follows the general principles of this disclosure and including the undocumented common knowledge or usual skill in the art of the disclosure
Art means.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by following claim
It points out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.
Claims (27)
1. a kind of model training systems, which is characterized in that possess node and model optimization node including model, the model possesses
Node and the model optimization node are the nodes in block chain network;
The model possesses node and is used for, and Data Analysis Model is written in block chain, the first block chain address is obtained, will be described
First block chain address is written in the intelligent contract stored in the block chain;
The model optimization node is used for, and is called the intelligent contract, is obtained the first block chain address;According to described first
Block chain address inquires the block chain, obtains the Data Analysis Model;Utilize preset the first training dataset training institute
Data Analysis Model is stated, model optimization information is obtained;The model optimization information is written in the block chain, the secondth area is obtained
Block chain address;It will be in the second block chain address write-in intelligent contract;
The model possesses node and is also used to, and calls the intelligent contract, obtains the second block chain address, according to described the
Two block chain addresses inquire the block chain, obtain the model optimization information;Using described in the model optimization information update
Data Analysis Model obtains target data analysis model.
2. model training systems according to claim 1, which is characterized in that the model possesses node and is used for:
Based on preset homomorphic encryption algorithm, the Data Analysis Model is encrypted using public key;
Encrypted Data Analysis Model is written in the block chain;
The model optimization node is used for:
Based on the homomorphic encryption algorithm, the encrypted Data Analysis Model is trained using first training dataset,
Obtain the model optimization information.
3. model training systems according to claim 2, which is characterized in that the model possesses node and is used for:
The model optimization information is decrypted using private key;
Using Data Analysis Model described in the model optimization information update after decryption, the target data analysis model is obtained.
4. model training systems according to claim 1, which is characterized in that the model training systems further include that model makes
With node, the model uses the node that node is in the block chain network;
The model possesses node and is also used to:
The target data analysis model is written in the block chain, third block chain address is obtained;
It will be in the third block chain address write-in intelligent contract;
The model is used for using node:
The intelligent contract is called, the third block chain address is obtained;
The block chain is inquired according to the third block chain address, obtains the target data analysis model;
It is analysed to data and inputs the target data analysis model, obtain analysis result.
5. model training systems according to claim 4, which is characterized in that the model possesses node and is used for:
Based on preset homomorphic encryption algorithm, encrypted using Data Analysis Model of the public key to the target;
The block chain is written into encrypted target data analysis model;
The model is used for using node:
It based on the homomorphic encryption algorithm, is analysed to data and inputs the encrypted Data Analysis Model, obtain described point
Analyse result.
6. model training systems according to claim 5, which is characterized in that the model is also used to using node:
The analysis result is written in the block chain, the 4th block chain address is obtained;
It will be in the 4th block chain address write-in intelligent contract;
The model possesses node and is also used to:
The intelligent contract is called, the 4th block chain address is obtained;
The block chain is inquired according to the 4th block chain address, obtains the analysis result;
The analysis result is decrypted using private key, and the 4th block chain address is written into the analysis result after decryption
In the block of the corresponding block chain;
The model is also used to using node:
The block chain is inquired according to the 4th block chain address, the analysis result after obtaining the decryption.
7. model training systems described according to claim 1~any one of 6, which is characterized in that the model optimization information
Gradient value including the Data Analysis Model after the Data Analysis Model or optimization after optimization.
8. a kind of model training method, which is characterized in that be applied to model optimization node, the model optimization node is block chain
Node in network, which comprises
The intelligent contract stored in block chain is called, obtains the first block chain address of Data Analysis Model, wherein described first
Block chain address is that model possesses node and obtains after the block chain is written in the Data Analysis Model and the intelligence is written
In energy contract, it is the node in the block chain network that the model, which possesses node,;
The block chain is inquired according to the first block chain address, obtains the Data Analysis Model;
Using preset the first training dataset training Data Analysis Model, model optimization information is obtained;
The model optimization information is written in the block chain, the second block chain address is obtained;
By in the second block chain address write-in intelligent contract, the second block chain address possesses for the model
Node obtains the model optimization information.
9. according to the method described in claim 8, it is characterized in that, the Data Analysis Model is that the model possesses node base
In preset homomorphic encryption algorithm, encryption is carried out using public key and is written in the block chain;
It is described to train the Data Analysis Model using the first preset training dataset, obtain model optimization information, comprising:
Based on the homomorphic encryption algorithm, the encrypted Data Analysis Model is trained using first training dataset,
Obtain the model optimization information.
10. method according to claim 8 or claim 9, which is characterized in that the model optimization information includes the data after optimization
The gradient value of Data Analysis Model after analysis model or optimization.
11. a kind of model training method, which is characterized in that possess node applied to model, it is block that the model, which possesses node,
Node in chain network, which comprises
Data Analysis Model is written in block chain, the first block chain address is obtained;
The first block chain address is written in the intelligent contract stored in the block chain, the first block chain address is used
It obtains the Data Analysis Model in model optimization node and is written in the intelligent contract and analyze mould for the data
Second block chain address of the model optimization information of type, wherein the model optimization node is the section in the block chain network
Point;
The intelligent contract is called, the second block chain address is obtained;
The block chain is inquired according to the second block chain address, obtains the model optimization information;
Using Data Analysis Model described in the model optimization information update, target data analysis model is obtained.
12. according to the method for claim 11, which is characterized in that described that Data Analysis Model is written in block chain, packet
It includes:
Based on preset homomorphic encryption algorithm, the Data Analysis Model is encrypted using public key;
Encrypted Data Analysis Model is written in the block chain.
13. according to the method for claim 12, which is characterized in that described to utilize number described in the model optimization information update
According to analysis model, comprising:
The model optimization information is decrypted using private key;
Utilize Data Analysis Model described in the model optimization information update after decryption.
14. according to the method for claim 11, which is characterized in that the method also includes:
The target data analysis model is written in the block chain, third block chain address is obtained;
By in the third block chain address write-in intelligent contract, the third block chain address uses node for model
Obtain the target data analysis model, wherein the model uses the node that node is in block chain network.
15. according to the method for claim 14, which is characterized in that it is described will the target data analysis model write-in described in
In block chain, comprising:
Based on preset homomorphic encryption algorithm, the target data analysis model is encrypted using public key;
Encrypted target data analysis model is written in the block chain.
16. a kind of model training apparatus, which is characterized in that be applied to model optimization node, the model optimization node is block
Node in chain network, described device include:
First calling module is configured as calling the intelligent contract stored in block chain, obtains the firstth area of Data Analysis Model
Block chain address, wherein the first block chain address is that model possesses node the area is being written in the Data Analysis Model
It obtains and is written in the intelligent contract after block chain, it is the node in the block chain network that the model, which possesses node,;
First enquiry module is configured as inquiring the block chain according to the first block chain address, obtains the data point
Analyse model;
First training module is configured as obtaining mould using preset the first training dataset training Data Analysis Model
Type optimizes information;
First writing module is configured as the model optimization information being written in the block chain, with obtaining the second block chain
Location;
Second writing module is configured as the second block chain address being written in the intelligent contract, second block
Chain address possesses node for the model and obtains the model optimization information.
17. device according to claim 16, which is characterized in that the Data Analysis Model is that the model possesses node
Based on preset homomorphic encryption algorithm, encryption is carried out using public key and is written in the block chain, the first training module includes:
First training submodule is configured as utilizing first training dataset training institute based on the homomorphic encryption algorithm
Encrypted Data Analysis Model is stated, the model optimization information is obtained.
18. device according to claim 16 or 17, which is characterized in that the model optimization information includes the number after optimization
According to the gradient value of the Data Analysis Model after analysis model or optimization.
19. a kind of model training apparatus, which is characterized in that possess node applied to model, it is block that the model, which possesses node,
Node in chain network, described device include:
Third writing module is configured as Data Analysis Model being written in block chain, obtains the first block chain address;
4th writing module is configured as the first block chain address intelligent contract stored in the block chain is written
In, the first block chain address is for the model optimization node acquisition Data Analysis Model and in the intelligent contract
Second block chain address of the write-in for the model optimization information of the Data Analysis Model, wherein the model optimization node
It is the node in the block chain network;
Second calling module is configured as calling the intelligent contract, obtains the second block chain address;
Second enquiry module is configured as inquiring the block chain according to the second block chain address, it is excellent to obtain the model
Change information;
Update module is configured as obtaining target data point using Data Analysis Model described in the model optimization information update
Analyse model.
20. device according to claim 19, which is characterized in that the third writing module includes:
First encryption submodule, is configured as based on preset homomorphic encryption algorithm, using public key to the Data Analysis Model
It is encrypted;
First write-in submodule, is configured as encrypted Data Analysis Model being written in the block chain.
21. device according to claim 20, which is characterized in that the update module includes:
First decryption submodule, is configured as that the model optimization information is decrypted using private key;
Submodule is updated, is configured as utilizing Data Analysis Model described in the model optimization information update after decryption.
22. device according to claim 19, which is characterized in that described device further include:
5th writing module is configured as the target data analysis model being written in the block chain, obtains third block
Chain address;
6th writing module is configured as the third block chain address being written in the intelligent contract, the third block
Chain address obtains the target data analysis model using node for model, wherein the model is block chain using node
Node in network.
23. device according to claim 22, which is characterized in that the 5th writing module includes:
Second encryption submodule is configured as analyzing the target data using public key based on preset homomorphic encryption algorithm
Model is encrypted;
Second write-in submodule, is configured as encrypted target data analysis model being written in the block chain.
24. a kind of model training apparatus, which is characterized in that be applied to model optimization node, the model optimization node is block
Node in chain network, described device include:
Processor;
Memory for storage processor executable instruction;
Wherein, the processor is configured to:
The intelligent contract stored in block chain is called, obtains the first block chain address of Data Analysis Model, wherein described first
Block chain address is that model possesses node and obtains after the block chain is written in the Data Analysis Model and the intelligence is written
In energy contract, it is the node in the block chain network that the model, which possesses node,;
Using preset the first training dataset training Data Analysis Model, model optimization information is obtained;
The model optimization information is written in the block chain, the second block chain address is obtained;
By in the second block chain address write-in intelligent contract, the second block chain address possesses for the model
Node obtains the model optimization information.
25. a kind of model training apparatus, which is characterized in that possess node applied to model, it is block that the model, which possesses node,
Node in chain network, described device include:
Processor;
Memory for storage processor executable instruction;
Wherein, the processor is configured to:
Data Analysis Model is written in block chain, the first block chain address is obtained;
The first block chain address is written in the intelligent contract stored in the block chain, the first block chain address is used
It obtains the Data Analysis Model in model optimization node and is written in the intelligent contract and analyze mould for the data
Second block chain address of the model optimization information of type, wherein the model optimization node is the section in the block chain network
Point;
The intelligent contract is called, the second block chain address is obtained;
The block chain is inquired according to the second block chain address, obtains the model optimization information;
Using Data Analysis Model described in the model optimization information update, target data analysis model is obtained.
26. a kind of computer readable storage medium, is stored thereon with computer program instructions, which is characterized in that the program instruction
The step of any one of claim 8~10 the method is realized when being executed by processor.
27. a kind of computer readable storage medium, is stored thereon with computer program instructions, which is characterized in that the program instruction
The step of any one of claim 11~15 the method is realized when being executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811160666.8A CN109493216B (en) | 2018-09-30 | 2018-09-30 | Model training method, device, system and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811160666.8A CN109493216B (en) | 2018-09-30 | 2018-09-30 | Model training method, device, system and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109493216A true CN109493216A (en) | 2019-03-19 |
CN109493216B CN109493216B (en) | 2021-02-09 |
Family
ID=65690061
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811160666.8A Active CN109493216B (en) | 2018-09-30 | 2018-09-30 | Model training method, device, system and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109493216B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977014A (en) * | 2019-03-22 | 2019-07-05 | 泰康保险集团股份有限公司 | Code error recognition methods, device, equipment and storage medium based on block chain |
CN110503202A (en) * | 2019-08-22 | 2019-11-26 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
CN110535938A (en) * | 2019-08-29 | 2019-12-03 | 腾讯科技(深圳)有限公司 | A kind of data processing method based on intelligent contract, equipment and storage medium |
CN110532320A (en) * | 2019-08-01 | 2019-12-03 | 立旃(上海)科技有限公司 | Training data management method and device based on block chain |
CN110543776A (en) * | 2019-08-30 | 2019-12-06 | 联想(北京)有限公司 | model processing method, model processing device, electronic equipment and medium |
CN110825797A (en) * | 2019-10-25 | 2020-02-21 | 烨链(上海)科技有限公司 | Data exchange method and device |
CN110991622A (en) * | 2019-08-22 | 2020-04-10 | 腾讯科技(深圳)有限公司 | Machine learning model processing method based on block chain network and node |
CN111091888A (en) * | 2019-11-25 | 2020-05-01 | 泰康保险集团股份有限公司 | Activity decision method, device, equipment and storage medium |
CN111723147A (en) * | 2019-03-21 | 2020-09-29 | 杭州海康威视数字技术股份有限公司 | Block chain-based data training method, device and equipment and storage medium |
CN112166445A (en) * | 2019-04-16 | 2021-01-01 | 华为技术有限公司 | Joint learning method and joint learning equipment based on block chain network |
CN112418433A (en) * | 2020-10-30 | 2021-02-26 | 联想(北京)有限公司 | Model training method, device and equipment |
CN112989390A (en) * | 2021-04-15 | 2021-06-18 | 深圳前海移联科技有限公司 | Block chain homogeneous data sharing method based on pluggable homomorphic encryption |
CN113609508A (en) * | 2021-08-24 | 2021-11-05 | 上海点融信息科技有限责任公司 | Block chain-based federal learning method, device, equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101774844B1 (en) * | 2017-04-18 | 2017-09-19 | 주식회사 더디엔에이시스템 | Self-learning module of artificial neural network model using neuro-blockchain combination |
CN107864198A (en) * | 2017-11-07 | 2018-03-30 | 济南浪潮高新科技投资发展有限公司 | A kind of block chain common recognition method based on deep learning training mission |
CN108491266A (en) * | 2018-03-09 | 2018-09-04 | 联想(北京)有限公司 | Data processing method, device based on block chain and electronic equipment |
CN108595185A (en) * | 2018-04-11 | 2018-09-28 | 暨南大学 | A method of ether mill intelligence contract is converted into super account book intelligence contract |
-
2018
- 2018-09-30 CN CN201811160666.8A patent/CN109493216B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101774844B1 (en) * | 2017-04-18 | 2017-09-19 | 주식회사 더디엔에이시스템 | Self-learning module of artificial neural network model using neuro-blockchain combination |
CN107864198A (en) * | 2017-11-07 | 2018-03-30 | 济南浪潮高新科技投资发展有限公司 | A kind of block chain common recognition method based on deep learning training mission |
CN108491266A (en) * | 2018-03-09 | 2018-09-04 | 联想(北京)有限公司 | Data processing method, device based on block chain and electronic equipment |
CN108595185A (en) * | 2018-04-11 | 2018-09-28 | 暨南大学 | A method of ether mill intelligence contract is converted into super account book intelligence contract |
Non-Patent Citations (1)
Title |
---|
A. BESIR KURTULMUS: "Trustless Machine Learning Contracts; Evaluating and Exchanging", 《ARXIV:1802.10185》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111723147A (en) * | 2019-03-21 | 2020-09-29 | 杭州海康威视数字技术股份有限公司 | Block chain-based data training method, device and equipment and storage medium |
CN109977014A (en) * | 2019-03-22 | 2019-07-05 | 泰康保险集团股份有限公司 | Code error recognition methods, device, equipment and storage medium based on block chain |
CN112166445A (en) * | 2019-04-16 | 2021-01-01 | 华为技术有限公司 | Joint learning method and joint learning equipment based on block chain network |
CN110532320B (en) * | 2019-08-01 | 2023-06-27 | 立旃(上海)科技有限公司 | Training data management method and device based on block chain |
CN110532320A (en) * | 2019-08-01 | 2019-12-03 | 立旃(上海)科技有限公司 | Training data management method and device based on block chain |
CN110503202A (en) * | 2019-08-22 | 2019-11-26 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
CN110991622A (en) * | 2019-08-22 | 2020-04-10 | 腾讯科技(深圳)有限公司 | Machine learning model processing method based on block chain network and node |
CN110535938B (en) * | 2019-08-29 | 2021-07-27 | 腾讯科技(深圳)有限公司 | Data processing method, equipment and storage medium based on intelligent contract |
CN110535938A (en) * | 2019-08-29 | 2019-12-03 | 腾讯科技(深圳)有限公司 | A kind of data processing method based on intelligent contract, equipment and storage medium |
US11733991B2 (en) | 2019-08-29 | 2023-08-22 | Tencent Technology (Shenzhen) Company Limited | Data processing method based on intelligent contract, device, and storage medium |
CN110543776A (en) * | 2019-08-30 | 2019-12-06 | 联想(北京)有限公司 | model processing method, model processing device, electronic equipment and medium |
CN110825797A (en) * | 2019-10-25 | 2020-02-21 | 烨链(上海)科技有限公司 | Data exchange method and device |
CN111091888A (en) * | 2019-11-25 | 2020-05-01 | 泰康保险集团股份有限公司 | Activity decision method, device, equipment and storage medium |
CN111091888B (en) * | 2019-11-25 | 2023-10-31 | 泰康保险集团股份有限公司 | Activity decision method, device, equipment and storage medium |
CN112418433A (en) * | 2020-10-30 | 2021-02-26 | 联想(北京)有限公司 | Model training method, device and equipment |
CN112989390A (en) * | 2021-04-15 | 2021-06-18 | 深圳前海移联科技有限公司 | Block chain homogeneous data sharing method based on pluggable homomorphic encryption |
CN113609508A (en) * | 2021-08-24 | 2021-11-05 | 上海点融信息科技有限责任公司 | Block chain-based federal learning method, device, equipment and storage medium |
CN113609508B (en) * | 2021-08-24 | 2023-09-26 | 上海点融信息科技有限责任公司 | Federal learning method, device, equipment and storage medium based on blockchain |
Also Published As
Publication number | Publication date |
---|---|
CN109493216B (en) | 2021-02-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109493216A (en) | Model training method, device, system and storage medium | |
Conti et al. | Looking ahead in pervasive computing: Challenges and opportunities in the era of cyber–physical convergence | |
US10157324B2 (en) | Systems and methods of updating user identifiers in an image-sharing environment | |
CN106973425A (en) | A kind of method and apparatus for connecting WAP | |
CN107106030A (en) | The dynamic wearable device operating condition detected based on planning chart | |
CN107667369A (en) | The use of the mobile device of configuration lock | |
CN104301528B (en) | The method and device of display information | |
CN110222721B (en) | Data processing method, data processing device, block chain node and storage medium | |
CN109615734A (en) | It is a kind of for unlocking the method and apparatus of vehicle arrangement | |
CN110458102A (en) | A kind of facial image recognition method and device, electronic equipment and storage medium | |
CN107730847A (en) | Sharing articles finding method and device | |
CN103916473A (en) | Travel information processing method and relative device | |
CN106204029A (en) | The method and apparatus of Quick Response Code payment | |
CN110532956A (en) | Image processing method and device, electronic equipment and storage medium | |
CN107038428A (en) | Vivo identification method and device | |
Lanka et al. | A review of research on emerging technologies of the Internet of Things and augmented reality | |
CN107454204A (en) | User profile mask method and device | |
CN104050236B (en) | Web site contents update reminding method, server and client side | |
Liu et al. | Vi-Fi: Associating moving subjects across vision and wireless sensors | |
Misra et al. | Dependable machine intelligence at the tactical edge | |
CN106203436A (en) | Image binaryzation method and device | |
CN107885464A (en) | Date storage method, device and computer-readable recording medium | |
Akter et al. | IoTSAMS: A novel framework for Internet of Things (IoT) based smart attendance management system | |
CN105956633A (en) | Search engine category identification method and apparatus | |
Aljawarneh | A Secure Smart Meter Application Framework |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |