CN111680793A - Block chain consensus method and system based on deep learning model training - Google Patents

Block chain consensus method and system based on deep learning model training Download PDF

Info

Publication number
CN111680793A
CN111680793A CN202010318933.0A CN202010318933A CN111680793A CN 111680793 A CN111680793 A CN 111680793A CN 202010318933 A CN202010318933 A CN 202010318933A CN 111680793 A CN111680793 A CN 111680793A
Authority
CN
China
Prior art keywords
uri
information
deep learning
block
data
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
Application number
CN202010318933.0A
Other languages
Chinese (zh)
Other versions
CN111680793B (en
Inventor
李引
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Zhongke Yide Technology Co ltd
Original Assignee
Guangzhou Zhongke Yide Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangzhou Zhongke Yide Technology Co ltd filed Critical Guangzhou Zhongke Yide Technology Co ltd
Priority to CN202010318933.0A priority Critical patent/CN111680793B/en
Publication of CN111680793A publication Critical patent/CN111680793A/en
Application granted granted Critical
Publication of CN111680793B publication Critical patent/CN111680793B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06Q20/00—Payment architectures, schemes or protocols
    • G06Q20/38—Payment protocols; Details thereof
    • G06Q20/382—Payment protocols; Details thereof insuring higher security of transaction
    • G06Q20/3825—Use of electronic signatures
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The invention discloses a block chain consensus method and a system based on deep learning model training, wherein the system comprises a parameter and data acquirer, a consensus algorithm scheduler, a deep learning model trainer and a consensus verifier; can be butted with the existing block chain system to replace common recognition algorithms such as POW/POS/POA and the like. The invention introduces the excessive computing power of the block chain into the deep learning model training, and leads investors to use the mining machines for the artificial intelligence model training through the excitation mechanism of the block chain, thereby being capable of guiding capital, computing power and energy to be put into more meaningful work and solving the problems of insufficient computing power and high cost. The POW calculation power of the block chain is used for deep learning model training calculation of big data, so that the cost is reduced, social resources are saved, and the calculation power is used for meaningful work.

Description

Block chain consensus method and system based on deep learning model training
Technical Field
The invention relates to the technical field of block chains, in particular to a block chain consensus method and system based on deep learning model training.
Background
The development of the block chain technology is generally agreed by enterprises, research institutions, universities and the like at home and abroad, and is considered as the core of the next generation of valuable internet. The main stream block chain platform bit currency, Ethengfang, Laite currency and the like generally adopt a POW consensus algorithm, and determine the nodes of the block by repeatedly performing hash value operation by an ore machine, wherein the node has higher input calculation power and has higher probability to obtain block reward. This mechanism enables trust establishment and value consensus to be achieved through machines and algorithms, investors invest a lot of money to buy mining machines and power, the computing power of a whole bitcoin network has reached 70EH/S by 2019 and 9 months, the computing power and energy are used by investors to perform hash value operation so as to obtain rewards, and the computing power and energy waste in this mode are very serious and problematic.
Deep Learning (DL) is a new research direction in the field of Machine Learning (ML), which is introduced into Machine Learning to make it closer to the original goal, Artificial Intelligence (AI). Deep learning is the intrinsic law and expression level of the learning sample data, and the information obtained in the learning process is very helpful for the interpretation of data such as characters, images and sounds. The final aim of the method is to enable the machine to have the analysis and learning capability like a human, and to recognize data such as characters, images and sounds. Deep learning is a complex machine learning algorithm, and achieves the effect in speech and image recognition far exceeding the prior related art. According to one statistic, the computational power used in artificial intelligence training tasks has been increasing exponentially since 2012, at a current rate of doubling every 3.5 months (in contrast to moore's law doubling every 18 months). Since 2012, the need for computing power has increased by over 300,000 times (and only 12 times if at the rate of moore's law). In the meantime, the improvement of hardware computing power has been an important factor for the rapid development of artificial intelligence. Therefore, the computational power plays an important role in the deep learning model training.
The mainstream block chain generally adopts consensus algorithms such as POW (point of sale) and the like to determine block nodes, so that the problem of energy and computational complexity waste exists.
Disclosure of Invention
In view of this, in order to solve the problem that the consensus algorithm of the block chain in the prior art wastes energy and computational power, the invention provides a block chain consensus method and system based on deep learning model training.
The invention solves the problems through the following technical means:
in one aspect, the invention provides a block chain consensus method based on deep learning model training, which comprises the following steps:
the main body collects the file information and returns a uniform resource locator (URI) after storing the file information in a file systemDocumentThe principal is configured for data collection transactions, collecting the award amount, URI according to the format { principal wallet addressDocumentSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes;
the main body obtains the stored file information, marks the file information, generates a marked file, stores the marked file in a file system and returns a uniform resource locator (URI)LabelingThe principal is configured to mark the transaction with data in the format { principal wallet address, mark the award amount, URIDocument，URILabelingSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes; wherein the URILabelingThe corresponding file is a pair URIDocumentInformation for marking corresponding files;
calculating and verifying by adopting a deep learning model training consensus algorithm, verifying the effectiveness of the transaction by an accounting node, and putting the transaction into a cache pool until the number n of the data sets A in the cache pool reaches a threshold value alpha;
the accounting node acquires all data marking transactions stored in a chain and all data marking transactions in a node buffer pool by using a parameter and data acquirer, acquires files and marking information corresponding to URI (Uniform resource identifier) files and URI marks, simultaneously acquires parameter values stored in a previous block or a previous block, initializes the deep learning neural network by using the parameter values, starts supervised learning, and automatically adjusts the network structure and the parameters by adopting an AutoML method until the prediction accuracy of the model is greater than a threshold value beta set by a system;
the accounting node completes calculation of the model, stores model parameters to a block head, generates a first block coinage transaction for recording accounting rewards obtained by the node, packages the transactions in the buffer pool into block bodies, combines the block head and the block bodies to generate blocks, broadcasts the blocks in the whole network, and searches for verification signatures of other verification nodes;
the other verification nodes receive the information of the new block, and the common identification verifier verifies the new block; when the deep learning consensus model is adopted, the verification node acquires all data marking transactions stored on a chain by using a parameter and data acquirer, acquires a file T and marking information F corresponding to a URI file and a URI mark, acquires parameter values stored in the block, predicts the T after initializing the deep learning neural network by using the parameter values, compares the T with the F, and calculates the accuracy threshold value gamma of the model; if the requirement of the threshold gamma is met, the verification node signs the request and returns to the accounting node;
after the accounting node receives the signatures of the verification nodes with a certain threshold proportion, the accounting node sends the block information again and attaches the signatures of all the verification nodes;
and other nodes receive the information of the new block, verify the signature information in the new block and add the requirement into the block chain.
Further, before the calculation and verification by using the deep learning model training consensus algorithm, the method further comprises the following steps:
and selecting a consensus algorithm by a 'consensus algorithm scheduler' of the accounting node, if the consensus algorithm is trained by a deep learning model, adopting the deep learning model to train the consensus algorithm, and otherwise, adopting a traditional block chain consensus algorithm to calculate and verify.
Further, the file information includes pictures, voice, video, and text.
Further, the underlying blockchain, if based on the bitcoin, lexel blockchain, will extend the new data collection type, here "URIDocumentStoring by using reserved fields in bitcoin transaction; if the EtherFang, EOS based blockchain platform with intelligent contract mechanism, then "URIDocumentThe method takes the data information as a parameter and calls the method of the intelligent contract, when the data information passes the verification of the intelligent contract, the data information is transferred to the 'main purse address' through the intelligent contract,the amount is the "collection reward amount".
Further, the underlying blockchain, if based on the bitcoin, lexel blockchain, will extend the new data collection type, here "URIDocument+URILabelingStoring by using reserved fields in bitcoin transaction; if the EtherFang, EOS based blockchain platform with intelligent contract mechanism, then "URIDocument+URILabelingThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the intelligent contract transfers the money to the main wallet address, and the money is marked with the reward money.
Further, the thresholds α, β, γ are decided by intelligent contract means and voting by all node participants.
In another aspect, the present invention provides a block chain consensus system based on deep learning model training, including:
a parameter and data acquirer for the main body to collect the file information and return the uniform resource locator URI after storing it in the file systemDocumentThe principal is configured for data collection transactions, collecting the award amount, URI according to the format { principal wallet addressDocumentSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes; the main body obtains the stored file information, marks the file information, generates a marked file, stores the marked file in a file system and returns a uniform resource locator (URI)LabelingThe principal is configured to mark the transaction with data in the format { principal wallet address, mark the award amount, URIDocument，URILabelingSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes; wherein the URILabelingThe corresponding file is a pair URIDocumentInformation for marking corresponding files;
the deep learning model trainer is used for calculating and verifying by adopting a deep learning model training consensus algorithm, and the accounting node verifies the effectiveness of the transaction and puts the transaction into a cache pool until the number n of the data sets A in the cache pool reaches a threshold value alpha; the accounting node acquires all data marking transactions stored in a chain and all data marking transactions in a node buffer pool by using a parameter and data acquirer, acquires files and marking information corresponding to URI (Uniform resource identifier) files and URI marks, simultaneously acquires parameter values stored in a previous block or a previous block, initializes the deep learning neural network by using the parameter values, starts supervised learning, and automatically adjusts the network structure and the parameters by adopting an AutoML method until the prediction accuracy of the model is greater than a threshold value beta set by a system; the accounting node completes calculation of the model, stores model parameters to a block head, generates a first block coinage transaction for recording accounting rewards obtained by the node, packages the transactions in the buffer pool into block bodies, combines the block head and the block bodies to generate blocks, broadcasts the blocks in the whole network, and searches for verification signatures of other verification nodes;
the consensus verifier is used for verifying the new block by the other verification nodes after receiving the information of the new block; when the deep learning consensus model is adopted, the verification node acquires all data marking transactions stored on a chain by using a parameter and data acquirer, acquires a file T and marking information F corresponding to a URI file and a URI mark, acquires parameter values stored in the block, predicts the T after initializing the deep learning neural network by using the parameter values, compares the T with the F, and calculates the accuracy threshold value gamma of the model; if the requirement of the threshold gamma is met, the verification node signs the request and returns to the accounting node; after the accounting node receives the signatures of the verification nodes with a certain threshold proportion, the accounting node sends the block information again and attaches the signatures of all the verification nodes; and other nodes receive the information of the new block, verify the signature information in the new block and add the requirement into the block chain.
Further, the block chain consensus system trained based on the deep learning model further includes:
and the consensus algorithm scheduler is used for selecting a consensus algorithm for the bookkeeping nodes, if the consensus algorithm is trained by the deep learning model, the computation and verification are performed by adopting the deep learning model training consensus algorithm, and if the consensus algorithm is not trained by the deep learning model, the computation and verification are performed by adopting the traditional block chain consensus algorithm.
Further, the underlying blockchain, if based on the bitcoin, lexel blockchain, will extend the new data collection type, here "URIDocumentStoring by using reserved fields in bitcoin transaction; if the EtherFang, EOS based blockchain platform with intelligent contract mechanism, then "URIDocumentThe method takes the data information as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the data information is transferred to a main wallet address through the intelligent contract, and the amount is the acquired reward amount;
underlying Block chain if based on Bingonian, Ledijian Block chains, a new data acquisition type, here "URI" will be extendedDocument+URILabelingStoring by using reserved fields in bitcoin transaction; if the EtherFang, EOS based blockchain platform with intelligent contract mechanism, then "URIDocument+URILabelingThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the intelligent contract transfers the money to the main wallet address, and the money is marked with the reward money.
Further, the file information includes pictures, voice, video and text; the thresholds α, β, γ are determined by intelligent contractual means and by voting by all node participants.
Compared with the prior art, the invention has the beneficial effects that at least:
the invention provides a method for carrying out deep learning model training on the data to be used as block chain consensus achievement aiming at big data such as pictures, videos, voice, texts and the like, and can apply the calculation power to valuable model mining.
The invention introduces the excessive computing power of the block chain into the deep learning model training, and leads investors to use the mining machines for the artificial intelligence model training through the excitation mechanism of the block chain, thereby being capable of guiding capital, computing power and energy to be put into more meaningful work and solving the problems of insufficient computing power and high cost.
The invention uses the POW calculation power of the block chain to carry out deep learning model training calculation of big data, reduces the cost, saves social resources and uses the calculation power to meaningful work.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a block chain consensus method based on deep learning model training according to the present invention;
FIG. 2 is a schematic structural diagram of the block chain consensus system based on deep learning model training according to the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below. It should be noted that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and all other embodiments obtained by those skilled in the art without any inventive work based on the embodiments of the present invention belong to the protection scope of the present invention.
Example 1
As shown in fig. 1, the present invention provides a block chain consensus method based on deep learning model training, which includes the following steps:
step S110: the main body collects file information such as pictures, voice, video, text and the like, stores the file information in a file system and returns a uniform resource locator (URI)DocumentThe principal is configured for data collection transactions, collecting the award amount, URI according to the format { principal wallet addressDocumentIt sends it to the node, which broadcasts the information to the neighboring nodes.
Bottom layer block chain if based on bitcoin, lexeme and other block chains, then new expansion will be madeThe type of data collection, here "URIDocument"may be stored using reserved fields in bitcoin transactions; if the blockchain platform with intelligent contract mechanism is based on EtherFang, EOS, etc.' URIDocumentThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the account is transferred to the 'main wallet address' through the intelligent contract, and the amount is the 'collection reward amount'.
Step S120: the main body acquires file information such as stored pictures, voice, videos, texts and the like, marks the file information, generates a marked file, stores the marked file in a file system, and returns a uniform resource locator (URI)LabelingThe principal is configured to mark the transaction with data in the format { principal wallet address, mark the award amount, URIDocument，URILabelingIt sends it to the node, which broadcasts the information to the neighboring nodes. Wherein the URILabelingThe corresponding file is a pair URIDocumentAnd labeling the corresponding files such as pictures, voice, videos, texts and the like.
Underlying blockchains if based on bitcoin, leybu, etc. blockchains, then a new data collection type, here "URI", will be extendedDocument+URILabeling"may be stored using reserved fields in bitcoin transactions; if the blockchain platform with intelligent contract mechanism is based on EtherFang, EOS, etc.' URIDocument+URILabelingThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the intelligent contract transfers the money to the main wallet address, and the money is marked with the reward money.
The data source may not be from data information stored on the chain, but a training set and a testing set of data are designated by a task issuing mode to carry out training and verification of the model.
Step S130: and (3) selecting a consensus algorithm by a 'consensus algorithm scheduler' of the accounting node, jumping to the step S140 if the consensus algorithm is trained for the deep learning model, and otherwise, calculating and verifying by adopting a traditional block chain consensus algorithm.
Step S140: the accounting node verifies the validity of the transaction and puts the transaction into the buffer pool until the number n of the data sets A in the buffer pool reaches a threshold value alpha.
Step S150: the accounting node acquires all data marking transactions stored in a chain and all data marking transactions in a node buffer pool by using a parameter and data acquirer, acquires files and marking information corresponding to URI files and URI marks, simultaneously acquires parameter values stored in a previous block or a previous block, initializes the deep learning neural network by using the parameter values, starts supervised learning, and automatically adjusts the network structure and the parameters by using an automatic ML (automated Machine learning) method until the prediction accuracy of the model is greater than a threshold value beta set by the system.
Step S160: and the accounting node completes the calculation of the model, stores the model parameters into a block head, generates a first block coinage transaction for recording the node to obtain accounting rewards, packages the transactions in the buffer pool into block bodies, combines the block head and the block bodies to generate blocks, broadcasts the blocks in the whole network, and searches for verification signatures of other verification nodes.
Step S170: the other verification nodes receive the information of the new block, and the consensus verifier verifies the new block. When the deep learning consensus model is adopted, the verification node acquires all data marking transactions stored on a chain by using a parameter and data acquirer, acquires a file T and marking information F corresponding to the URI file and the URI marking, acquires parameter values stored in the block, predicts the T after initializing the deep learning neural network by using the parameter values, compares the T with the F, and calculates a series of threshold values gamma such as accuracy of the model. If the threshold gamma requirement is met, the verification node signs it and returns the accounting node.
And step S180, after the accounting node receives the signatures of the verification nodes with a certain threshold proportion, the accounting node sends the block information again and attaches the signatures of all the verification nodes.
And step S190, after the other nodes receive the information of the new block and verify that the signature information in the new block passes, adding the requirement into the block chain.
The thresholds α, β, γ, etc. may be decided by intelligent contracts, etc. and voted on by all node participants.
Example 2
As shown in fig. 2, the present invention further provides a block chain consensus system based on deep learning model training, which includes a parameter and data acquirer, a consensus algorithm scheduler, a deep learning model trainer, and a consensus verifier.
A parameter and data acquirer for main body to collect file information such as picture, voice, video, text, etc. and to return uniform resource locator URIDocumentThe principal is configured for data collection transactions, collecting the award amount, URI according to the format { principal wallet addressDocumentIt sends it to the node, which broadcasts the information to the neighboring nodes.
Underlying blockchains if based on bitcoin, leybu, etc. blockchains, then a new data collection type, here "URI", will be extendedDocument"may be stored using reserved fields in bitcoin transactions; if the blockchain platform with intelligent contract mechanism is based on EtherFang, EOS, etc.' URIDocumentThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the account is transferred to the 'main wallet address' through the intelligent contract, and the amount is the 'collection reward amount'.
The main body acquires file information such as stored pictures, voice, videos, texts and the like, marks the file information, generates a marked file, stores the marked file in a file system, and returns a uniform resource locator (URI)LabelingThe principal is configured to mark the transaction with data in the format { principal wallet address, mark the award amount, URIDocument，URILabelingIt sends it to the node, which broadcasts the information to the neighboring nodes. Wherein the URILabelingThe corresponding file is a pair URIDocumentAnd labeling the corresponding files such as pictures, voice, videos, texts and the like.
The bottom layer block chain is based on Bite and Lete coinsWait for blockchains, then a new data acquisition type, here "URI", will be extendedDocument+URILabeling"may be stored using reserved fields in bitcoin transactions; if the blockchain platform with intelligent contract mechanism is based on EtherFang, EOS, etc.' URIDocument+URILabelingThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the intelligent contract transfers the money to the main wallet address, and the money is marked with the reward money.
The data source may not be from data information stored on the chain, but a training set and a testing set of data are designated by a task issuing mode to carry out training and verification of the model.
And the consensus algorithm scheduler is used for selecting a consensus algorithm for the bookkeeping nodes, if the consensus algorithm is trained by the deep learning model, the computation and verification are performed by adopting the deep learning model training consensus algorithm, and if the consensus algorithm is not trained by the deep learning model, the computation and verification are performed by adopting the traditional block chain consensus algorithm.
And the deep learning model trainer is used for verifying the effectiveness of the transaction by the accounting node and putting the transaction into the buffer pool until the number n of the data sets A in the buffer pool reaches a threshold value alpha.
The accounting node acquires all data marking transactions stored in a chain and all data marking transactions in a node buffer pool by using a parameter and data acquirer, acquires files and marking information corresponding to URI files and URI marks, simultaneously acquires parameter values stored in a previous block or a previous block, initializes the deep learning neural network by using the parameter values, starts supervised learning, and automatically adjusts the network structure and the parameters by using an automatic ML (automated Machine learning) method until the prediction accuracy of the model is greater than a threshold value beta set by the system.
And the accounting node completes the calculation of the model, stores the model parameters into a block head, generates a first block coinage transaction for recording the node to obtain accounting rewards, packages the transactions in the buffer pool into block bodies, combines the block head and the block bodies to generate blocks, broadcasts the blocks in the whole network, and searches for verification signatures of other verification nodes.
And the consensus verifier is used for verifying the new block by the other verification nodes after receiving the information of the new block. When the deep learning consensus model is adopted, the verification node acquires all data marking transactions stored on a chain by using a parameter and data acquirer, acquires a file T and marking information F corresponding to the URI file and the URI marking, acquires parameter values stored in the block, predicts the T after initializing the deep learning neural network by using the parameter values, compares the T with the F, and calculates a series of threshold values gamma such as accuracy of the model. If the threshold gamma requirement is met, the verification node signs it and returns the accounting node.
And after the accounting node receives the signatures of the verification nodes with a certain threshold proportion, the accounting node sends the block information again and attaches the signatures of all the verification nodes.
And other nodes receive the information of the new block, verify the signature information in the new block and add the requirement into the block chain.
The thresholds α, β, γ, etc. may be decided by intelligent contracts, etc. and voted on by all node participants.
The invention provides a method and a system for achieving block chain node consensus by deep learning model training, which are divided into a consensus algorithm scheduler, a parameter and data acquirer, a deep learning model trainer and a consensus verifier. The system can be butted with the existing block chain system to replace common recognition algorithms such as POW/POS/POA and the like.
The invention provides a method for carrying out deep learning model training on the data to be used as block chain consensus achievement aiming at big data such as pictures, videos, voice, texts and the like, and can apply the calculation power to valuable model mining.
The invention introduces the excessive computing power of the block chain into the deep learning model training, and leads investors to use the mining machines for the artificial intelligence model training through the excitation mechanism of the block chain, thereby being capable of guiding capital, computing power and energy to be put into more meaningful work and solving the problems of insufficient computing power and high cost.
The invention uses the POW calculation power of the block chain to carry out deep learning model training calculation of big data, reduces the cost, saves social resources and uses the calculation power to meaningful work.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (10)

1. A block chain consensus method based on deep learning model training is characterized by comprising the following steps:
the main body collects the file information and returns a uniform resource locator (URI) after storing the file information in a file systemDocumentThe principal is configured for data collection transactions, collecting the award amount, URI according to the format { principal wallet addressDocumentSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes;
the main body obtains the stored file information, marks the file information, generates a marked file, stores the marked file in a file system and returns a uniform resource locator (URI)LabelingThe principal is configured to mark the transaction with data in the format { principal wallet address, mark the award amount, URIDocument，URILabelingSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes; wherein the URILabelingThe corresponding file is a pair URIDocumentInformation for marking corresponding files;
calculating and verifying by adopting a deep learning model training consensus algorithm, verifying the effectiveness of the transaction by an accounting node, and putting the transaction into a cache pool until the number n of the data sets A in the cache pool reaches a threshold value alpha;
the accounting node acquires all data label transactions stored in a chain and all data label transactions in a node buffer pool by using a parameter and data acquirer, acquires files and label information corresponding to the URI file and the URI label, simultaneously acquires parameter values stored in a previous block or a previous block, initializes a deep learning neural network by using the parameter values, starts supervised learning, and automatically adjusts a network structure and parameters by using an AutoML method until the prediction accuracy of a model is greater than a threshold value beta set by a system;
the accounting node completes calculation of the model, stores model parameters to a block head, generates a first block coinage transaction for recording accounting rewards obtained by the node, packages the transactions in the buffer pool into block bodies, combines the block head and the block bodies to generate blocks, broadcasts the blocks in the whole network, and searches for verification signatures of other verification nodes;
the other verification nodes receive the information of the new block, and the consensus verifier verifies the new block; when the deep learning consensus model is adopted, the verification node acquires all data marking transactions stored on a chain by using a parameter and data acquirer, acquires a file T and marking information F corresponding to the URI file and the URI marking, acquires parameter values stored in the block, predicts the T after initializing the deep learning neural network by using the parameter values, compares the T with the F, and calculates the accuracy threshold gamma of the model; if the requirement of the threshold gamma is met, the verification node signs the request and returns to the accounting node;
after the accounting node receives the signatures of the verification nodes with a certain threshold proportion, the accounting node sends the block information again and attaches the signatures of all the verification nodes;
and other nodes receive the information of the new block, verify the signature information in the new block and add the requirement into the block chain.
2. The block chain consensus method based on deep learning model training of claim 1, wherein before the calculating and verifying by using the deep learning model training consensus algorithm, further comprising:
and selecting a consensus algorithm by a consensus algorithm scheduler of the accounting node, if the consensus algorithm is trained by a deep learning model, adopting the deep learning model to train the consensus algorithm, and otherwise, adopting a traditional block chain consensus algorithm to calculate and verify.
3. The block chain consensus method based on deep learning model training of claim 1, wherein the document information comprises pictures, voice, video, and text.
4. The block chain consensus method based on deep learning model training as claimed in claim 1, wherein the underlying block chain, if based on Bitty, Letty block chains, will extend new data collection type, here URIDocumentThe reserved field in the bitcoin transaction is adopted for storage; URI if EtherFang, EOS based blockchain platform with intelligent contract mechanismDocumentThe method is used as a parameter and calls a method of an intelligent contract, when the data information passes the verification of the intelligent contract, the data information is transferred to the address of a main wallet through the intelligent contract, and the amount is the acquired reward amount.
5. The block chain consensus method based on deep learning model training as claimed in claim 1, wherein the underlying block chain, if based on Bitty, Letty block chains, will extend new data collection type, here "URI"Document+URILabelingStoring by using reserved fields in bitcoin transaction; if the EtherFang, EOS based blockchain platform with intelligent contract mechanism, then "URIDocument+URILabelingThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the data information is transferred to the address of the main wallet through the intelligent contract, and the amount is the marked reward amount.
6. The block chain consensus method based on deep learning model training of claim 1, wherein the thresholds α, β, γ are decided by intelligent contract means and voted by all node participants.
7. A block chain consensus system based on deep learning model training, comprising:
a parameter and data acquirer for the main body to collect the file information and return the uniform resource locator URI after storing it in the file systemDocumentThe principal is configured for data collection transactions, collecting the award amount, URI according to the format { principal wallet addressDocumentSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes; the main body obtains the stored file information, marks the file information, generates a marked file, stores the marked file in a file system and returns a uniform resource locator (URI)LabelingThe principal is configured to mark the transaction with data in the format { principal wallet address, mark the award amount, URIDocument，URILabelingSending the information to the nodes, and broadcasting the information to the adjacent nodes by the nodes; wherein the URILabelingThe corresponding file is a pair URIDocumentInformation for marking corresponding files;
the deep learning model trainer is used for calculating and verifying by adopting a deep learning model training consensus algorithm, and the accounting node verifies the effectiveness of the transaction and puts the transaction into a cache pool until the number n of the data sets A in the cache pool reaches a threshold value alpha; the accounting node acquires all data label transactions stored in a chain and all data label transactions in a node buffer pool by using a parameter and data acquirer, acquires files and label information corresponding to the URI file and the URI label, simultaneously acquires parameter values stored in a previous block or a previous block, initializes a deep learning neural network by using the parameter values, starts supervised learning, and automatically adjusts a network structure and parameters by using an AutoML method until the prediction accuracy of a model is greater than a threshold value beta set by a system; the accounting node completes calculation of the model, stores model parameters to a block head, generates a first block coinage transaction for recording accounting rewards obtained by the node, packages the transactions in the buffer pool into block bodies, combines the block head and the block bodies to generate blocks, broadcasts the blocks in the whole network, and searches for verification signatures of other verification nodes;
the consensus verifier is used for verifying the new block by the other verification nodes after receiving the information of the new block; when the deep learning consensus model is adopted, the verification node acquires all data marking transactions stored on a chain by using a parameter and data acquirer, acquires a file T and marking information F corresponding to the URI file and the URI marking, acquires parameter values stored in the block, predicts the T after initializing the deep learning neural network by using the parameter values, compares the T with the F, and calculates the accuracy threshold gamma of the model; if the requirement of the threshold gamma is met, the verification node signs the request and returns to the accounting node; after the accounting node receives the signatures of the verification nodes with a certain threshold proportion, the accounting node sends the block information again and attaches the signatures of all the verification nodes; and other nodes receive the information of the new block, verify the signature information in the new block and add the requirement into the block chain.
8. The deep learning model training-based blockchain consensus system of claim 7, wherein the deep learning model training-based blockchain consensus system further comprises:
and the consensus algorithm scheduler is used for selecting a consensus algorithm for the bookkeeping nodes, if the consensus algorithm is trained by the deep learning model, the computation and verification are performed by adopting the deep learning model training consensus algorithm, and if the consensus algorithm is not trained by the deep learning model, the computation and verification are performed by adopting the traditional block chain consensus algorithm.
9. The deep learning model training-based blockchain consensus system as claimed in claim 7, wherein the underlying blockchain extends new data collection types if based on bitcoin, leyburn blockchains, where URI isDocumentThe reserved field in the bitcoin transaction is adopted for storage; URI if EtherFang, EOS based blockchain platform with intelligent contract mechanismDocumentThe method is used as a parameter and calls a method of the intelligent contract, and after the data information passes the verification of the intelligent contract, the data information passes the intelligent contractTransferring accounts to the address of the main wallet, wherein the amount is the acquired reward amount;
underlying Block chain if based on Bingonian, Ledijian Block chains, a new data acquisition type, here "URI" will be extendedDocument+URILabelingStoring by using reserved fields in bitcoin transaction; if the EtherFang, EOS based blockchain platform with intelligent contract mechanism, then "URIDocument+URILabelingThe method is used as a parameter and calls a method of the intelligent contract, when the data information passes the verification of the intelligent contract, the data information is transferred to the address of the main wallet through the intelligent contract, and the amount is the marked reward amount.
10. The deep learning model training-based blockchain consensus system of claim 7, wherein the file information comprises pictures, voice, video, and text; the thresholds α, β, γ are determined by intelligent contractual means and by voting by all node participants.
CN202010318933.0A 2020-04-21 2020-04-21 Block chain consensus method and system based on deep learning model training Active CN111680793B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010318933.0A CN111680793B (en) 2020-04-21 2020-04-21 Block chain consensus method and system based on deep learning model training

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010318933.0A CN111680793B (en) 2020-04-21 2020-04-21 Block chain consensus method and system based on deep learning model training

Publications (2)

Publication Number Publication Date
CN111680793A true CN111680793A (en) 2020-09-18
CN111680793B CN111680793B (en) 2023-06-09

Family

ID=72451667

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010318933.0A Active CN111680793B (en) 2020-04-21 2020-04-21 Block chain consensus method and system based on deep learning model training

Country Status (1)

Country Link
CN (1) CN111680793B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112364388A (en) * 2020-10-28 2021-02-12 中车工业研究院有限公司 Sensor data authentication method and device based on block chain
CN112418335A (en) * 2020-11-27 2021-02-26 北京云聚智慧科技有限公司 Model training method based on continuous image frame tracking and labeling and electronic equipment
CN112487103A (en) * 2020-12-25 2021-03-12 昆明理工大学 Trusted deep learning data set sharing system based on intelligent contract of block chain
CN112819072A (en) * 2021-02-01 2021-05-18 西南民族大学 Supervised classification method and system
CN112967148A (en) * 2021-03-30 2021-06-15 电子科技大学长三角研究院(衢州) Novel block chain consensus mechanism for intelligent Internet of things computing power service
CN113254980A (en) * 2021-07-07 2021-08-13 南京可信区块链与算法经济研究院有限公司 Workload certification consensus method and system for deep learning
CN113837300A (en) * 2021-09-29 2021-12-24 上海海事大学 Automatic driving cross-domain target detection method based on block chain
CN114282586A (en) * 2020-09-27 2022-04-05 中兴通讯股份有限公司 Data annotation method, system and electronic equipment

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
US20190012595A1 (en) * 2017-07-07 2019-01-10 Pointr Data, Inc. Neural network consensus using blockchain
CN109543726A (en) * 2018-11-06 2019-03-29 联动优势科技有限公司 A kind of method and device of training pattern
CN110197285A (en) * 2019-05-07 2019-09-03 清华大学 Security cooperation deep learning method and device based on block chain
CN110535836A (en) * 2019-08-12 2019-12-03 安徽师范大学 A kind of trust block chain common recognition method of based role classification

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190012595A1 (en) * 2017-07-07 2019-01-10 Pointr Data, Inc. Neural network consensus using blockchain
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
CN109543726A (en) * 2018-11-06 2019-03-29 联动优势科技有限公司 A kind of method and device of training pattern
CN110197285A (en) * 2019-05-07 2019-09-03 清华大学 Security cooperation deep learning method and device based on block chain
CN110535836A (en) * 2019-08-12 2019-12-03 安徽师范大学 A kind of trust block chain common recognition method of based role classification

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
MENG SHEN等: "Privacy-Preserving Support Vector Machine Training Over Blockchain-Based Encrypted IoT Data in Smart Cities" *
方俊杰;雷凯;: "面向边缘人工智能计算的区块链技术综述" *

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114282586A (en) * 2020-09-27 2022-04-05 中兴通讯股份有限公司 Data annotation method, system and electronic equipment
CN112364388A (en) * 2020-10-28 2021-02-12 中车工业研究院有限公司 Sensor data authentication method and device based on block chain
CN112418335A (en) * 2020-11-27 2021-02-26 北京云聚智慧科技有限公司 Model training method based on continuous image frame tracking and labeling and electronic equipment
CN112418335B (en) * 2020-11-27 2024-04-05 北京云聚智慧科技有限公司 Model training method based on continuous image frame tracking annotation and electronic equipment
CN112487103A (en) * 2020-12-25 2021-03-12 昆明理工大学 Trusted deep learning data set sharing system based on intelligent contract of block chain
CN112487103B (en) * 2020-12-25 2023-06-06 昆明理工大学 Deep learning data set trusted sharing system based on blockchain intelligent contract
CN112819072A (en) * 2021-02-01 2021-05-18 西南民族大学 Supervised classification method and system
CN112819072B (en) * 2021-02-01 2023-07-18 西南民族大学 Supervision type classification method and system
CN112967148B (en) * 2021-03-30 2024-02-20 电子科技大学长三角研究院(衢州) Block chain consensus mechanism for intelligent Internet of things computing service
CN112967148A (en) * 2021-03-30 2021-06-15 电子科技大学长三角研究院(衢州) Novel block chain consensus mechanism for intelligent Internet of things computing power service
CN113254980B (en) * 2021-07-07 2022-02-15 南京可信区块链与算法经济研究院有限公司 Workload certification consensus method and system for deep learning
CN113254980A (en) * 2021-07-07 2021-08-13 南京可信区块链与算法经济研究院有限公司 Workload certification consensus method and system for deep learning
CN113837300A (en) * 2021-09-29 2021-12-24 上海海事大学 Automatic driving cross-domain target detection method based on block chain
CN113837300B (en) * 2021-09-29 2024-03-12 上海海事大学 Automatic driving cross-domain target detection method based on block chain

Also Published As

Publication number Publication date
CN111680793B (en) 2023-06-09

Similar Documents

Publication Publication Date Title
CN111680793A (en) Block chain consensus method and system based on deep learning model training
CN110751261A (en) Training method and system and prediction method and system of neural network model
CN112613501A (en) Information auditing classification model construction method and information auditing method
CN110751287B (en) Training method and system and prediction method and system for neural network model
CN110889291B (en) Report evaluation method based on deep learning and blockchain
CN113014566B (en) Malicious registration detection method and device, computer readable medium and electronic device
CN111681091A (en) Financial risk prediction method and device based on time domain information and storage medium
CN110751285B (en) Training method and system and prediction method and system for neural network model
CN106067019A (en) The method and device of Text region is carried out for image
CN111506785B (en) Social text-based network public opinion topic identification method and system
CN111680098A (en) Block chain system for data acquisition, data annotation, AI model training and verification
CN109948735A (en) A kind of multi-tag classification method, system, device and storage medium
CN112257709A (en) Signboard photo auditing method and device, electronic equipment and readable storage medium
CN115392237B (en) Emotion analysis model training method, device, equipment and storage medium
CN114626102A (en) Block chain-based electronic certificate transfer method, device, equipment and storage medium
CN115687643A (en) Method for training multi-mode information extraction model and information extraction method
Kharchevnikova et al. Efficient video face recognition based on frame selection and quality assessment
CN111061948A (en) User label recommendation method and device, computer equipment and storage medium
CN110858326A (en) Method, device, equipment and medium for model training and acquiring additional characteristic data
CN111353689B (en) Risk assessment method and device
CN117290508A (en) Post-loan text data processing method and system based on natural language processing
CN111680076A (en) Block chain consensus method and system based on association rule model training
US20190266500A1 (en) Method and system for processing multi-request applications
CN113590786A (en) Data prediction method, device, equipment and storage medium
CN111680099A (en) Block chain consensus method and system based on decision tree model training

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