CN108805282A - Deep learning data sharing method, storage medium based on block chain mode - Google Patents
Deep learning data sharing method, storage medium based on block chain mode Download PDFInfo
- Publication number
- CN108805282A CN108805282A CN201810397795.2A CN201810397795A CN108805282A CN 108805282 A CN108805282 A CN 108805282A CN 201810397795 A CN201810397795 A CN 201810397795A CN 108805282 A CN108805282 A CN 108805282A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- data
- learning data
- block chain
- metadata
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides a kind of deep learning data sharing method, storage medium based on block chain mode, and method includes:Storage depth learning data and Target Modeling model to block chain network each node, wherein in block chain network, the Target Modeling model is protected for the deep learning data sharing;Deep learning data are split into the metadata with corresponding data attribute tags;When receiving the deep learning request of a corresponding Target Modeling model, determines corresponding data attribute label according to the Target Modeling model, obtain corresponding metadata;According to the corresponding metadata, extracts deep learning data corresponding with the Target Modeling model and learnt.The present invention not only realizes the mode for optimizing deep learning, but also can also realize settling at one go for enhancing study, significantly improves the efficiency of study;Further, the multiplexing of learning data and the saving of learning cost and time are realized.
Description
Technical field
The present invention relates to deep learning data fields, the deep learning data sharing based on block chain mode is particularly related to
Method, storage medium.
Background technology
Machine deep learning scheme common at present is the general modeling algorithm that Google DeepMind team uses.It adopts
With the deep learning model of convolutional neural networks, several robots are connected with each other, the shared experience mutually learnt and trial number
According to, after nearly million actions of experience, the machine of networking can gradually start to correct it is whole I, achieve the effect that self-teaching.That
These data can as this mode of learning experimental data it is single store, and other action learnings can not be used as
Support data, that is, cannot achieve the multiplexing of data.
Simultaneously as machine is independently to be learnt, study can all retain mass data every time, and these data make
It is unique in permission and range.Due to handled problem and think that demand to be achieved is one between several machines
It causes, so data can be directly invoked without processing.It is used but although problem and demand are consistent
Enhancing learning algorithm may be inconsistent.In revenue function, by research strategy, research revenue function, research based on model
The algorithm research of enhancing three kinds of different angles of study, different machines learn the required time and cost is centainly different.
Further, existing machine deep learning scheme, each study are required for from the beginning, even if wherein very much
It has all grasped part, it is also desirable to it relearns, cannot achieve the targetedly corresponding sample data of direct learning objective model, from
And cause the waste of unnecessary time and cost.
In conclusion it is necessary to provide a kind of deep learning data sharing method that can overcome the above problem and a kind of meters
Calculation machine storage medium.
Invention content
The technical problem to be solved by the present invention is to:A kind of deep learning data sharing side based on block chain mode is provided
Method, storage medium can have targetedly only learning objective learning data, significantly improve the efficiency of deep learning, simultaneously also
It can realize the reusable of learning data.
In order to solve the above-mentioned technical problem, the technical solution adopted by the present invention is:
A kind of deep learning data sharing method based on block chain mode, including:
Storage deep learning data and Target Modeling model to block chain network each node, wherein the deep learning
In block chain network, the Target Modeling model is protected for data sharing;
Deep learning data are split into the metadata with corresponding data attribute tags;
When receiving the deep learning request of a corresponding Target Modeling model, determined according to the Target Modeling model
Corresponding data attribute label, obtains corresponding metadata;
According to the corresponding metadata, deep learning data corresponding with the Target Modeling model are extracted
It practises.
Another technical solution provided by the invention is:
A kind of computer storage media, is stored thereon with computer program, can be realized when which is executed by processor
The above-mentioned deep learning data sharing method based on block chain mode.
The beneficial effects of the present invention are:Using the open characteristic storage deep learning data of block chain, depth is realized
Practise the transparent shared and reusable of data;Characteristic is weighed really using block chain and stores privately owned Target Modeling model, ensures target
The safety of modeler model;By the way that deep learning data are carried out granulating dismantling into metadata, each metadata is with correspondence
Data attribute label;In study, corresponding data attribute is determined according to Target Modeling model, then direct basis mark
Label quote metadata, and the learning data for extracting corresponding model is learnt, without all being carried out to all learning datas
Study causes resource and time upper unnecessary waste.The present invention quotes accurate and quick acquisition target by label and learns
Data can have and targetedly be learnt, realize settling at one go for enhancing study, significantly improve the effect of deep learning
Rate.
Description of the drawings
Fig. 1 is a kind of flow diagram of the deep learning data sharing method based on block chain mode of the present invention;
Fig. 2 be the embodiment of the present invention one the deep learning data sharing method based on block chain mode framework composition and
Interaction schematic diagram.
Specific implementation mode
To explain the technical content, the achieved purpose and the effect of the present invention in detail, below in conjunction with embodiment and coordinate attached
Figure is explained.
The design of most critical of the present invention is:By disassembling deep learning data at corresponding data attribute tags
Metadata determines corresponding data attribute, then direct basis label quotes member in study according to Target Modeling model
Data are extracted learning data corresponding with model and are learnt.
Explanation of technical terms of the present invention:
Technical term | It explains |
Deep learning data | Sample data as study |
Target Modeling model | Need the object module reached |
Metadata | The data with label are split out from sample data |
Data attribute | Including essential attributes such as pigment value, type of action, time, authors |
Fig. 1 is please referred to, the present invention provides a kind of deep learning data sharing method based on block chain mode, including:
Storage deep learning data and Target Modeling model to block chain network each node, wherein the deep learning
In block chain network, the Target Modeling model is protected for data sharing;
Deep learning data are split into the metadata with corresponding data attribute tags;
When receiving the deep learning request of a corresponding Target Modeling model, determined according to the Target Modeling model
Corresponding data attribute label, obtains corresponding metadata;
According to the corresponding metadata, deep learning data corresponding with the Target Modeling model are extracted
It practises.
As can be seen from the above description, the beneficial effects of the present invention are:By the way that data are really weighed, it is shared after, make deep learning
Data can disassemble into the metadata with data attribute label and be stored in the service of block chain network, reach granulating, this
Sample has a large amount of learning data that can be re-used in the new deep learning of machine, and when the model data and target of granulating
It, can direct learning success when modeler model is completely the same.The enhancing indoctrination session of machine settles at one go, need not prepare again big
The experimental data support study of amount, without all data are learnt from the beginning, there are no must repeat experience failure could complete to learn
It practises, significantly improves the efficiency studied in depth.
Further, described that deep learning data are split into the metadata with corresponding data attribute tags, specially:
Deep learning data are disassembled into multiple metadata;
Assign each metadata at least one label, the label record has a number of current meta data, and with work as
The corresponding characteristic value of a data attribute of preceding metadata.
Further, described to determine corresponding data attribute label according to the Target Modeling model, it obtains corresponding
Metadata, specially:
The characteristic value of corresponding at least one data attribute is determined according to the Target Modeling model;
Corresponding label is determined according to identified at least one characteristic value;
Corresponding metadata is obtained according to identified label.
Seen from the above description, deep learning data are split, and according to splitting obtained each metadata respectively
Corresponding data attribute is marked with tagged manner, realizes quoting based on label, using label as corresponding learning data
Index can fast and accurately position according to label and obtain corresponding learning data, realize needed for targetedly only obtaining
Learning data is learnt.
Further, described to be learnt, later, further include:
Generate new deep learning data;
The new deep learning data are stored to each node of block chain network;
The new deep learning data are split into the new metadata with corresponding data attribute tags.
Seen from the above description, after the completion of study, new a large amount of learning data will be generated, by equally being stored
In block chain node, and metadata is split into, as the data supporting of study next time, can not only constantly enriched and perfect
Deep learning database improves the value of learning database, and can also realize the multiplexing of data.
Further, further include:
Metadata is stored in the metadatabase of each node of block chain network.
Seen from the above description, separate storage metadata realizes the classified and stored of data in specific metadatabase, side
Just it quickly and accurately calls.
Further, the deep learning data are split according to the least unit of block chain data.
Seen from the above description, learning data is carried out to the fractionation of least unit, is realized maximum to learning data
Refinement, the further degree of simplifying for improving the learning data quoted according to label and accuracy.
Another technical solution provided by the invention is:
A kind of computer storage media, is stored thereon with computer program, can be realized when which is executed by processor
The above-mentioned deep learning data sharing method based on block chain mode.
As can be seen from the above description, the beneficial effects of the present invention are:It corresponds to one of ordinary skill in the art will appreciate that real
All or part of flow in existing above-mentioned technical proposal, is that relevant hardware can be instructed to realize by computer program
, the program can be stored in a computer-readable storage medium, and the program is when being executed, it may include such as above-mentioned each
The flow of method.
Embodiment one
Referring to Fig.1 and 2, the present embodiment provides a kind of deep learning data sharing methods based on block chain mode.
By block chain Techno-sharing, the data characteristic of true power, relevance, metadata can be extracted from information resources, for member
Data are tagged, these metadata can be as the data supporting of deep learning.
The method of the present embodiment includes the following steps:
S1:Build block chain network service platform;
Build a block chain network service platform, i.e. block chain network, the block chain network service platform it is each
Node stores identical data, and open use.
S2:Storage deep learning data and Target Modeling model to block chain network each node.
Specifically, usually Target Modeling model and deep learning data are relevant property, it just can guarantee that target is built in this way
The realization of mould model.The Target Modeling model, which is deep learning target, needs modeling systems library to be achieved model, that is, learns mesh
Mark data model;The deep learning data are to need a large amount of data to do the exercises in completing object procedure, by a large amount of
Data gradually complete last learning objective, this partial data is referred to as to the support data of result, i.e. Learning Support data.
Deep learning data and Target Modeling model are all stored to each node of block chain network, wherein pass through block
Deep learning data, are stored in each node of block chain by shared, the open and clear characteristic of chain network with open, sharing mode
On, and it is labeled as deep learning database;Weigh characteristic really by block chain network, by the model data of Target Modeling model with
Shielded mode is stored on each node of block chain, and carries out signature and hash cryptographic operations to data, is labeled as target
Model library.
Wherein, the true power refers to being confirmed as privately owned by perhaps block in this block, others can not check and use.
Since object module library needs the target completed, current each company all carrying out deep learning for my this deep learning, but
It is that everybody progress is different, the objective result of core is secrecy, only when achievement has been completed, can just announce technology,
So needing to maintain secrecy to the content that I is doing.
The deep learning database is to service the sample experimental data needed for deep learning by block chain network to believe
Breath is stored in the position of all nodes, and support of these data as deep learning is shared in the service of block chain network;It is described
Object module library data carry out signature and Encryption Algorithm utilized the anonymity of block chain network chain by privately owned protection
With extremely strong safety protection modeling database.
S3:All deep learning data are split into the metadata with corresponding data attribute tags;
Specifically, can be by disassembling all deep learning data in deep learning database at multiple first numbers
According to;The fineness of dismantling can be pre-configured, and realize the flexible control of dismantling precision.It is preferred that the particle according to block chain data is special
Property, the least unit of a data is disassembled for the precision of 10-8 powers, realizes that data refine to the greatest extent, improve most
The deep learning data obtained afterwards simplify degree.It is preferable to use Tika to utilize existing parsing class libraries, detects and extract member
Data.
Wherein, the particle characteristics refer to that the data of block chain can be granulated;Using bit coin as example, 1
Piece bit coin can split into -8 powers that least unit is 10.
Each metadata will be endowed at least one label, and each label record has the number of current meta data, with
And the characteristic value corresponding to a data attribute of current meta data.The metadata be both granulating learning data and
It is equivalent to index, the information of data attribute is recorded by label, for supporting the work(such as historical data, resource lookup, file record
Can, available support data can be quoted by metadata quickly.
The metadata is stored in the metadatabase of each node of block chain, and realization is targetedly called, while again can
Better classified and stored.
S4:When receiving the deep learning request of a corresponding Target Modeling model, according to the Target Modeling model
It determines corresponding data attribute label, obtains corresponding metadata;
When machine receives the deep learning request of a corresponding Target Modeling model, such as skip melt pit this target mould
Type is learnt.The Target Modeling model can be not present in advance in object module library, that is, belong to new object module,
But it is the need to ensure that in deep learning database that there are relative learning datas, just can guarantee can realize this by study
Object module.
Specific learning process is as follows:
According to the deep learning request received, determines and ask corresponding Target Modeling model corresponding at least one
The characteristic value of data attribute;It can specifically determine according to the model data of the Target Modeling model received and realize that the target is built
The characteristic value of the data attribute of learning data needed for mould model;Data attribute is usually multiple, thus may determine that going out multiple
Characteristic value;
Then, it determines corresponding label according to identified at least one characteristic value, then is obtained according to identified label
Take corresponding metadata.Metadatabase can be traversed by characteristic value, obtain label substance and matched with identified characteristic value
Metadata.
Subsequently, according to identified metadata, go out from the magnanimity deep learning extracting data of deep learning database
The corresponding deep learning data of the metadata, that is, the deep learning data for corresponding to the Target Modeling model are learnt.By
Although in metadata itself being that deep learning data are split, and imperfect, cannot be used directly for learning, therefore,
It needs to quote the corresponding deep learning data of corresponding with request Target Modeling data by identified metadata and be used as
The data supporting of deep learning.
That is, when starting deep learning, characteristic that is open and being cited is recorded according to block chain, when privately owned
Target Modeling database request call learning data when, in metadatabase by the label of definition quote needs data make
For the data supporting of deep learning.Described quote refers to that every transaction record can be seen certainly when block chain transaction authentication
With transfer, all public datas can be cited, and refer herein to all data all and be can be cited and core
Thought, publicly-owned data can use.
S5:After the completion of the deep learning of the above-mentioned request of correspondence, further include:
New deep learning data will be generated according to this study;Then the new deep learning data are stored to area
Each node of block chain network, storage mode are for example above-mentioned;It also needs to tear the new deep learning data open according to aforesaid way
It is divided into the new metadata with corresponding data attribute tags, and stores in metadata to metadatabase.
After machine is completed deep learning every time, a large amount of new learning data generated is deposited according to aforesaid way
Storage is taken into for meta-data preservation in metadatabase, the number as next deep learning again in public block chain
According to support.It realizes the recycling of data, while helping gradually to improve the corresponding deep learning data of Target Modeling model
Useful value, help more efficient completion to learn.
Embodiment two
The present embodiment correspond to above-described embodiment one specifically use scene.
Deep learning machine starts to learn virtual person of low position is how scene leaping over obstacles under various circumstances advances always.
Obstacle place includes three kinds of enclosure wall, melt pit, stump obstacle patterns;Person of low position needs to bypass enclosure wall, skips melt pit, climbs over three kinds of stump
Different operations.Deep learning is handled image by convolutional neural networks, judges what obstacle front is, then small
People can pass through different mode leaping over obstacles.
In stump obstacle, person of low position can skip stump, but stump have different height it, the spring of person of low position can become
Difference, then generating different data in the case where carrying out convolutional neural networks judgement.These data meeting deep-drawn goes out can be different
The pigment value of height, person of low position can go the pigment value for attempting such height whether can be skipped each time.And if it find that can not
When skipping, it can be moved on by the way of bypassing, detour when, it still can according to pigment value judge will around how far and be
It is no to skip.And if it find that when his pigment value is negative, then this obstacle is exactly a melt pit, person of low position still needs
It skips, some holes can be shorter, and some then can be long, then the different long-jump abilities of person of low position can be tempered.
These either high jumps, are walked around wall, skip melt pit, and a new person of low position will be carried out by a large amount of data
Study, allows him to go to complete such action, especially if when larger change occurs for topography and geomorphology, then this person of low position is just
It needs more to practice data to go to be learnt.
By a kind of deep learning data processing method based on block chain mode of embodiment one be used in person of low position study across
In this scene of obstacle-overpass, a block platform chain can be built first, when there is person of low position to start study, can be stored down all
Learning Support data, that is, deep learning data and person of low position finally modeler model to be achieved, i.e. Target Modeling model.Depth
The data that support data learn as person of low position are practised, it is shared, is disclosed, and when there is data generation, institute can be written immediately
In some block chains;Modeler model data are privately owned, when founding a model, it to be protected to learn in my person of low position
Nobody can first learn to arrive than me before completion.This just uses the mechanism of public key and private key in block chain network well,
Under anonymity feature, and can open and clearization.In my object module library, learning outcome can be provided, is sought by result
Label is looked for, for example, running is as a result, my removal search all and related label of running, such as lift leg, swing arm etc., pass through label
Finding has the metadata of these labels.
In Learning Support data, -8 power concepts that the data least unit by block chain is 10 can extract
Metadata, each metadata carry the Special Significance label of oneself, this is the characteristic of metadata.We complete front person of low position
Run, jump, around etc. the data pick-ups of operations go out meta-data preservation within a block, here, the explanation of metadata is exactly:The study of machine
Be according to the pigment value of image judge, so metadata it is corresponding be exactly image pigment value, including person of low position run, jump, around operation
When leg interval, knee bend, arms swing, lift torso up erect, the pigments value such as frontal scene obstacle.Metadata is also to be shared in area
On block chain.
If that thering is new person of low position to still want to study high jump, being walked around wall, skip melt pit, it can walk, jump by high jump, around wall
Melt pit these object modules one Target Modeling model of configuration is crossed, is placed in my object module library;In learning process, by obtaining
It takes the corresponding learning outcome of Target Modeling model to be learnt to determine label, metadatabase is then gone to directly invoke corresponding member
Data, then the sample data before dismantling, i.e. Learning Support data are found by the label in metadata,;By calling these
It practises support data and carries out deep learning, so that it may to allow person of low position to directly reach high jump, the effect of melt pit is walked, skipped around wall, in short
It, i.e., the Learning Support data being traceable to before dismantling by metadata are learnt.
If that person of low position need learn be creep either climb rocks?So our metadata will be extremely
Useful, in that time that metadata generates, he just stores characteristic value, resource, record of Learning Support data etc., i.e., with mark
The information such as the data attribute of the corresponding Learning Support data of label form storage can be found related to creeping by metadata
Learning Support data of bending over, can find and climb rocks similar Learning Support data of takeofing.Since all metadata have
Multiple labels, then the Learning Support data finally to be learnt can automatically generate.That is, either climbing rocks according to creeping
It determines label, then determines metadata (with relevant Learning Support data of bending over of creeping) according to label, then pass through member
Data determine the Learning Support data finally learnt (with relevant Learning Support data of bending over of creeping, and its correlation
Learning Support data of takeofing as similar in rock-climbing).
The study of person of low position need not be from the beginning, it is only necessary to creep and climb according to the Learning Support data generated
The operation study stepped on, to understanding how to do, is then gone without as other deep learnings, being ceaselessly to fail at the beginning
It carries out.That is, it is completely corresponding if it is scene and final goal and the data stored in present block chain network, then will be direct
Association.Such as person of low position is allowed to learn one step of race per second on level land, if there is such data in block chain, this person of low position can be straight
It connects and learns the action per second for running a step, need not learn from step is started.
All newly-generated learning datas can all extract metadata, be stored in block chain, then block entire in this way
Chain network can be more and more wider, and the content of storage can be more and more, and deep learning database can be increasingly with Target Modeling model library
It is perfect.After being assigned per next fresh target learning tasks, similar metadata and deep learning data can be all found, is directly assisted
Study.
Embodiment three
The present embodiment corresponding embodiment one provides a kind of computer storage media, is stored thereon with computer program, the journey
Sequence can realize the depth based on block chain mode described in above-described embodiment one or embodiment two when being executed by processor
Practise all steps that data sharing method is included.
In conclusion a kind of deep learning data sharing method based on block chain mode provided by the invention, storage are situated between
Matter not only realizes the mode for optimizing deep learning, but also can also realize settling at one go for enhancing study, significantly improves study
Efficiency;Further, the multiplexing of learning data and the saving of learning cost and time are realized.
Example the above is only the implementation of the present invention is not intended to limit the scope of the invention, every to utilize this hair
Equivalents made by bright specification and accompanying drawing content are applied directly or indirectly in relevant technical field, include similarly
In the scope of patent protection of the present invention.
Claims (7)
1. a kind of deep learning data sharing method based on block chain mode, which is characterized in that including:
Storage deep learning data and Target Modeling model to block chain network each node, wherein the deep learning data
It is shared in block chain network, the Target Modeling model is protected;
Deep learning data are split into the metadata with corresponding data attribute tags;
When receiving the deep learning request of a corresponding Target Modeling model, determines and correspond to according to the Target Modeling model
Data attribute label, obtain corresponding metadata;
According to the corresponding metadata, extracts deep learning data corresponding with the Target Modeling model and learnt.
2. the deep learning data sharing method as described in claim 1 based on block chain mode, which is characterized in that described to incite somebody to action
Deep learning data split into the metadata with corresponding data attribute tags, specially:
Deep learning data are disassembled into multiple metadata;
Assign each metadata at least one label, the label record has a number of current meta data, and with current member
The corresponding characteristic value of a data attribute of data.
3. the deep learning data sharing method as claimed in claim 2 based on block chain mode, which is characterized in that it is described according to
Corresponding data attribute label is determined according to the Target Modeling model, obtains corresponding metadata, specially:
The characteristic value of corresponding at least one data attribute is determined according to the Target Modeling model;
Corresponding label is determined according to identified at least one characteristic value;
Corresponding metadata is obtained according to identified label.
4. the deep learning data sharing method as described in claim 1 based on block chain mode, which is characterized in that it is described into
Row learns, and later, further includes:
Generate new deep learning data;
The new deep learning data are stored to each node of block chain network;
The new deep learning data are split into the new metadata with corresponding data attribute tags.
5. the deep learning data sharing method as described in claim 1 based on block chain mode, which is characterized in that also wrap
It includes:
Metadata is stored in the metadatabase of each node of block chain network.
6. the deep learning data sharing method as described in claim 1 based on block chain mode, which is characterized in that according to area
The least unit of block chain data splits the deep learning data.
7. a kind of computer storage media, is stored thereon with computer program, which is characterized in that when the program is executed by processor
It can realize the deep learning data sharing method based on block chain mode described in claim 1-6 any one.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810397795.2A CN108805282A (en) | 2018-04-28 | 2018-04-28 | Deep learning data sharing method, storage medium based on block chain mode |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810397795.2A CN108805282A (en) | 2018-04-28 | 2018-04-28 | Deep learning data sharing method, storage medium based on block chain mode |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108805282A true CN108805282A (en) | 2018-11-13 |
Family
ID=64093498
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810397795.2A Pending CN108805282A (en) | 2018-04-28 | 2018-04-28 | Deep learning data sharing method, storage medium based on block chain mode |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108805282A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109685213A (en) * | 2018-12-29 | 2019-04-26 | 百度在线网络技术(北京)有限公司 | A kind of acquisition methods, device and the terminal device of training sample data |
CN109710691A (en) * | 2018-12-20 | 2019-05-03 | 清华大学 | A kind of mixing block chain model construction method based on deep learning |
CN110058922A (en) * | 2019-03-19 | 2019-07-26 | 华为技术有限公司 | A kind of method, apparatus of the metadata of extraction machine learning tasks |
CN110784507A (en) * | 2019-09-05 | 2020-02-11 | 贵州人和致远数据服务有限责任公司 | Fusion method and system of population information data |
CN113254977A (en) * | 2021-06-24 | 2021-08-13 | 中电科新型智慧城市研究院有限公司 | Sandbox service construction method and device, electronic equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106651346A (en) * | 2016-11-28 | 2017-05-10 | 上海凯岸信息科技有限公司 | Block chain-based credit investigation data sharing and trading system |
CN107659429A (en) * | 2017-08-11 | 2018-02-02 | 四川大学 | Data sharing method based on block chain |
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 |
CN107947940A (en) * | 2017-11-29 | 2018-04-20 | 树根互联技术有限公司 | A kind of method and device of data exchange |
-
2018
- 2018-04-28 CN CN201810397795.2A patent/CN108805282A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106651346A (en) * | 2016-11-28 | 2017-05-10 | 上海凯岸信息科技有限公司 | Block chain-based credit investigation data sharing and trading system |
CN107659429A (en) * | 2017-08-11 | 2018-02-02 | 四川大学 | Data sharing method based on block chain |
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 |
CN107947940A (en) * | 2017-11-29 | 2018-04-20 | 树根互联技术有限公司 | A kind of method and device of data exchange |
Non-Patent Citations (2)
Title |
---|
A.BESIR KURTULMUS EL AL.: "《Trustless Machine Learning Contracts;Evaluating and Exchanging Machine Learning Models on the Ethereum Blockchain》", 《ARXIV》 * |
TRENT MCCONAGHY: "《how blockchains could transform artificial intelligence》", 《HTTP:/DATACONOMY.COM/2016/12/BLOCKCHAINS-FOR-ARTIFICIAL-INTELLIGENCE/》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109710691A (en) * | 2018-12-20 | 2019-05-03 | 清华大学 | A kind of mixing block chain model construction method based on deep learning |
CN109710691B (en) * | 2018-12-20 | 2020-08-28 | 清华大学 | Mixed block chain model construction method based on deep learning |
CN109685213A (en) * | 2018-12-29 | 2019-04-26 | 百度在线网络技术(北京)有限公司 | A kind of acquisition methods, device and the terminal device of training sample data |
CN110058922A (en) * | 2019-03-19 | 2019-07-26 | 华为技术有限公司 | A kind of method, apparatus of the metadata of extraction machine learning tasks |
CN110058922B (en) * | 2019-03-19 | 2021-08-20 | 华为技术有限公司 | Method and device for extracting metadata of machine learning task |
CN110784507A (en) * | 2019-09-05 | 2020-02-11 | 贵州人和致远数据服务有限责任公司 | Fusion method and system of population information data |
CN110784507B (en) * | 2019-09-05 | 2022-12-09 | 贵州人和致远数据服务有限责任公司 | Fusion method and system of population information data |
CN113254977A (en) * | 2021-06-24 | 2021-08-13 | 中电科新型智慧城市研究院有限公司 | Sandbox service construction method and device, electronic equipment and storage medium |
CN113254977B (en) * | 2021-06-24 | 2022-03-18 | 中电科新型智慧城市研究院有限公司 | Sandbox service construction method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108805282A (en) | Deep learning data sharing method, storage medium based on block chain mode | |
Fang et al. | PAWS—A deployed game-theoretic application to combat poaching | |
CN104598611B (en) | The method and system being ranked up to search entry | |
CN105138802B (en) | A kind of gun barrel intelligent design system and design method | |
Fang et al. | Ssml: Self-supervised meta-learner for en route travel time estimation at baidu maps | |
CN114386694A (en) | Drug molecule property prediction method, device and equipment based on comparative learning | |
CN110046907A (en) | Contract based on block chain signs method and device, electronic equipment, storage medium | |
Hinrichs et al. | Transfer learning through analogy in games | |
CN110033269A (en) | Contract based on block chain consults method and device, electronic equipment, storage medium | |
CN110458351A (en) | Area management method, device, equipment and readable storage medium storing program for executing based on flow of the people | |
Chou et al. | Biological-based genetic algorithms for optimized disaster response resource allocation | |
Nissen et al. | Repository support for multi-perspective requirements engineering | |
CN108255059B (en) | Robot control method based on simulator training | |
CN112115276A (en) | Intelligent customer service method, device, equipment and storage medium based on knowledge graph | |
CN104077524B (en) | Training method and viruses indentification method and device for viruses indentification | |
Aljefri et al. | Modeling misperception of options and preferences in the graph model for conflict resolution | |
Duan et al. | MOBA game item recommendation via relation-aware graph attention network | |
Yang et al. | High‐Performance Computing Analysis and Location Selection of Logistics Distribution Center Space Based on Whale Optimization Algorithm | |
CN109977998A (en) | Information processing method and device, storage medium and electronic device | |
CN116051320A (en) | Multitasking attention knowledge tracking method and system for online learning platform | |
Howard et al. | Assessing evolutionary terrain generation methods for curriculum reinforcement learning | |
Chen et al. | Leapfrogging and partial recapitulation as latecomer strategies | |
Kuo et al. | Applying hybrid learning approach to RoboCup's strategy | |
CN112581026B (en) | Joint path planning method for logistics robot on alliance chain | |
Rodriguez et al. | An IR-based artificial bee colony approach for traceability link recovery |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181113 |