CN109376844A - The automatic training method of neural network and device recommended based on cloud platform and model - Google Patents
The automatic training method of neural network and device recommended based on cloud platform and model Download PDFInfo
- Publication number
- CN109376844A CN109376844A CN201811278821.6A CN201811278821A CN109376844A CN 109376844 A CN109376844 A CN 109376844A CN 201811278821 A CN201811278821 A CN 201811278821A CN 109376844 A CN109376844 A CN 109376844A
- Authority
- CN
- China
- Prior art keywords
- neural network
- sample data
- model
- data
- models
- 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
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)
- Image Analysis (AREA)
Abstract
The neural network automatic training method provided in an embodiment of the present invention recommended based on cloud platform and model and device, under cloud computing platform with cloud by way of existing for the various machine learning clouds that form of a large amount of computers, it is that cloud computing platform default has carried various machine learning models, neural network model is selected by the data type of the sample data inputted according to user, the sample data is trained automatically, solve the problems, such as that machine learning modeling is inconvenient, it is worried for cumbersome details to make the user do not need, the business of oneself can be focused more on, be conducive to innovate and reduce cost.
Description
Technical field
The present embodiments relate to field of artificial intelligence, more particularly, to what is recommended based on cloud platform and model
The automatic training method of neural network and device.
Background technique
Machine learning is a branch of artificial intelligence.The research of artificial intelligence is to attach most importance to from " reasoning " to " to know
Know " attach most importance to, then attach most importance to " study ", a nature, clearly train of thought.Obviously, machine learning is to realize artificial intelligence
An approach, i.e., solve the problems in artificial intelligence by means of machine learning.Machine learning had developed at nearly more than 30 years
One multi-field cross discipline, is related to the multiple subjects such as probability theory, statistics, Approximation Theory, convextiry analysis, computational complexity theory.
Machine Learning Theory be mainly design and analyze it is some allow computer can automatic " study " algorithm.Machine learning algorithm is one
Class is automatically analyzed from data obtains rule, and the algorithm that assimilated equations predict unknown data.Because in learning algorithm
A large amount of statistical theory is related to, machine learning and inferencial statistics contact especially closely, also referred to as Statistical Learning Theory.
Algorithm design aspect, machine Learning Theory concern may be implemented, effective learning algorithm.
Deep learning is that a kind of method based on to data progress representative learning, the benefit of deep learning are in machine learning
Highly effective algorithm, which is extracted, with the feature learning and layered characteristic of non-supervisory formula or Semi-supervised obtains feature by hand to substitute.Depth
Habit is a new field in machine learning research, and motivation is to establish, simulate the nerve net that human brain carries out analytic learning
Network imitates the mechanism of human brain to explain data, such as image, sound and text etc..
But sticked to designed by the demand of its user itself in the prior art, such as the use of certain electric business platform
Family hobby analysis, electric power system data analysis platform, genetic analysis medically etc..System constructed by these schemes is all
It can only be designed for specific mesh, if analysis demand changes, inevitably need to modify to it, need to expend greatly
Manpower is constantly modified.Due to machine learning techniques it is numerous and complicated be responsible for, developing instrument is very wide with frame, many algorithm difficulty compared with
Greatly, it is related to mathematical and computer sciences, it is desirable that there is more high-quality professional to carry out exploitation design.
Summary of the invention
The embodiment of the present invention, which provides, a kind of to be overcome the above problem or at least is partially solved the flat based on cloud of the above problem
The automatic training method of neural network and device that platform and model are recommended.
In a first aspect, the embodiment of the invention provides a kind of neural networks recommended based on cloud platform and model to train automatically
Method, comprising:
The sample data of user's input is obtained, the data type based on the sample data is selected from default recommended models
Multiple neural network models；
Multiple neural network models are trained automatically based on the sample data, obtain multiple initial nerve nets
Network model；
It is based respectively on accuracy rate and calculating speed to be ranked up multiple initial neural network models, and passes through Web
Interactive interface gives back user, recommend respectively the wherein initial neural network model of accuracy rate first, calculating speed first just
Beginning neural network model to select as target nerve network model for user, and provides the target nerve network model
Detailed downloading data.
Second aspect, the embodiment of the invention provides a kind of neural networks recommended based on cloud platform and model to train automatically
Device, comprising:
Model selects cloud platform, for obtaining the sample data of user's input, the data type based on the sample data
Multiple neural network models are selected from default recommended models；
Model training cloud platform, for being instructed automatically based on the sample data to multiple neural network models
Practice, obtains multiple initial neural network models；
Model recommends cloud platform, for being based respectively on accuracy rate and calculating speed to multiple initial neural network models
It is ranked up, and user is given back by Web interactive interface, recommend the initial neural network mould of wherein accuracy rate first respectively
The initial neural network model of type, calculating speed first, to be selected as target nerve network model for user, and described in offer
The detailed downloading data of target nerve network model.
The third aspect, the embodiment of the present invention provides a kind of electronic equipment, including memory, processor and is stored in memory
Computer program that is upper and can running on a processor, is realized when the processor executes described program as first aspect provides
Method the step of.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating
Machine program is realized as provided by first aspect when the computer program is executed by processor the step of method.
The embodiment of the present invention proposes a kind of automatic training method of neural network and dress based on cloud platform and model recommendation
Set, under cloud computing platform with cloud by way of existing for the various machine learning clouds that form of a large amount of computers, be cloud computing
Platform default has carried various machine learning models, selects nerve net by the data type of the sample data inputted according to user
Network model trains the sample data automatically, solves the problems, such as that machine learning modeling is inconvenient, to make the user do not need
It is worried for cumbersome details, the business of oneself can be focused more on, is conducive to innovate and reduce cost.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is to be illustrated according to the automatic training method of neural network of the embodiment of the present invention recommended based on cloud platform and model
Figure；
Fig. 2 is to be illustrated according to the automatic training device of neural network of the embodiment of the present invention recommended based on cloud platform and model
Figure；
Fig. 3 is the entity structure schematic diagram according to the electronic equipment of the embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
It about the technical solution of neural metwork training is sticked to set by the demand of its user itself in the prior art
Meter, for example, the user preferences analysis of certain electric business platform, electric power system data analysis platform, genetic analysis medically etc..
For application, system constructed by these schemes can only all be designed for specific mesh, if data type has
Variation, or analysis demand change, and inevitably need to modify to it, need to expend great manpower and constantly modified.
For system level, this analysis system for specific purpose, routine interface disunity can not be mutual between homologous ray
Communication.And since height customizes, cause to lack wide usage, portable and multiplexing capacity, is a kind of waste of resource.From
For exploitation level, it is responsible for since machine learning techniques are numerous and complicated, developing instrument is very wide with frame, and many algorithm difficulty are larger, relates to
And mathematical and computer sciences, it is desirable that there is more high-quality professional to carry out exploitation design.
As being all to stick in the prior art designed by the demand of its user itself, technology is numerous and complicated to be responsible for, exploitation
Tool is very wide with frame, and many algorithm difficulty are larger, is related to mathematical and computer sciences, it is desirable that has more high-quality professional people
Member carries out exploitation design, big using difficulty for ordinary user, therefore various embodiments of the present invention are directed to and use mind in ordinary user
Solution is provided through network training, specially combines cloud computing technology and the machine learning techniques Coping with Reality problem side of providing
Just efficiently method can focus more on the business of oneself, be conducive to make the user do not need worried for cumbersome details
Innovate and reduce cost.Expansion explanation and introduction will be carried out by multiple embodiments below.
Fig. 1 is a kind of neural network recommended based on cloud platform and model provided in an embodiment of the present invention training side automatically
Method, comprising:
S1, the sample data for obtaining user's input, the data type based on the sample data is from default recommended models
Select multiple neural network models；
S2, multiple neural network models are trained based on the sample data automatically, obtains multiple initial minds
Through network model；
S3, accuracy rate and calculating speed are based respectively on multiple initial neural network models is ranked up, and passed through
Web interactive interface gives back user, recommends the wherein initial neural network model of accuracy rate first, calculating speed first respectively
Initial neural network model to select as target nerve network model for user, and provides the target nerve network model
Detailed downloading data.
In the present embodiment, in the present embodiment, in S1, the sample data inputted by restful interface captures user,
It needs first for the preset neural network model of sample data, every kind of data type is corresponding with multiple default recommended models, i.e., multiple
Preset neural network model, when receiving the sample data of a certain data type, it is necessary first to the number of judgement sample data
According to type, neural network model is selected further according to data type, the neural network model is carried out by above-mentioned sample data
Automatic training solves the problems, such as that machine learning modeling is inconvenient, to make the user do not need as cumbersome details and worried, Neng Gougeng
Add the business for being absorbed in oneself, is conducive to innovate and reduce cost.Finally, multiple neural network models can be obtained, then can again into
Row verifying, chooses optimal neural network model.
In the present embodiment, also it can determine that user needs by auxiliary information by increasing auxiliary information in sample data
It solves the problems, such as, when having selected a variety of neural network models according to data type, according to neural network model in different application
Superiority-inferiority under scene selects an optimal neural network model, is trained to data.In this case, there is no need to
The problem of accuracy rate and calculating speed are calculated in step S3, it is only necessary to which final target nerve network model is recommended into user i.e.
It can.
Specifically, in the present embodiment, the data type of above-mentioned sample data include text, document, image, sound, when
Between sequence；
When data type is text, the problem of being applied to include sentiment analysis, name Entity recognition, part-of-speech tagging,
Semantic role label；To each problem, an optimal neural network model is all selected；Other as data type be document when,
Solve the problems, such as be the theme modeling, document classification；When image, solve the problems, such as image recognition, more Object identifyings or picture search；Sound
Sound problem to be solved is speech recognition；Time series data problem to be solved is forecast analysis.It all can be by being in advance every
A problem selects an optimal neural network model, avoids multiple training.
In the present embodiment, it is additionally provided with markup information in sample data, is trained with facilitating.
In the present embodiment, every kind of data type is preset with multiple default recommended models, in the sample of identification user's input
After the data type of data, selected from default neural network model multiple, the number of preference pattern can be set.
In the present embodiment, above method step is completed by cloud computing, by under cloud computing platform with cloud
Form existing for a large amount of computers compositions various machine learning clouds, be that cloud computing platform default has carried various machine learning
Algorithm, another aspect, the initial modeling cloud that can be made of computer cluster, search space generally estimate cloud, method discovery cloud, EM
(Expectation Maximization Algorithm, EM algorithm) algorithm supports cloud, valuation functions cloud, calculates
Cloud, machine learning algorithm expand cloud, to embody the advantage of cloud, by a large amount of computing resource calculate ordinary user be difficult to or
The suitable parameter that the machine learning for needing long-time numerical behavior to go out uses, while again including being interacted with the Web that user interacts
Interface, machine learning input/output module and cloud management module, to support the operation of cloud computing platform.Make making for machine learning
With the constraint for getting rid of environment, the efficient computing capability of cloud computing platform and the transparency have been given full play to, has been reduced to the greatest extent
Machine learning uses threshold, so that user is without finding suitable machine by experiment repeatedly in the more machine learning methods of comforming
Device learning method is solved in practical application machine learning, model selection it is difficult to predict property, parameter adjust artificial experience
The disadvantages of property, ordinary user's difficulty of learning.
On the basis of the above embodiments, the data type of above-mentioned sample data include text, document, image, sound, when
Between sequence.
In the present embodiment, in order to meet the different demands of user, the training of different types of data sample data is supported, altogether
Support the sample data of five kinds of data types, specially text, document, image, sound and time series, wherein text input is
Term vector needs to carry out Gauss and corrects transformation, and the input of document is word frequency probability or TF-IDF (term frequency-
Inverse document frequency, the common weighting technique of information retrieval data mining), binary system mould can be switched to
Type, image input are binary system or N/A, need to carry out Gauss when being N/A and correct transformation, sound, time series are all N/A, are needed
It carries out Gauss and corrects transformation.
On the basis of the various embodiments described above, the data type based on above-mentioned sample data selects neural network model, tool
Body includes:
If the data type of above-mentioned sample data is text, default recommended models include recurrent neural tensor network
(Recursive Neural Tensor Network, RNTN), deepness belief network (Deep Belief Network, DBN),
Stack noise reduction autocoder and deepness auto encoder；
If the data type of above-mentioned sample data is document, default recommended models include deepness auto encoder, depth letter
Read network and stack noise reduction autocoder；Specifically, above-mentioned deepness auto encoder includes a DBN or stack noise reduction
Autocoder (Stacked Denoising Autoencoders, SDA)；
If the data type of above-mentioned sample data is image, default recommended models include deepness belief network, convolutional Neural
Network (Convolutional Neural Network, CNN), recurrent neural tensor network RNTN and deepness auto encoder；
If the data type of above-mentioned sample data is sound, default recommended models include convolutional neural networks CNN, depth letter
Read network DBN and Recognition with Recurrent Neural Network (Recurrent Neural Network, RNN)；Speech recognition then randomly selects one,
Or multiple neural network models, it is taken after training optimal.
If the data type of above-mentioned sample data is time series, default recommended models include convolutional neural networks CNN, depth
Spend belief network DBN and Recognition with Recurrent Neural Network RNN.
Specifically, in the present embodiment, according to the sample data of different types of data, selecting most suitable neural network mould
Type is trained, wherein multiple neural network models can be chosen to a kind of data type, by a variety of neural network models into
Row training, and optimal models therein are chosen, it can be improved the accuracy of model after training.
Specifically, deepness belief network is substantially a kind of graphical representation network with generative capacity, i.e., it generates and works as
Preceding exemplary all probable values, by such as Boltzmann machine (Restricted Boltzmann Machine, RBM) stacking
At, the invisible layer of each sub-network is next layer of visible layer, and hidden layer or sightless layer are not interconnection, but
Have ready conditions mutually it is independent, text detection, text modeling classification, image, voice, in terms of have very
Important application.
On the basis of the various embodiments described above, when the data type of sample data is text, sentiment analysis selects RNTN to make
For optimal neural network model, Entity recognition is named to select DBN as optimal neural network model, part-of-speech tagging, semanteme
Role's label selects stack noise reduction autocoder, deepness auto encoder as optimal neural network model respectively；
When the data type of sample data is document, theme models selected depth autocoder as optimal nerve net
Network model；Document classification selected depth belief network or stack noise reduction autocoder, select one；
When the data type of sample data is image, image recognition selected depth belief network is as optimal neural network
Model, more Object identifyings select one and select CNN or RNTN as optimal neural network model, and a selection recurrence mind is then selected in picture search
Through tensor network RNTN or deepness auto encoder as optimal neural network model.
When the data type of sample data is time series, then one or multiple neural network models are randomly selected, instructed
It is taken after white silk optimal.
On the basis of the various embodiments described above, the sample data is trained automatically based on the neural network model
Before, further includes:
Based on the input format of the neural network model, the sample data is formatted, sample data is made
Format and corresponding neural network model input format it is unified.
On the basis of the various embodiments described above, after the sample data and the markup information that obtain user's input, further includes:
Above-mentioned sample data and markup information are encrypted and stored to cloud.
It in the present embodiment, can be in input sample data and mark if user needs to maintain secrecy to its training sample
When information, selection carries out secrecy operation to sample data, markup information, then after cloud encrypts sample data and markup information,
It is stored again to cloud.
On the basis of the various embodiments described above, accuracy rate and calculating speed are based respectively on to multiple initial neural networks
Model is ranked up, and is specifically included:
Partial data is randomly selected from the sample data as test set, the test set is separately input to each
In initial neural network model, the accuracy rate of each initial neural network model is verified, and records each initial neural network mould
The calculating speed of type.
In the present embodiment, corresponding when only having selected a kind of neural network model according to the data type of sample data
The corresponding neural network model of the maximum history typical event of only one similarity, also only one target mind of finally training
Through network model；If when having selected a variety of neural network models according to the data type of sample data, each neural network mould
Type corresponds to the maximum history typical event of a similarity, that is, is corresponding with a neural network model, and final training obtains multiple
Initial neural network model, at this point, then needing to test obtained initial neural network model, wherein effect is optimal for selection
Initial neural network model as target nerve network model, in the present embodiment, pass through accuracy rate and calculating speed two
Aspect is ranked up, selection wherein the initial neural network model of accuracy rate first, calculating speed first initial neural network
Model, to be selected as target nerve network model for user.
On the basis of the various embodiments described above, before above-mentioned sample data is divided into training set and test set, further includes:
It is unitized, after digitization to above-mentioned sample data, successively carries out missing values processing, noise data processing, number
According to cleaning, data integration, data transformation and data reduction process.
In the present embodiment, due in training, test process, needing to be converted to sample data and neural network model
The identical format of input data because in the present embodiment, by the input lattice of each neural network model under each data type
Formula is unified, after receiving sample data, after the format unification of sample data, digitization, successively progress missing values processing,
Noise data processing, data scrubbing, data integration, data transformation and data reduction process.
The present embodiment additionally provides a kind of automatic training device of neural network recommended based on cloud platform and model, based on upper
The method for stating each embodiment, as shown in Fig. 2, including that model selection cloud platform 30, model training cloud platform 40 and model recommend cloud
Platform 50, in which:
Model selects cloud platform 30 to obtain the sample data of user's input, and the data type based on the sample data is from pre-
If selecting multiple neural network models in recommended models；
Model training cloud platform 40 is based on the sample data and is trained automatically to multiple neural network models, obtains
To multiple initial neural network models；
Model recommend cloud platform 50 be based respectively on accuracy rate and calculating speed to multiple initial neural network models into
Row sequence, and gives back user by Web interactive interface, recommend respectively wherein the initial neural network model of accuracy rate first,
The initial neural network model of calculating speed first to select as target nerve network model for user, and provides the mesh
Mark the detailed downloading data of neural network model.
Fig. 3 is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention, as shown in figure 3, the electronic equipment
It may include: processor (processor) 810,820, memory communication interface (Communications Interface)
(memory) 830 and communication bus 840, wherein processor 810, communication interface 820, memory 830 pass through communication bus 840
Complete mutual communication.Processor 810 can call the meter that is stored on memory 830 and can run on processor 810
Calculation machine program, to execute the automatic training method of neural network of the various embodiments described above offer recommended based on cloud platform and model,
For example,
S1, the sample data for obtaining user's input, the data type based on the sample data is from default recommended models
Select multiple neural network models；
S2, multiple neural network models are trained based on the sample data automatically, obtains multiple initial minds
Through network model；
S3, accuracy rate and calculating speed are based respectively on multiple initial neural network models is ranked up, and passed through
Web interactive interface gives back user, recommends the wherein initial neural network model of accuracy rate first, calculating speed first respectively
Initial neural network model to select as target nerve network model for user, and provides the target nerve network model
Detailed downloading data.
In addition, the logical order in above-mentioned memory 830 can be realized by way of SFU software functional unit and conduct
Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally
The technical solution of the inventive embodiments substantially part of the part that contributes to existing technology or the technical solution in other words
It can be embodied in the form of software products, which is stored in a storage medium, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes the present invention respectively
The all or part of the steps of a embodiment the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory
(ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk
Etc. the various media that can store program code.
The embodiment of the present invention also provides a kind of non-transient computer readable storage medium, is stored thereon with computer program,
The computer program is implemented to carry out recommending based on cloud platform and model for the various embodiments described above offers when being executed by processor
The automatic training method of neural network, for example,
S1, the sample data for obtaining user's input, the data type based on the sample data is from default recommended models
Select multiple neural network models；
S2, multiple neural network models are trained based on the sample data automatically, obtains multiple initial minds
Through network model；
S3, accuracy rate and calculating speed are based respectively on multiple initial neural network models is ranked up, and passed through
Web interactive interface gives back user, recommends the wherein initial neural network model of accuracy rate first, calculating speed first respectively
Initial neural network model to select as target nerve network model for user, and provides the target nerve network model
Detailed downloading data.
The embodiment of the present invention also provides the present embodiment and discloses a kind of computer program product, the computer program product packet
The computer program being stored in non-transient computer readable storage medium is included, the computer program includes program instruction, when
When described program instruction is computer-executed, computer is able to carry out such as the above-mentioned nerve net recommended based on cloud platform and model
The automatic training method of network, for example,
S1, the sample data for obtaining user's input, the data type based on the sample data is from default recommended models
Select multiple neural network models；
S2, multiple neural network models are trained based on the sample data automatically, obtains multiple initial minds
Through network model；
S3, accuracy rate and calculating speed are based respectively on multiple initial neural network models is ranked up, and passed through
Web interactive interface gives back user, recommends the wherein initial neural network model of accuracy rate first, calculating speed first respectively
Initial neural network model to select as target nerve network model for user, and provides the target nerve network model
Detailed downloading data.
In conclusion a kind of neural network recommended based on cloud platform and model provided in an embodiment of the present invention is trained automatically
Method and apparatus, under cloud computing platform with cloud by way of existing for the various machine learning clouds that form of a large amount of computers,
It is that cloud computing platform default has carried various machine learning models, is selected by the data type of the sample data inputted according to user
Neural network model is selected, the sample data is trained automatically, solves the problems, such as that machine learning modeling is inconvenient, to make
User is worried without being cumbersome details, can focus more on the business of oneself, be conducive to innovate and reduce cost.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of automatic training method of neural network recommended based on cloud platform and model characterized by comprising
The sample data of user's input is obtained, the data type based on the sample data selects multiple from default recommended models
Neural network model；
Multiple neural network models are trained automatically based on the sample data, obtain multiple initial neural network moulds
Type；
It is based respectively on accuracy rate and calculating speed to be ranked up multiple initial neural network models, and passes through Web interaction
User is given back at interface, recommends the initial mind of the wherein initial neural network model of accuracy rate first, calculating speed first respectively
Through network model, to select as target nerve network model for user, and the detailed of the target nerve network model is provided
Downloading data.
2. the neural network automatic training method according to claim 1 recommended based on cloud platform and model, feature are existed
In the sample data for obtaining user's input specifically includes:
The sample data inputted by restful interface captures user, the data type of the sample data be text, document,
Image, sound or time series.
3. the neural network automatic training method according to claim 2 recommended based on cloud platform and model, feature are existed
In, the data type based on the sample data selects multiple neural network models from default recommended models, it specifically includes:
If the data type of the sample data is text, presetting recommended models includes recurrent neural tensor network RNTN, depth
Spend belief network DBN, stack noise reduction autocoder and deepness auto encoder；
If the data type of the sample data is document, presetting recommended models includes deepness auto encoder, depth conviction
Network and stack noise reduction autocoder；
If the data type of the sample data is image, presetting recommended models includes deepness belief network, convolutional Neural net
Network CNN, recurrent neural tensor network RNTN and deepness auto encoder；
If the data type of the sample data is sound, presetting recommended models includes convolutional neural networks CNN, depth conviction
Network DBN and Recognition with Recurrent Neural Network RNN；
If the data type of the sample data is time series, presetting recommended models includes convolutional neural networks CNN, depth
Belief network DBN and Recognition with Recurrent Neural Network RNN.
4. the neural network automatic training method according to claim 2 recommended based on cloud platform and model, feature are existed
In before being trained automatically based on the sample data to multiple neural network models progress, further includes:
Based on the input format of the neural network model, the sample data is formatted, makes the lattice of sample data
Formula is unified with the input format of corresponding neural network model.
5. the neural network automatic training method according to claim 1 recommended based on cloud platform and model, feature are existed
In after the sample data and the markup information that obtain user's input, further includes:
The sample data and markup information are encrypted and stored to cloud.
6. the neural network automatic training method according to claim 1 recommended based on cloud platform and model, feature are existed
In being based respectively on accuracy rate and calculating speed and be ranked up to multiple initial neural network models, specifically included:
Partial data is randomly selected from the sample data as test set, the test set is separately input to each initial
In neural network model, the accuracy rate of each initial neural network model is verified, and records each initial neural network model
Calculating speed.
7. the neural network automatic training method according to claim 5 recommended based on cloud platform and model, feature are existed
In before the sample data is divided into training set and test set, further includes:
It is unitized, after digitization to the sample data, successively carries out missing values processing, noise data is handled, data are clear
Reason, data integration, data transformation and data reduction process.
8. a kind of automatic training device of neural network recommended based on cloud platform and model characterized by comprising
Model selects cloud platform, and for obtaining the sample data of user's input, the data type based on the sample data is from pre-
If selecting multiple neural network models in recommended models；
Model training cloud platform is obtained for being trained automatically based on the sample data to multiple neural network models
To multiple initial neural network models；
Model recommends cloud platform, carries out for being based respectively on accuracy rate and calculating speed to multiple initial neural network models
Sequence, and user is given back by Web interactive interface, recommend initial neural network model, the meter of wherein accuracy rate first respectively
The initial neural network model for calculating speed regulation one to select as target nerve network model for user, and provides the target
The detailed downloading data of neural network model.
9. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor realizes method as described in any one of claim 1 to 7 when executing described program
The step of.
10. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the calculating
The step of machine program realizes method as described in any one of claim 1 to 7 when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811278821.6A CN109376844A (en) | 2018-10-30 | 2018-10-30 | The automatic training method of neural network and device recommended based on cloud platform and model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811278821.6A CN109376844A (en) | 2018-10-30 | 2018-10-30 | The automatic training method of neural network and device recommended based on cloud platform and model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109376844A true CN109376844A (en) | 2019-02-22 |
Family
ID=65390769
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811278821.6A Pending CN109376844A (en) | 2018-10-30 | 2018-10-30 | The automatic training method of neural network and device recommended based on cloud platform and model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109376844A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934512A (en) * | 2019-03-28 | 2019-06-25 | 努比亚技术有限公司 | A kind of training method and system of prediction model |
| CN110046706A (en) * | 2019-04-18 | 2019-07-23 | 腾讯科技（深圳）有限公司 | Model generating method, device and server |
| CN110135289A (en) * | 2019-04-28 | 2019-08-16 | 北京天地玛珂电液控制系统有限公司 | A kind of underground coal mine intelligent use cloud service platform based on deep learning |
| CN110278370A (en) * | 2019-06-21 | 2019-09-24 | 上海摩象网络科技有限公司 | Automatically generate method, apparatus, the electronic equipment of shooting controlling mechanism |
| CN110334955A (en) * | 2019-07-08 | 2019-10-15 | 北京字节跳动网络技术有限公司 | Processing method, device, equipment and the storage medium of index evaluation |
| CN110443310A (en) * | 2019-08-07 | 2019-11-12 | 浙江大华技术股份有限公司 | Compare update method, server and the computer storage medium of analysis system |
| CN110532113A (en) * | 2019-08-30 | 2019-12-03 | 北京地平线机器人技术研发有限公司 | Information processing method, device, computer readable storage medium and electronic equipment |
| CN111276254A (en) * | 2020-01-13 | 2020-06-12 | 印迹信息科技（北京）有限公司 | Medical open platform system and diagnosis and treatment data processing method |
| CN111695670A (en) * | 2019-03-11 | 2020-09-22 | 深圳市茁壮网络股份有限公司 | Neural network model training method and device |
| CN111882059A (en) * | 2020-07-17 | 2020-11-03 | 北京市商汤科技开发有限公司 | Data processing method and device, electronic equipment and storage medium |
| CN112101529A (en) * | 2020-10-14 | 2020-12-18 | 杭州海康威视数字技术股份有限公司 | Cross-platform deployment method and framework for neural network model inference |
| CN112214470A (en) * | 2019-07-11 | 2021-01-12 | 纬创资通股份有限公司 | Data acquisition device and data operation system and method |
| CN112883654A (en) * | 2021-03-24 | 2021-06-01 | 国家超级计算天津中心 | Model training system based on data driving |
| CN113408634A (en) * | 2021-06-29 | 2021-09-17 | 深圳市商汤科技有限公司 | Model recommendation method and device, equipment and computer storage medium |
| CN113472866A (en) * | 2021-06-23 | 2021-10-01 | 北京理工大学珠海学院 | Cloud neural network data interface method based on TCP (Transmission control protocol) |
| CN113553816A (en) * | 2020-11-16 | 2021-10-26 | 厦门市和家健脑智能科技有限公司 | Method and device for automatically optimizing scale based on intelligent model small data volume sample |
| CN113673706A (en) * | 2020-05-15 | 2021-11-19 | 富泰华工业（深圳）有限公司 | Machine learning model training method and device and electronic equipment |
| CN114579822A (en) * | 2021-12-13 | 2022-06-03 | 北京市建筑设计研究院有限公司 | Method and device for pushing modeling tool, electronic equipment and storage medium |
| CN116070675A (en) * | 2023-03-06 | 2023-05-05 | 西南交通大学 | Side slope neural network model selection method, device, equipment and readable storage medium |
| CN116955592A (en) * | 2023-07-21 | 2023-10-27 | 广州拓尔思大数据有限公司 | Data processing method and system based on visual reasoning result |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101782976A (en) * | 2010-01-15 | 2010-07-21 | 南京邮电大学 | Automatic selection method for machine learning in cloud computing environment |
| US20160306798A1 (en) * | 2015-04-16 | 2016-10-20 | Microsoft Corporation | Context-sensitive content recommendation using enterprise search and public search |
| CN107169031A (en) * | 2017-04-17 | 2017-09-15 | 广东工业大学 | It is a kind of to recommend method based on the picture materials that depthmeter reaches |
| CN108431833A (en) * | 2015-12-23 | 2018-08-21 | 多伦多道明银行 | End-to-end depth collaborative filtering |
| CN108446388A (en) * | 2018-03-22 | 2018-08-24 | 平安科技（深圳）有限公司 | Text data quality detecting method, device, equipment and computer readable storage medium |
-
2018
- 2018-10-30 CN CN201811278821.6A patent/CN109376844A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101782976A (en) * | 2010-01-15 | 2010-07-21 | 南京邮电大学 | Automatic selection method for machine learning in cloud computing environment |
| US20160306798A1 (en) * | 2015-04-16 | 2016-10-20 | Microsoft Corporation | Context-sensitive content recommendation using enterprise search and public search |
| CN108431833A (en) * | 2015-12-23 | 2018-08-21 | 多伦多道明银行 | End-to-end depth collaborative filtering |
| CN107169031A (en) * | 2017-04-17 | 2017-09-15 | 广东工业大学 | It is a kind of to recommend method based on the picture materials that depthmeter reaches |
| CN108446388A (en) * | 2018-03-22 | 2018-08-24 | 平安科技（深圳）有限公司 | Text data quality detecting method, device, equipment and computer readable storage medium |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111695670A (en) * | 2019-03-11 | 2020-09-22 | 深圳市茁壮网络股份有限公司 | Neural network model training method and device |
| CN109934512A (en) * | 2019-03-28 | 2019-06-25 | 努比亚技术有限公司 | A kind of training method and system of prediction model |
| CN109934512B (en) * | 2019-03-28 | 2024-02-09 | 努比亚技术有限公司 | Prediction model training method and training system |
| CN110046706B (en) * | 2019-04-18 | 2022-12-20 | 腾讯科技（深圳）有限公司 | Model generation method and device and server |
| CN110046706A (en) * | 2019-04-18 | 2019-07-23 | 腾讯科技（深圳）有限公司 | Model generating method, device and server |
| CN110135289A (en) * | 2019-04-28 | 2019-08-16 | 北京天地玛珂电液控制系统有限公司 | A kind of underground coal mine intelligent use cloud service platform based on deep learning |
| CN110278370A (en) * | 2019-06-21 | 2019-09-24 | 上海摩象网络科技有限公司 | Automatically generate method, apparatus, the electronic equipment of shooting controlling mechanism |
| CN110334955A (en) * | 2019-07-08 | 2019-10-15 | 北京字节跳动网络技术有限公司 | Processing method, device, equipment and the storage medium of index evaluation |
| CN110334955B (en) * | 2019-07-08 | 2021-09-14 | 北京字节跳动网络技术有限公司 | Index evaluation processing method, device, equipment and storage medium |
| CN112214470B (en) * | 2019-07-11 | 2024-05-17 | 纬创资通股份有限公司 | Data acquisition device and data operation system and method |
| CN112214470A (en) * | 2019-07-11 | 2021-01-12 | 纬创资通股份有限公司 | Data acquisition device and data operation system and method |
| US11297157B2 (en) | 2019-07-11 | 2022-04-05 | Wistron Corporation | Data capturing device and data calculation system and method |
| TWI773907B (en) * | 2019-07-11 | 2022-08-11 | 緯創資通股份有限公司 | Data capturing apparatus and data calculation system and method |
| CN110443310A (en) * | 2019-08-07 | 2019-11-12 | 浙江大华技术股份有限公司 | Compare update method, server and the computer storage medium of analysis system |
| CN110443310B (en) * | 2019-08-07 | 2022-08-09 | 浙江大华技术股份有限公司 | Updating method of comparison analysis system, server and computer storage medium |
| CN110532113A (en) * | 2019-08-30 | 2019-12-03 | 北京地平线机器人技术研发有限公司 | Information processing method, device, computer readable storage medium and electronic equipment |
| CN111276254A (en) * | 2020-01-13 | 2020-06-12 | 印迹信息科技（北京）有限公司 | Medical open platform system and diagnosis and treatment data processing method |
| CN113673706A (en) * | 2020-05-15 | 2021-11-19 | 富泰华工业（深圳）有限公司 | Machine learning model training method and device and electronic equipment |
| CN111882059A (en) * | 2020-07-17 | 2020-11-03 | 北京市商汤科技开发有限公司 | Data processing method and device, electronic equipment and storage medium |
| CN112101529A (en) * | 2020-10-14 | 2020-12-18 | 杭州海康威视数字技术股份有限公司 | Cross-platform deployment method and framework for neural network model inference |
| CN113553816A (en) * | 2020-11-16 | 2021-10-26 | 厦门市和家健脑智能科技有限公司 | Method and device for automatically optimizing scale based on intelligent model small data volume sample |
| CN113553816B (en) * | 2020-11-16 | 2024-09-10 | 厦门市和家健脑智能科技有限公司 | Method and device for automatically optimizing scale based on small data volume sample of intelligent model |
| CN112883654A (en) * | 2021-03-24 | 2021-06-01 | 国家超级计算天津中心 | Model training system based on data driving |
| CN113472866A (en) * | 2021-06-23 | 2021-10-01 | 北京理工大学珠海学院 | Cloud neural network data interface method based on TCP (Transmission control protocol) |
| CN113408634B (en) * | 2021-06-29 | 2022-07-05 | 深圳市商汤科技有限公司 | Model recommendation method and device, equipment and computer storage medium |
| CN113408634A (en) * | 2021-06-29 | 2021-09-17 | 深圳市商汤科技有限公司 | Model recommendation method and device, equipment and computer storage medium |
| CN114579822A (en) * | 2021-12-13 | 2022-06-03 | 北京市建筑设计研究院有限公司 | Method and device for pushing modeling tool, electronic equipment and storage medium |
| CN116070675B (en) * | 2023-03-06 | 2023-06-09 | 西南交通大学 | Side slope neural network model selection method, device, equipment and readable storage medium |
| CN116070675A (en) * | 2023-03-06 | 2023-05-05 | 西南交通大学 | Side slope neural network model selection method, device, equipment and readable storage medium |
| CN116955592A (en) * | 2023-07-21 | 2023-10-27 | 广州拓尔思大数据有限公司 | Data processing method and system based on visual reasoning result |
| CN116955592B (en) * | 2023-07-21 | 2024-02-09 | 广州拓尔思大数据有限公司 | Data processing method and system based on visual reasoning result |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109376844A (en) | The automatic training method of neural network and device recommended based on cloud platform and model | |
| CN109635918A (en) | The automatic training method of neural network and device based on cloud platform and preset model | |
| KR102071582B1 (en) | Method and apparatus for classifying a class to which a sentence belongs by using deep neural network | |
| CN108334605B (en) | Text classification method and device, computer equipment and storage medium | |
| Kosinski et al. | Mining big data to extract patterns and predict real-life outcomes. | |
| CN111507768B (en) | Potential user determination method and related device | |
| CN110781321A (en) | Multimedia content recommendation method and device | |
| CN109271493A (en) | A kind of language text processing method, device and storage medium | |
| CN112163165A (en) | Information recommendation method, device, equipment and computer readable storage medium | |
| CN109816438B (en) | Information pushing method and device | |
| US11423307B2 (en) | Taxonomy construction via graph-based cross-domain knowledge transfer | |
| Liu et al. | Winning solutions and post-challenge analyses of the ChaLearn AutoDL challenge 2019 | |
| CN111859149A (en) | Information recommendation method and device, electronic equipment and storage medium | |
| Coelho et al. | Building Machine Learning Systems with Python: Explore machine learning and deep learning techniques for building intelligent systems using scikit-learn and TensorFlow | |
| CN110309114A (en) | Processing method, device, storage medium and the electronic device of media information | |
| Sosnovshchenko et al. | Machine learning with Swift: artificial intelligence for iOS | |
| CN111563158A (en) | Text sorting method, sorting device, server and computer-readable storage medium | |
| Concolato et al. | Data science: A new paradigm in the age of big-data science and analytics | |
| CN115577316A (en) | User personality prediction method based on multi-mode data fusion and application | |
| EP4261763A1 (en) | Apparatus and method for providing user's interior style analysis model on basis of sns text | |
| CN112417267A (en) | User behavior analysis method and device, computer equipment and storage medium | |
| CN109242927B (en) | Advertisement template generation method and device and computer equipment | |
| Chanaa et al. | Context-aware factorization machine for recommendation in massive open online courses (MOOCs) | |
| CN114202402A (en) | Behavior characteristic prediction method and device | |
| CN110717539A (en) | Dimension reduction model training method, retrieval method and device based on artificial intelligence |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190222 |
|
| RJ01 | Rejection of invention patent application after publication |