CN107832358A - A kind of distributed SVM optimization methods and system - Google Patents

A kind of distributed SVM optimization methods and system Download PDF

Info

Publication number
CN107832358A
CN107832358A CN201710999243.4A CN201710999243A CN107832358A CN 107832358 A CN107832358 A CN 107832358A CN 201710999243 A CN201710999243 A CN 201710999243A CN 107832358 A CN107832358 A CN 107832358A
Authority
CN
China
Prior art keywords
svm
cxn
training
distributed
optimization methods
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
Application number
CN201710999243.4A
Other languages
Chinese (zh)
Inventor
刘小东
蒋杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Aiyouwei Software Development Co Ltd
Original Assignee
Shanghai Aiyouwei Software Development Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Aiyouwei Software Development Co Ltd filed Critical Shanghai Aiyouwei Software Development Co Ltd
Priority to CN201710999243.4A priority Critical patent/CN107832358A/en
Publication of CN107832358A publication Critical patent/CN107832358A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2453—Query optimisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2465—Query processing support for facilitating data mining operations in structured databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28—Databases characterised by their database models, e.g. relational or object models
    • G06F16/284—Relational databases
    • G06F16/285—Clustering or classification

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The embodiment of the present application discloses a kind of distributed SVM optimization methods and system, is related to intelligent terminal technical field.Methods described includes：Obtain sample data set；Iteration is performed based on Tensor Flow platforms；Dividing subset carries out distributed SVM training, and merging obtains data set CXn；Judge whether CXn/CXn 1 ratio P is more than threshold value；If so, judging whether iteration terminates；If so, the SVM of output training.The distributed SVM optimization methods and system of the application, based on Tensor Flow platforms, dividing subset carries out distributed SVM training, then exports the SVM of training, ensures precision of prediction, improves training effectiveness.

Description

A kind of distributed SVM optimization methods and system
Technical field
The application is related to intelligent terminal technical field, more particularly to distributed SVM method and system.
Background technology
With the development of mobile Internet, mobile terminal and data pick-up, data are quick-fried to occur beyond the speed of the imagination The growth of hairdo.Within following a period of time, big data is by as enterprise, society and the important strategic resource of State-level.Enter Enter the big data epoch, useful value is obtained from big data and excavates hiding data rule as important topic, for sea How amount data are classified, and are stored, management, and analysis etc. becomes key issue.In order to solve the above problems, Google is proposed Google FileSystem frameworks are used for storage management mass file；MapReduce frameworks are used to analyze and process number of files According to BigTable is used to manage various books；Apache is proposed Hadoop big data development platforms.Support vector machines (Support Vector Machines) is one of most efficient method in small-scale data mining algorithm, and SVM can be supported linearly The data mining of model and nonlinear model.It is difficult but traditional SVM methods are only applicable to small-scale data mining quantity Point is how SVM good characteristic is applied into large-scale data.
Wherein, as the release of Google MapReduce parallel process models, SVM can be by the way that training sample be carried out Uniformly random cutting, then sub- sample set is trained, and zygote sample set is to obtain final SVMs.So And the model is limited to the particularity of linear model and collective data, and excessive dividing subset module, cause training Accuracy rate reduces.
Accordingly, it is desired to provide a kind of distributed SVM optimization methods and system, based on Tensor Flow platforms, dividing subset Distributed SVM training is carried out, then exports the SVM of training, ensures precision of prediction, improves training effectiveness.
The content of the invention
According to the first aspect of some embodiments of the present application, there is provided a kind of distributed SVM optimization methods, applied to end Hold in (for example, electronic equipment etc.), methods described can include：Obtain sample data set；Performed based on Tensor Flow platforms Iteration；Dividing subset carries out distributed SVM training, and merging obtains data set CXn；Judge whether CXn/CXn-1 ratio P is more than Threshold value；If so, judging whether iteration terminates；If so, the SVM of output training.
In certain embodiments, methods described may further include：If CXn/CXn-1 ratio P is less than threshold value, division Subset carries out distributed SVM training, and merging obtains data set CXn.
In certain embodiments, methods described may further include：If iteration does not terminate, put down based on Tensor Flow Platform performs iteration.
In certain embodiments, the distributed SVM training of the dividing subset progress further comprises：Based on Tensor Flow platforms, sample data set is divided into M block subsets, TX；SVM training is carried out to the M blocks TX subsets.
In certain embodiments, the merging obtains data set CXn and further comprised：Merge TX SVM set, obtain CX1。
In certain embodiments, methods described may further include：CX1 is divided into M block subsets, CTX；To the M Block subset carries out SVM training；Merge CTX SVM set, obtain CX2.
In certain embodiments, whether the ratio P for judging CXn/CXn-1 further comprises more than threshold value：Judge Whether CX2/CX1 ratio is more than threshold value.
In certain embodiments, the distributed SVM of the dividing subset progress is trained for being evenly dividing M block subsets.
In certain embodiments, the M blocks subset is six pieces of subsets.
According to the second aspect of some embodiments of the present application, there is provided a system, including：One memory, by with It is set to data storage and instruction；One is established the processor to communicate with memory, wherein, when performing the instruction in memory, The processor is configured as：Obtain sample data set；Iteration is performed based on Tensor Flow platforms；Dividing subset is divided Cloth SVM is trained, and merging obtains data set CXn；Judge whether CXn/CXn-1 ratio P is more than threshold value；If so, judge iteration Whether terminate；If so, the SVM of output training.
Therefore, according to the distributed SVM optimization methods and system of some embodiments of the present application, based on Tensor Flow Platform, dividing subset carries out distributed SVM training, then exports the SVM of training, ensures precision of prediction, improves training effectiveness.
Brief description of the drawings
To more fully understand and illustrating some embodiments of the present application, below with reference to the description of accompanying drawing reference implementation example, In the drawings, same digital number indicates corresponding part in the accompanying drawings.
Fig. 1 is the illustrative diagram of the Environment System provided according to some embodiments of the present application.
Fig. 2 is the exemplary cell schematic diagram that the electronic functionalities provided according to some embodiments of the present application configure.
Fig. 3 is the exemplary process diagram of the distributed SVM optimization methods provided according to some embodiments of the present application.
Fig. 4 is the exemplary process diagram of the distributed SVM training methods provided according to some embodiments of the present application.
Fig. 5 is that the exemplary of coding flow of the Tensor Flow platforms provided according to some embodiments of the present application is shown It is intended to.
Embodiment
Below with reference to accompanying drawing description for ease of Integrated Understanding the application as defined in claim and its equivalent Various embodiments.These embodiments include various specific details in order to understand, but these be considered only as it is exemplary.Cause This, it will be appreciated by those skilled in the art that carrying out variations and modifications without departing from this to various embodiments described here The scope and spirit of application.In addition, briefly and to be explicitly described the application, the application will be omitted to known function and structure Description.
The term and phrase used in description below and claims is not limited to literal meaning, and be merely can Understand and as one man understand the application.Therefore, for those skilled in the art, it is possible to understand that, there is provided to the various implementations of the application The description of example is only the purpose to illustrate, rather than limitation appended claims and its application of Equivalent definitions.
Below in conjunction with the accompanying drawing in the application some embodiments, the technical scheme in the embodiment of the present application is carried out clear Chu, it is fully described by, it is clear that described embodiment is only some embodiments of the present application, rather than whole embodiments. Based on the embodiment in the application, those of ordinary skill in the art are obtained all under the premise of creative work is not made Other embodiment, belong to the scope of the application protection.
It should be noted that the term used in the embodiment of the present application is only merely for the mesh of description specific embodiment , and it is not intended to be limiting the application." one " of singulative used in the embodiment of the present application and appended claims, "one", " one kind ", " described " and "the" be also intended to including most forms, unless context clearly shows that other implications.Also It should be appreciated that term "and/or" used herein refers to and list items purposes comprising one or more mutually bindings are any Or it is possible to combine.Expression " first ", " second ", " described the first " and " described the second " be used for modify respective element without Consideration order or importance, are used only for distinguishing a kind of element and another element, without limiting respective element.
Terminal according to the application some embodiments can be electronic equipment, the electronic equipment can include smart mobile phone, PC (PC, such as tablet personal computer, desktop computer, notebook, net book, palm PC PDA), mobile phone, e-book Reader, portable media player (PMP), audio/video player (MP3/MP4), video camera, virtual reality device And one or more of combinations in wearable device etc. (VR).According to some embodiments of the present application, the wearable device Type of attachment (such as wrist-watch, ring, bracelet, glasses or wear-type device (HMD)), integrated type (such as electronics can be included Clothes), decorated type (such as pad skin, tatoo or built in electronic device) etc., or several combination.In some realities of the application Apply in example, the electronic equipment can be flexible, be not limited to the said equipment, or can be one kind in above-mentioned various equipment Or several combination.In this application, term " user " can be indicated using the people of electronic equipment or setting using electronic equipment Standby (such as artificial intelligence electronic equipment).
The embodiment of the present application provides a kind of distributed SVM optimization methods.For the ease of understanding the embodiment of the present application, below The embodiment of the present application is described in detail refer to the attached drawing.
Fig. 1 is the illustrative diagram of the Environment System 100 provided according to some embodiments of the present application.Such as Fig. 1 Shown, Environment System 100 can include electronic equipment 110, network 120 and server 130 etc..Electronic equipment 110 can be with Including bus 111, processor 112, memory 113, input/output module 114, display 115, communication module 116 and physics Key 117 etc..In some embodiments of the present application, electronic equipment 110 can omit one or more elements, or can enter one Step includes one or more of the other element.
Bus 111 can include circuit.The circuit can be with one or more element (examples in interconnection electronics 110 Such as, bus 111, processor 112, memory 113, input/output module 114, display 115, communication module 116 and secondary or physical bond 117).The circuit can also be realized between one or more elements in electronic equipment 110 communication (for example, obtain and/or Send information).
Processor 112 can include one or more coprocessors (Co-processor), application processor (AP, Application Processor) and communication processor (Communication Processor).As an example, processor 112 can perform with the control of one or more elements of electronic equipment 110 and/or data processing (for example, log-on data is trained Deng operation).
Memory 113 can be with data storage.The data can include other with one or more of electronic equipment 110 The related instruction of element or data.For example, the data can include the initial data of the before processing of processor 112, intermediate data And/or the data after processing.Memory 113 can include impermanent memory memory and/or permanent memory memory.Make For example, memory 113 can store initial data, sample data etc..
According to some embodiments of the present application, memory 113 can store software and/or program.Described program can wrap Include kernel, middleware, API (API, Application Programming Interface) and/or using journey Sequence (or " application ").
At least a portion of the kernel, the middleware or the API can include operating system (OS, Operating System).As an example, the kernel can be controlled or managed for performing other programs (for example, middle Part, API and application program) in realize operation or function system resource (for example, bus 111, processor 112nd, memory 113 etc.).In addition, the kernel can provide interface.The interface can by the middleware, it is described should One or more elements of electronic equipment 110 are accessed with DLL or the application program to control or management system resource.
The middleware can be as the intermediate layer of data transfer.The data transfer can allow API or Application program is with the kernel communication exchanging data.As an example, the middleware can be handled from the application program One or more task requests of acquisition.For example, the middleware can be to one or more application assigned electronic equipments The priority of 110 system resource (for example, bus 111, processor 112, memory 113 etc.), and processing it is one or Multiple tasks are asked.The API can be that the application program is used to control from the kernel or the middleware The interface of function is provided.The API can also include one or more interfaces or function (for example, instruction).It is described Function can be used for starting control, data channel control, security control, Control on Communication, document control, window control, text control System, image procossing, information processing etc..
Input/output module 114 can send what is inputted from user or external equipment to the other elements of electronic equipment 110 Instruction or data.Input/output module 114 can also be defeated by the instruction of the other elements acquisition from electronic equipment 110 or data Go out to user or external equipment.In certain embodiments, input/output module 114 can include input block, and user can lead to Cross the input block input information or instruction.
Display 115 can be with display content.The content can to user show all kinds (for example, text, image, Video, icon and/or symbol etc., or several combinations).Display 115 can include liquid crystal display (LCD, Liquid Crystal Display), light emitting diode (LED, Light-Emitting Diode) display, Organic Light Emitting Diode (OLED, Organic Light Emitting Diode) display, Micro Electro Mechanical System (MEMS, Micro Electro Mechanical Systems) display or electric paper display etc., or several combinations.Display 115 can include display Screen, touch-screen etc..The display screen can show sample data etc..In certain embodiments, display 115 can be shown virtually Key.The touch-screen can obtain the input of the virtual key.Display 115 can be obtained by the touch-screen and inputted.Institute Touch input, gesture input, action input, the input close to input, electronic pen or user's body part can be included by stating input (for example, hovering input).
Communication module 116 can configure the communication between equipment.In certain embodiments, Environment System 100 can be with Further comprise electronic equipment 140.As an example, the communication between the equipment can include electronic equipment 110 and other set Communication between standby (for example, server 130 or electronic equipment 140).For example, communication module 116 can by radio communication or Wire communication is connected to network 120, realizes and communicates with other equipment (for example, server 130 or electronic equipment 140).
The radio communication can include microwave communication and/or satellite communication etc..The radio communication can include honeycomb Communication is (for example, global mobile communication (GSM, Global System for Mobile Communications), CDMA (CDMA, Code Division Multiple Access), 3G (Third Generation) Moblie (3G, The 3rd Generation Telecommunication), forth generation mobile communication (4G), the 5th third-generation mobile communication (5G), Long Term Evolution (LTE, Long Term Evolution), Long Term Evolution upgrade version (LTE-A, LTE-Advanced), WCDMA (WCDMA, Wideband Code Division Multiple Access), UMTS (UMTS, Universal Mobile Telecommunications System), WiMAX (WiBro, Wireless Broadband) etc., or several combinations.According to some embodiments of the present application, the radio communication can include wireless local Net (WiFi, Wireless Fidelity), bluetooth, low-power consumption bluetooth (BLE, Bluetooth Low Energy), ZigBee protocol (ZigBee), near-field communication (NFC, Near Field Communication), magnetic safe transmission, radio frequency and body area network (BAN, Body Area Network) etc., or several combinations.According to some embodiments of the present application, the wire communication can include GLONASS (Glonass/GNSS, Global Navigation Satellite System), global positioning system System (GPS, Global Position System), Beidou navigation satellite system or Galileo (European GPS) Deng.The wire communication can include USB (USB, Universal Serial Bus), high-definition media interface (HDMI, High-Definition Multimedia Interface), proposed standard 232 (RS-232, Recommend Standard 232), and/or plain old telephone service (POTS, Plain Old Telephone Service) etc., or it is several The combination of kind.
Secondary or physical bond 117 can be used for user mutual.Secondary or physical bond 117 can include one or more entity keys.In some realities Apply in example, user can be with the function of self-defined secondary or physical bond 117.As an example, secondary or physical bond 117 can send instruction.The instruction Log-on data transmission etc. can be included.
In certain embodiments, electronic equipment 110 may further include sensor.The sensor can be included but not It is limited to light sensor, acoustic sensor, gas sensor, chemical sensor, voltage sensitive sensor, temp-sensitive sensor, fluid to pass Sensor, biology sensor, laser sensor, Hall sensor, position sensor, acceleration transducer, intelligence sensor etc., or Several combinations.
In certain embodiments, electronic equipment 110 may further include infrared equipment, image capture device etc..As Example, the infrared equipment can identify by infrared ray mode of delivery, and blink, watch the technical limit spacing eyes such as identification attentively Information.For example, the infrared equipment is acted come certification user profile by gathering the blink of user.As an example, described image Collecting device can include camera, iris device etc..The camera can realize the functions such as eyeball tracking.The iris dress Authentication (for example, certification user profile) can be carried out using iris recognition technology by putting.The iris device can include rainbow Film camera, the iris camera can obtain iris information, and the iris information can be stored in memory 113.
Network 120 can include communication network.The communication network can include computer network (for example, LAN (LAN, Local Area Network) or wide area network (WAN, Wide Area Network)), internet and/or telephone network Deng, or several combinations.Network 120 can be to the other equipment in Environment System 100 (for example, electronic equipment 110, clothes Business device 130, electronic equipment 140 etc.) send information.
Server 130 can be by the other equipment in the connection Environment System 100 of network 120 (for example, electronic equipment 110th, electronic equipment 140 etc.).In certain embodiments, server 130 can enter line number by network 120 and electronic equipment 110 According to transmission etc..For example, server 130 can send sample data etc., electronic equipment 110 by network 120 to electronic equipment 110 Information etc. can be sent to server 130 by network 120.
Electronic equipment 140 can be identical or different with electronic equipment 110 type.According to some embodiments of the present application, The part or all of operation performed in electronic equipment 110 can be in another equipment or multiple equipment (for example, electronic equipment 140 And/or server 130) in perform.In certain embodiments, when electronic equipment 110 be automatically or in response to request perform it is a kind of or When multiple functions and/or service, electronic equipment 110 can ask other equipment (for example, electronic equipment 140 and/or server 130) perform function and/or service are substituted.In certain embodiments, electronic equipment 110 is in addition to perform function or service, further Perform relative one or more functions.In certain embodiments, other equipment is (for example, electronic equipment 140 and/or clothes Business device 130) asked function or other related one or more functions can be performed, implementing result can be sent to electricity Sub- equipment 110.Electronic equipment 110 can repeat result or further handle implementing result, to provide asked function Or service.As an example, electronic equipment 110 can use cloud computing, distributed computing technology and/or client-server end to calculate meter Calculate etc., or several combinations.In certain embodiments, can be included according to the difference of cloud computing service property, the cloud computing Public cloud, private clound and mixed cloud etc..For example, electronic equipment 110 can carry out data transmission with electronic equipment 140.
It should be noted that the description for Environment System 100 above, only for convenience of description, can not be this Shen It please be limited within the scope of illustrated embodiment.It is appreciated that for those skilled in the art, the principle based on the system can Each element can be combined on the premise of without departing substantially from the principle, or forms subsystem and be connected with other elements, To implementing the various modifications and variations on the above method and systematic difference field progress form and details.For example, network environment System 100 may further include database.In another example electronic equipment 110 can not include secondary or physical bond 117 etc..It is all such The deformation of class, within the protection domain of the application.
Fig. 2 is the exemplary cell block diagram that the electronic functionalities provided according to some embodiments of the present application configure.Such as Shown in Fig. 2, processor 112 can include processing module 200, and the processing module 200 can include acquiring unit 210, processing Unit 220, control unit 230, determining unit 240.
According to some embodiments of the present application, acquiring unit 210 can obtain data.In certain embodiments, the number According to that can include information, described information can include but is not limited to text, image, audio, video, action, gesture, sound, eye Eyeball (for example, iris information etc.), breath, light etc., or several combinations.In certain embodiments, described information can include but It is not limited to input information, system information and/or communication information etc..As an example, acquiring unit 210 can pass through input/output Module 114, the touch-screen of display 115, secondary or physical bond 117 and/or sensor obtain the input information of electronic equipment 110.It is described Other equipment (for example, electronic equipment 140) and/or the input of user can be included by inputting information, for example, key-press input, touch-control Input, gesture input, action input, remote input, transmission input, eyes input, sound input, breath input, light input etc., Or several combination.The obtaining widget of the input information can include but is not limited to infrared equipment, image capture device, sensing Device etc., or several combinations.As an example, acquiring unit 210 can obtain sample data etc. by sensor.
In certain embodiments, acquiring unit 210 can obtain the communication information by network 120.The communication information can With including application software information, communication signal (for example, voice signal, vision signal etc.), short message etc..In some embodiments In, acquiring unit 210 can obtain system information by network 120, memory 113 and/or sensor.The system information can With include but is not limited to the system mode of electronic equipment 110, presupposed information, memory 113 store information (for example, iris is recognized Demonstrate,prove information etc.) etc., or several combinations.
In certain embodiments, described information can include instruction.The instruction includes user instruction and/or system command Deng, or several combinations.The instruction can include triggering command, certification instruction, fill in instruction etc., or several combinations.Institute Certification user profile instruction etc. can be included by stating certification instruction.As an example, when user presses secondary or physical bond (for example, shortcut etc.) When, electronic equipment 110 can perform iteration etc..
According to some embodiments of the present application, processing unit 220 can be with processing data.In certain embodiments, processing is single Member 220 can handle sample data etc..As an example, sample data can be evenly dividing as multiple subsets by processing unit 220. For example, sample data set can be divided into M block subsets TX etc. by processing unit 220.In another example processing unit 220 can be by CX1 It is divided into M block subsets CTX etc..In certain embodiments, processing unit 220 can be with merging data set etc..As an example, processing Unit 220 can merge TX SVM set, obtain CX1.
According to some embodiments of the present application, control unit 230 can be with controlled training process.In certain embodiments, control Unit 230 processed can be based on Tensor Flow platforms and perform iteration etc..In certain embodiments, control unit 230 can export The SVM of training.
According to some embodiments of the present application, determining unit 240 can determine information.In certain embodiments, it is it is determined that single Member 240 can determine magnitude relationship between data set ratio and threshold value etc..In certain embodiments, determining unit 240 can be with Determine whether iteration terminates.
It should be noted that described above for the unit in processing module 200, only for convenience of description, can not be this Application is limited within the scope of illustrated embodiment.It is appreciated that for those skilled in the art, the principle based on the system, Unit may be combined on the premise of without departing substantially from the principle, or form submodule and connect with other units Connect, the various modifications and variations in form and details are carried out to the function of implementing above-mentioned module and unit.For example, processing module 200 can further analytic unit, for certification current user information and uniformity etc. of storage user profile.In another example place Reason module 200 may further include memory cell, and the memory cell can store sample data, intermediate data etc..Such as Such deformation, within the protection domain of the application.
Fig. 3 is the exemplary process diagram of the distributed SVM optimization methods provided according to some embodiments of the present application.Such as figure Shown in 3, flow 300 can be realized by processing module 200.
301, sample data set is obtained.Operation 301 can be realized by the acquiring unit 210 of processing module 200.One In a little embodiments, acquiring unit 210 can obtain data by input/output module 114.
302, iteration is performed based on Tensor Flow platforms.Operation 302 can pass through the control list of processing module 200 Member 230 is realized.In certain embodiments, control unit 230 can be based on Tensor Flow platforms and perform the sample data set Iteration.A series of calculating flow graph that the Tensor Flow platforms are made up of nodes, Tensor Flow are a uses DFD (Data Flow), the software library for science numerical computations.The Tensor Flow platforms can be a variety of flat Platform unfolding calculation, such as one or more of desktop computer CPU (or GPS), server, mobile device etc..It is described Tensor Flow platforms are mainly used in the research in terms of machine learning and deep neural network, and its versatility makes it also extensively should For data mining Large-scale parallel computing field.The coding flow of the Tensor Flow platforms is as shown in Figure 5.
303, dividing subset carries out distributed SVM training, and merging obtains data set CXn.Operation 303 can pass through processing The processing unit 220 of module 200 is realized.In certain embodiments, processing unit 220 can be carried out equal to the sample data set Even division.For example, sample data set can be evenly dividing as M block TX subsets by processing unit 220.In certain embodiments, locate Reason unit 220 can merge TX SVM set, obtain data set CXn.
304, judge whether CXn/CXn-1 ratio P is more than threshold value.Operation 304 can be by processing module 200 really Order member 240 is realized.In certain embodiments, determining unit 240 may determine that CXn/CXn-1 ratio P and threshold value take 0.96 Magnitude relationship.
If CXn/CXn-1 ratio P is more than threshold value, into operation 305,305, judge whether iteration terminates.Operation 305 can be realized by the determining unit 240 of processing module 200.In certain embodiments, determining unit 240 can determine to instruct Whether experienced data set causes model that there is the division hyperplane of maximum spacing to judge whether iteration terminates.
306, the SVM of training is exported.Operation 306 can be realized by the control unit 230 of processing module 200.One In a little embodiments, control unit 230 can export the SVM of training by input/output module 114.
According to some embodiments of the present application, SVM algorithm can include：Gather training sample data collection X={ (x1,y1), (x2,y2),…,(xn,yn)},yi∈ { -1 ,+1 }, a division plane is found in training sample set X, by different class samples Separate, division hyperplane is described by following linear equation, as shown in Equation 1：
wTX+b=0 (formula 1)
Wherein, w is normal vector, determines the direction of hyperplane；B be offset determine hyperplane and primary plane away from From；The distance d of sample space x to hyperplane (w, b) is represented by, as shown in Equation 2：
For the sample data divided, (xi,yi)∈X；If yi=+1, then there is wTX+b ＞ 0；If yi, then there is w=- 1Tx+ B ＞ 0, as shown in Equation 3：
In training sample data set, nearest sample point cause formula (3) set up, two foreign peoples to hyperplane away from Distance s are designated as from sum；As shown in Equation 4：
To solve the division hyperplane of maximum distance s, then it is equivalent to meet that the constraints w and b of formula (3) cause minimum Change | | w | |2, then have following SVM basic model formula, as shown in Equation 5：
Its " dual problem " can be obtained using method of Lagrange multipliers to formula (5), specifically, to every treaty of formula (5) Shu Tianjia Lagrange multipliers αi>=0, then the Lagrangian of the problem can be written as, as shown in Equation 6：
Solve to L (w, b, α) and b is that zero bias derivative can obtain：
(7) are substituted into formula (6) and eliminate w and b, are considering the constraints of formula (8), you can obtain the dual problem of formula (5) For as shown in Equation 9：
F (x) model is solved, α is solved by the result of calculation of the 4th step, w and b is then obtained and can obtain F (x) models：
It should be noted that the description for flow 300 above, only for convenience of description, can not be limited in the application Within the scope of illustrated embodiment.It is appreciated that for those skilled in the art, the principle based on the system, may not carry on the back On the premise of from the principle, each operation is combined, or forms sub-process and other operative combinations, in implementation State the various modifications and variations in flow and the function progress form and details of operation.For example, flow 300 may further include The operation such as SVM training is carried out to the subset of division.Such deformation, within the protection domain of the application.
Fig. 4 is the exemplary process diagram of the distributed SVM optimization methods provided according to some embodiments of the present application.Such as figure Shown in 4, flow 400 can be realized by processing module 200.In certain embodiments, flow 400 can be grasped in flow 300 Make a kind of 303 implementation.
401, based on Tensor Flow platforms, sample data set is divided into M block subsets, TX.Operation 401 can lead to The processing unit 220 for crossing processing module 200 is realized.In certain embodiments, processing unit 220 can be carried out to sample data set It is evenly dividing.For example, sample data set can be evenly dividing as 6 pieces of subsets TX, TX1-TX6 by processing unit 220.
402, SVM training is carried out to the M blocks TX subsets.Operation 402 can pass through the control unit of processing module 200 230 realize.In certain embodiments, control unit 230 can individually train the M block TX subsets being evenly dividing, for example, TX1- TX6。
403, merge TX SVM set, obtain CX1.Operation 403 can pass through the processing unit 220 of processing module 200 Realize.In certain embodiments, processing unit 220 can individually training obtains corresponding SVM set merging by M block TX subsets, Obtain CX1.
404, CX1 is divided into M block subsets, CTX.Operation 404 can pass through the processing unit 220 of processing module 200 Realize.In certain embodiments, processing unit 220 can be evenly dividing to CX1.For example, processing unit 220 can incite somebody to action CX1 is evenly dividing as 6 pieces of subsets CTX, CTX1-CTX6.
405, SVM training is carried out to the M blocks CTX subsets.Operation 405 can pass through the control list of processing module 200 Member 230 is realized.In certain embodiments, control unit 230 can individually train the M block CTX subsets being evenly dividing, for example, CTX1-CTX6。
406, merge CTX SVM set, obtain CX2.Operation 406 can pass through the processing unit of processing module 200 220 realize.In certain embodiments, processing unit 220 can individually training obtains corresponding SVM set conjunction by M block CTX subsets And obtain CX2.
It should be noted that the description for flow 400 above, only for convenience of description, can not be limited in the application Within the scope of illustrated embodiment.It is appreciated that for those skilled in the art, the principle based on the system, may not carry on the back On the premise of from the principle, each operation is combined, or forms sub-process and other operative combinations, in implementation State the various modifications and variations in flow and the function progress form and details of operation.For example, to make CXn/CXn-1 ratio P Meet threshold condition, flow 400 can circulate execution operation 401 to operation 403, or circulation and perform operation 404 to operation 406 etc. Operation.Such deformation, within the protection domain of the application.
In summary, according to the distributed SVM optimization methods and system of the embodiment of the present application, put down based on TensorFlow Platform, dividing subset carries out distributed SVM training, then exports the SVM of training, ensures precision of prediction, improves training effectiveness.
It should be noted that the above embodiments are intended merely as example, the application is not limited to such example, but can To carry out various change.
It should be noted that in this manual, term " comprising ", "comprising" or its any other variant are intended to Nonexcludability includes, so that process, method, article or equipment including a series of elements not only will including those Element, but also the other element including being not expressly set out, or it is this process, method, article or equipment also to include Intrinsic key element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that Other identical element also be present in process, method, article or equipment including the key element.
Finally, it is to be noted that, a series of above-mentioned processing are not only included with order described here in temporal sequence The processing of execution, and the processing including performing parallel or respectively rather than in chronological order.
One of ordinary skill in the art will appreciate that realize all or part of flow in above-described embodiment method, being can be with To be completed by the related hardware of computer program instructions, described program can be stored in a computer-readable recording medium, The program is upon execution, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic disc, CD, read-only storage (Read-Only Memory, ROM) or random access memory (Random Access Memory, RAM) Deng.
Above disclosed is only some preferred embodiments of the application, it is impossible to the right model of the application is limited with this Enclose, one of ordinary skill in the art will appreciate that all or part of flow of above-described embodiment is realized, and will according to the application right Made equivalent variations are sought, still falls within and invents covered scope.

Claims (10)

  1. A kind of 1. distributed SVM optimization methods, it is characterised in that including：
    Obtain sample data set；
    Iteration is performed based on Tensor Flow platforms；
    Dividing subset carries out distributed SVM training, and merging obtains data set CXn；
    Judge whether CXn/CXn-1 ratio P is more than threshold value；
    If so, judging whether iteration terminates；
    If so, the SVM of output training.
  2. 2. distributed SVM optimization methods according to claim 1, it is characterised in that further comprise：
    If CXn/CXn-1 ratio P is less than threshold value, dividing subset carries out distributed SVM training, and merging obtains data set CXn.
  3. 3. distributed SVM optimization methods according to claim 1, it is characterised in that further comprise：
    If iteration does not terminate, iteration is performed based on Tensor Flow platforms.
  4. 4. distributed SVM optimization methods according to claim 1, it is characterised in that the dividing subset carries out distributed SVM training further comprises：
    Based on Tensor Flow platforms, sample data set is divided into M block subsets, TX；
    SVM training is carried out to the M blocks TX subsets.
  5. 5. distributed SVM optimization methods according to claim 4, it is characterised in that the merging obtains data set CXn and entered One step includes：
    Merge TX SVM set, obtain CX1.
  6. 6. distributed SVM optimization methods according to claim 5, it is characterised in that further comprise：
    CX1 is divided into M block subsets, CTX；
    SVM training is carried out to the M blocks subset；
    Merge CTX SVM set, obtain CX2.
  7. 7. distributed SVM optimization methods according to claim 6, it is characterised in that the ratio of the judgement CXn/CXn-1 Whether P further comprises more than threshold value：
    Judge whether CX2/CX1 ratio is more than threshold value.
  8. 8. the distributed SVM optimization methods according to claim any one of 1-7, it is characterised in that the dividing subset is entered The distributed SVM of row is trained for being evenly dividing M block subsets.
  9. 9. distributed SVM optimization methods according to claim 8, it is characterised in that the M blocks subset is six pieces of subsets.
  10. A 10. system, it is characterised in that including：
    One memory, is configured as data storage and instruction；
    One is established the processor to communicate with memory, wherein, when performing the instruction in memory, the processor is configured For：
    Obtain sample data set；
    Iteration is performed based on Tensor Flow platforms；
    Dividing subset carries out distributed SVM training, and merging obtains data set CXn；
    Judge whether CXn/CXn-1 ratio P is more than threshold value；
    If so, judging whether iteration terminates；
    If so, the SVM of output training.
CN201710999243.4A 2017-10-20 2017-10-20 A kind of distributed SVM optimization methods and system Pending CN107832358A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710999243.4A CN107832358A (en) 2017-10-20 2017-10-20 A kind of distributed SVM optimization methods and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710999243.4A CN107832358A (en) 2017-10-20 2017-10-20 A kind of distributed SVM optimization methods and system

Publications (1)

Publication Number Publication Date
CN107832358A true CN107832358A (en) 2018-03-23

Family

ID=61648815

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710999243.4A Pending CN107832358A (en) 2017-10-20 2017-10-20 A kind of distributed SVM optimization methods and system

Country Status (1)

Country Link
CN (1) CN107832358A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110867959A (en) * 2019-11-13 2020-03-06 上海迈内能源科技有限公司 Intelligent monitoring system and monitoring method for electric power equipment based on voice recognition

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110867959A (en) * 2019-11-13 2020-03-06 上海迈内能源科技有限公司 Intelligent monitoring system and monitoring method for electric power equipment based on voice recognition

Similar Documents

Publication Publication Date Title
CN107425579A (en) A kind of intelligent charging method and system
CN108334947A (en) A kind of the SGD training methods and system of intelligent optimization
CN108228811A (en) Information recommendation method and system
CN107423100A (en) A kind of methods, devices and systems of terminal OTA firmware updatings
CN107124560A (en) A kind of self-heterodyne system, medium and method
CN107423585A (en) The concealed application method and system of a kind of application
CN107786979A (en) A kind of multiple terminals shared communication method and system
CN107590276A (en) A kind of article map-indication method and system
CN107341378A (en) For the method for the access right for setting application
CN109379431A (en) A kind of contact person information updating method, cloud server and terminal
CN107689227A (en) A kind of voice de-noising method and system based on data fusion
CN107862518A (en) A kind of method of payment and system based on terminal location
CN107220531A (en) A kind of information processing method of convenient login
CN108897479A (en) A kind of terminal touch control method and system
CN107832358A (en) A kind of distributed SVM optimization methods and system
CN206908680U (en) Intelligent terminal with secure log function
CN107566978A (en) A kind of tracking terminal method and system based on intelligent Neural Network
CN108021350A (en) A kind of terminal output volume method of adjustment and system
CN108093312A (en) Interactive approach and system based on barrage technology
CN107690024A (en) Electronic equipment and its control method
CN107277625A (en) A kind of management method and system of video playback history
CN107613106A (en) Similar APP automatic selecting method and system
CN107402690A (en) A kind of global collecting method and system
CN109413266A (en) It is a kind of based on the data back up method and terminal that fall detection
CN108664998A (en) A kind of picture training method and system that paraphrase is reinforced

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

Application publication date: 20180323

WD01 Invention patent application deemed withdrawn after publication