CN109933307A - A kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology - Google Patents
A kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology Download PDFInfo
- Publication number
- CN109933307A CN109933307A CN201910136456.3A CN201910136456A CN109933307A CN 109933307 A CN109933307 A CN 109933307A CN 201910136456 A CN201910136456 A CN 201910136456A CN 109933307 A CN109933307 A CN 109933307A
- Authority
- CN
- China
- Prior art keywords
- model
- ontology
- mhg
- feature
- meta
- 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
Abstract
The present invention devises a kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology, specifically: the semantic information of different semantic hierarchies is modeled and unified formalized model is provided, it realizes jumping between each semantic hierarchies, constitutes criss-cross semantic structure;Using the domain features model of extension as the description of embedded intelligence control field inner member business semantics basis, initial embedded intelligence control field conceptual architecture is generated in an automated way;The transition designed using component semantics as feature to pel component is defined the function distribution and interaction semantics of each machine learning algorithm pel component.Present invention improves over existing knowledge expression frame non-functional feature description and feature realize changeability in terms of existing for deficiency, improve software code reuse efficiency, be convenient for embedded intelligence control software product mass customization production.
Description
Technical field
The present invention relates to a kind of formalized description frame of machine learning algorithm module and the methods of encapsulation, and in particular to one
It plants based on ontology, towards the machine learning algorithm modular form describing framework and packaging method of intelligent controller software development.
Background technique
The embedded novel intelligent controller software systems of field of industrial manufacturing have scale multiplicity, complexity height, coupling
The features such as degree is close, life cycle is long is very suitable for software modularity exploitation.Domestic and foreign scholars open in embedded software product
Hair and software reuse etc. have done a large amount of research work, but there are still part deficiencies.
It is existing based on the feature modeling method of domain analysis to the field of the non-functional aspect such as performance, expense, handling capacity
Signature analysis then considers less, and cannot support customized feature constraint relationship, in terms of algoritic module instantiation, only from can
The angle for becoming Feature Selection discusses the production of software product, and it is diversified and embedding with intelligent plant not account for feature implementation
The problems such as entering the selection and positioning of the corresponding algorithm pel component of formula application field feature, it is embedded in terms of knowledge representation
Software product engineering is differently formed Semantic hierarchy because exploitation participant's focus, and existing characteristic model is showed only as feature
Relationship graph, and lack unified formalized model, it is not able to satisfy the needs of semantic diversity and consistency check.
Summary of the invention
The present invention is directed to intelligent controller field, and existing knowledge is expressed frame and realized in the description of non-functional feature and feature
Changeability description etc. Shortcomings, propose a kind of machine learning algorithm modular form describing framework and envelope based on ontology
Dress method.This method includes formalized description frame, field of intelligent control characteristic spectrum, machine learning algorithm figure based on ontology
The encapsulation and description of first component.
Formalized description frame based on ontology is using unified formalized model to the semantic information of different semantic hierarchies
Unified formalized description is modeled and provided, realizes jumping between each semantic hierarchies, constitutes criss-cross semantic knot
Structure, to meet the semantic multifarious exposition need of deep learning algoritic module.
Field of intelligent control characteristic spectrum is using the domain features model of extension as embedded intelligence control field inner member
The description basis of business semantics, generates initial embedded intelligence control field conceptual architecture in an automated way.
The mistake that the encapsulation and description of machine learning algorithm pel component are designed using component semantics as feature to pel component
Cross, the function distribution and interaction semantics of each machine learning algorithm pel component be defined, as associated components encapsulation and
The basis of description.
Detailed description of the invention
Fig. 1 is that the present invention is based on the field of intelligent control characteristic spectrum meta-models of ontology;
Fig. 2 is that the design of machine learning algorithm pel framing component is realized the present invention is based on multi-mode;
Fig. 3 is that the module of the hiding alternative changing features of the present invention realizes component design;
Fig. 4 is method frame figure of the invention.
Specific embodiment
For the technology contents that the present invention will be described in detail, construction feature, the purpose and effect realized, below in conjunction with embodiment party
Formula simultaneously cooperates attached drawing to be explained.
To support to define the ontology of description domain knowledge semanteme for different application fields, and further it is abstracted and is characterized
Construction feature map and its meta-model, machine learning algorithm module describing framework are based on famous ontology construction tool Prot é g
é carries out secondary development, by the way that feature extraction and inference function construction feature modeling tool are appropriately modified and increased to it.
Prot é g é is by the ontology construction tool of the open source of Stanford University's exploitation, and function is very powerful, it can be achieved that ontology is general
The definition and maintenance of thought, conceptual relation, concept attribute and axiom.By interactive interface, user can be edited in domain body
Hold and establish body construction, with term common in Cover Characteristics map and its meta-model, concept, and entirety --- part, one
As --- the conceptual relations such as specialization.On this basis, characteristics map and its meta-model are established and straight with visualized graphs
Sight is shown.Since the software products such as pel component not yet developed at this time, the location of Artifact instances of ontology
Equal data attribute informations will specify in the component development stage.By being integrated with ontology inferences engines such as Jena, Racer, base
Consistency and validation verification can be carried out to characteristics map according to meta-model in the feature modeling tool of Prot é g é.
Intelligent controller machine learning algorithm module encapsulation tool is quasi- to carry out developing plug building by Integrated Development Environment
Domain architecture and component Integrated Development Environment.The Integrated Development Environment passes through feature of the importing based on Prot é g é first and builds
Die worker has the characteristics map OWL generated and describes file generated characteristic model, and on this basis, developer can be by figure circle
Face visual development and compilation and design model, and different design patterns is used according to changing features componentbased development guide
Automatic code generating frame can effectively improve the standardization level of pel component, hide variable information, improve pel component development
And multiplexing efficiency.Integrated Development Environment realizes the seamless combination with component base, the software products such as component that developed simultaneously
It can be transmitted directly to component base and carry out storage and version management etc., and according to its Outside Access address update pair in component base
The data attribute informations such as the location for the Artifact instances of ontology answered.The interface of Integrated Development Environment be divided into architecture,
Characteristic model, design a model, Code Edit, mode binding, information description etc. multiple views and editing area.Mode binding view
To support to indicate architecture component in characteristic model and the correspondence role undertaken in designing a model, and in text edit area
Show that related source code etc. provides navigation feature.Tight association between each view and editing area can accomplish modification
See.
Field of intelligent control characteristic spectrum is extended characteristic model, and introduces ontology on its basis as meta-model
Description basis, introduce hypergraph also directed to the semantic multifarious feature of embedded intelligent controller and built towards different semantic hierarchies
Found unified formalized model and framework for representing knowledge.One characteristic spectrum is seven tuples, FM=(F, D, C, A, λD,λC,
λA), wherein F is the characteristic set in characteristics map;D is the characteristic relation set in characteristics map;C is in characteristics map
Feature non-functional constraint set, features the constraint condition that should meet when feature is selected;A is the mark sheet in characteristics map
The assets product set shown.Assets product is the specific implementation of feature, indicates the difference in product line and its Software Development
Stage realizes the possibility value of the software product of feature functionality, reflects the variability of feature implementation;λDIndicate that set F is arrived
The mapping of set D power set, i.e. λD: F → P (D), wherein P (D) indicates the power set of set D, and meets:AndλCIndicate set F to the mapping of set C power set, i.e. λC:F→
P (C), wherein P (C) indicates the power set of set C, and meets:And
Characteristic spectrum meta-model: feature and its incidence relation, non-functional constraint expression and component product are all abstracted
For meta-model basic element.Corresponding with the Web Ontology Language OWL that W3C recommends, meta-model modeling element is subdivided into ontology class
Element, object attribute element, data attribute element and data type element.Wherein ontology dvielement indicates the language in characteristics map
Adopted main body;Object attribute element characterizes the incidence relation between ontology dvielement in the form of the object properties of ontology dvielement,
Its domain and codomain are all ontology dvielements;Data attribute element describes the attribute information of ontology dvielement, and domain is
Ontology dvielement, codomain are data type element.
Characteristic spectrum meta-model based on ontology is as shown in Figure 1, feature (Feature), feature binding
(FeatureBind), feature association (Association), feature constraint (Constraint), pel component product
(Artifact) etc. it is defined as ontology class;And constraint object (restrictsObject), feature realize pel component product
(hasArtifact), role executes (playedBy), role includes that (hasRole) etc. is defined as Ontology attribute, is used for
Establish semantic subjective relationship net;Feature name (name), constraint expression (expression), component product reference (location) etc.
It is defined as the data attribute of corresponding ontology class, the characteristic attribute for descriptive semantics main body.
It is directed to the hypergraph model of knowledge mapping and manufacturing knowledge by using for reference, can establish the machine learning based on hypergraph and calculate
Method module knowledge representation formalized model: FMKRM=(MHG, MHG', φM,FMHG,FMHG',φF, Φ), in which: MHG is one
A Directed Hypergraph, MHG=(CM,λCM,RM,λRM,IM), the meta-model ontology basis Directed Hypergraph of referred to as FMKRM, in which: MHG's
Vertex set CMFor the meta-model ontology basic concept collection of FMKRM;The line set R of MHGMIt is general for the meta-model ontology basis of FMKRM
Read set of relations;The incident line set I of MHGMFor the meta-model ontology basis side the incident collection of FMKRM;MHG' is one
Directed Hypergraph, MHG'=(XM,λXM,MM,λMM,IM'), the meta-model Ontology of referred to as FMKRM jumps hypergraph, in which: MHG'
Vertex set XMFor the meta-model ontology element collection of FMKRM, and XM=CM∪RM∪IM;The line set M of MHG'MFor FMKRM's
Meta-model this volume elements set of relations;The incident line set I of MHG'M' it is the side the meta-model of FMKRM this volume elements incident collection;
CM, RM, IM, MM, IM' meetIn MHG', only allow that there are a connection vertex
With while incident while, that is, meet:|λXM(x)∩λMM(m)|≤1。φM:MM→CMIt is MHG
Characteristics map meta-model Ontology jumps function set, indicates from this volume elements of meta-model set of relations to ontology basic concept collection
Mapping, which is injection, i.e.,There is φM(m)≠φM(m');x∈XMIf x with
M is incident relationship in MHG', then the meta-model Ontology of m jumps result φM(m), it is not belonging to the connected subgraph of x
MHG [x]:IfThen
The actual vector that machine learning algorithm pel component is realized as module is stand-alone development and participates in the basic of multiplexing
Unit can be combined acquisition as the encapsulation class to object obtained by realization signature analysis.Object has determined the number in component
According to and operation, the affiliated layer of structure of feature and Mutability analysis component design and interactive mode has then been determined.It is developed based on feature
Component mainly include three types: complete the basic component of atomic function, correspondence corresponding to the leaf feature in characteristic model
The framing component of lower level operations interaction flow control is realized in the feature for possessing exploded relationship and corresponding to the spy comprising variant
Sign provides the alterable features component of variant selection.
Basic component is mainly used for realizing the basic function that leaf feature externally describes, including control important in control field
Process, controlled entity and relevant control rule processed etc..Based on the control logic process object that signature analysis determines, control member function
It can be expressed as one group of relevant operation, and show as component interface and service is externally provided.OWL language is generallyd use to control at present
Use-case model and controlled entity model processed is described, and the aggregation of use-case and class is carried out according to the principle of high cohesion lower coupling, and
The interface of control member is designed on this basis, and then obtains design specification with complete design process.
Framing component is not directly handled control logic, but sub to realizing as the encapsulation of schedule management object
The component operating interactive process of feature is controlled.The function services interface that framing component externally provides is the behaviour of subcharacter component
It combines, will be forwarded to correlator characteristic members for the calling of framing component and handled, and sub- characteristic members are operated and are handed over
Mutually is coordinated and dispatched.Such as Fig. 2, framing component design can usually be realized using a variety of design patterns, such as synthesis model, door
Surface model, template method mode etc..
For alterable features component, since embedded controller software design must take into consideration how the selection for making alterable features
Influence to other feature is as small as possible, so that the changeability of feature is confined in one or a few component, therefore in mould
In block modeling process, realizes the component of public character feature and realize that the component of changeability feature must be separated to improve pel
Reusability of the component product in Product Family development process.For the influence localization for generating variability, variable spy is realized
The modular structure of sign is needed changing features Information hiding.The modular structure design of the corresponding object of encapsulation alterable features is answered
There is provided unified interface externally to adapt to all variants of character pair, or different to hide using the mould plate technique of parametrization
Optinal plan.Feature can be used for determining template parameter, be instantiated by the selection of parameter attribute value to module.
For the variable characteristic for externally hiding alternative feature, component design scheme shown in Fig. 3 can be used, it will
The common portion for the object that alternative feature determines is modeled as interface component (Interface component), and difference
Part is then encapsulated as concrete implementation component (Concrete component).Wherein client element (client component)
It is to have to use or realize that the characteristic members relied on are realized with the alterable features being modeled.Plant component (Factory
Component it) is instantiated using getInstance method and a specific implementation component for returning to interface component gives client's structure
Part, client element can access to specific component by the abstraction interface of interface component.This implementation allow by
Alternative feature is tied to software product at runtime and has an impact without the other parts to product.If
BuildTime is bound, then interface component can be designed as using specific component or its feature realized as actual parameter
Parameterize component.Therefrom it can also be seen that generalization/specialization and implemented-by feature structure
Relationship can be called respectively by inheritance mechanism and interface and be realized.
Or feature can be used parameterized template and realize that variability is hidden.Wherein proxy component (Proxy Component)
The dependence of the use to or feature is encapsulated, and the request of client element is transmitted to the specific structure for the object that encapsulation or feature determines
Part.As shown above, the character pair client that client element Clientcomponentl is realized is passed to as parameter
Proxy component simultaneously calls its operation method, and operation method finds realization feature relevant to client use dependence
Specific component collection (Concrete componentF2 and Concretecomponent F3), and forward a request to them.
Due to concealing the specific component of respond request to client element, thus the changeability of or feature will not have an impact outside.
The above description is only an embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or the equivalent process variation that bright specification and accompanying drawing content are done, are applied directly or indirectly in other the relevant technologies
Field is similarly included in the invention patent protection scope.
Claims (8)
1. a kind of intelligent controller machine learning algorithm modular form describing framework and packaging method based on ontology, feature
It is, comprising:
Formalized description frame based on ontology;
Field of intelligent control characteristic spectrum;And
Machine learning algorithm pel component base.
2. according to the method described in claim 1, it is characterized by: the formalized description frame based on ontology uses unification
Formalized model unified formalized description is modeled and provided to the semantic information of different semantic hierarchies, realize each semanteme
Jumping between level constitutes criss-cross semantic structure, to meet the semantic multifarious expression of deep learning algoritic module
Demand.
3. according to the method described in claim 2, it is characterized by: the field of intelligent control characteristic spectrum to characteristic model into
Row extension, and description basis of the ontology as meta-model is introduced on its basis, it is semantic more also directed to embedded intelligent controller
The characteristics of sample, introduces hypergraph and establishes unified formalized model and framework for representing knowledge towards different semantic hierarchies.
4. method according to claim 1-3, it is characterised in that: each characteristic spectrum is one seven yuan
Group, FM=(F, D, C, A, λD,λC,λA), wherein
F is the characteristic set in characteristics map;
D is the characteristic relation set in characteristics map;
C is the feature non-functional constraint set in characteristics map, features the constraint condition that should meet when feature is selected;
A is the assets product set of the character representation in characteristics map;
λDIndicate set F to the mapping of set D power set, i.e. λD: F → P (D), wherein P (D) indicates the power set of set D, and full
Foot:And
λCIndicate set F to the mapping of set C power set, i.e. λC: F → P (C), wherein P (C) indicates the power set of set C, and full
Foot:And
5. according to the method described in claim 4, expression exists it is characterized by: the assets product is the specific implementation of feature
The different phase of product line and its Software Development realizes the possibility value of the software product of feature functionality, reflects feature reality
The variability of existing scheme.
6. according to the method described in claim 1, it is characterized by: the field of intelligent control characteristic spectrum is by the field of extension
Description basis of the characteristic model as embedded intelligence control field inner member business semantics, generates initially in an automated way
Embedded intelligence control field conceptual architecture.
7. according to the method described in claim 1, it is characterized by: the encapsulation and description of the machine learning algorithm pel component
The transition designed using component semantics as feature to pel component, function distribution to each machine learning algorithm pel component and
Interaction semantics are defined, the basis for encapsulating and describing as associated components.
8. according to the method described in claim 1, characterized by further comprising: by use for reference be directed to knowledge mapping and manufacturing knowledge
Hypergraph model, can establish the machine learning algorithm module knowledge representation formalized model based on hypergraph: FMKRM=(MHG,
MHG',φM,FMHG,FMHG',φF, Φ), in which:
MHG is a Directed Hypergraph, MHG=(CM,λCM,RM,λRM,IM), the meta-model ontology basis Directed Hypergraph of referred to as FMKRM,
Wherein: the vertex set C of MHGMFor the meta-model ontology basic concept collection of FMKRM;The line set R of MHGMFor the meta-model of FMKRM
Ontology basic concept set of relations;The incident line set I of MHGMFor the meta-model ontology basis side the incident collection of FMKRM;
MHG' is a Directed Hypergraph, MHG'=(XM,λXM,MM,λMM,IM'), the meta-model Ontology of referred to as FMKRM jumps hypergraph,
Wherein:
The vertex set X of MHG'MFor the meta-model ontology element collection of FMKRM, and XM=CM∪RM∪IM;The line set M of MHG'MFor
The meta-model of FMKRM this volume elements set of relations;The incident line set I of MHG'M' it is this volume elements of the meta-model of FMKRM
The side incident collection;
CM, RM, IM, MM, IM' meet
In MHG', only allow there are one connect vertex and while incident while, that is, meet:|
λXM(x)∩λMM(m)|≤1;
φM:MM→CMIt is that the characteristics map meta-model Ontology of MHG jumps function set, indicates from meta-model this volume elements relationship
Collect the mapping of ontology basic concept collection, which is injection, i.e.,There is φM(m)≠φM
(m');x∈XMIf x and m are incident relationship in MHG', the meta-model Ontology of m jumps result
φM(m), it is not belonging to the connected subgraph MHG [x] of x:IfThen
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910136456.3A CN109933307A (en) | 2019-02-18 | 2019-02-18 | A kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910136456.3A CN109933307A (en) | 2019-02-18 | 2019-02-18 | A kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109933307A true CN109933307A (en) | 2019-06-25 |
Family
ID=66985889
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910136456.3A Pending CN109933307A (en) | 2019-02-18 | 2019-02-18 | A kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109933307A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111160564A (en) * | 2019-12-17 | 2020-05-15 | 电子科技大学 | Chinese knowledge graph representation learning method based on feature tensor |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060265699A1 (en) * | 2005-05-20 | 2006-11-23 | Sun Microsystems, Inc. | Method and apparatus for pattern-based system design analysis using a meta model |
CN105426628A (en) * | 2015-12-10 | 2016-03-23 | 中南大学 | Complex product heterogeneous model fusing method |
US20170262996A1 (en) * | 2016-03-11 | 2017-09-14 | Qualcomm Incorporated | Action localization in sequential data with attention proposals from a recurrent network |
CN108189043A (en) * | 2018-01-10 | 2018-06-22 | 北京飞鸿云际科技有限公司 | A kind of method for inspecting and crusing robot system applied to high ferro computer room |
-
2019
- 2019-02-18 CN CN201910136456.3A patent/CN109933307A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060265699A1 (en) * | 2005-05-20 | 2006-11-23 | Sun Microsystems, Inc. | Method and apparatus for pattern-based system design analysis using a meta model |
CN105426628A (en) * | 2015-12-10 | 2016-03-23 | 中南大学 | Complex product heterogeneous model fusing method |
US20170262996A1 (en) * | 2016-03-11 | 2017-09-14 | Qualcomm Incorporated | Action localization in sequential data with attention proposals from a recurrent network |
CN108189043A (en) * | 2018-01-10 | 2018-06-22 | 北京飞鸿云际科技有限公司 | A kind of method for inspecting and crusing robot system applied to high ferro computer room |
Non-Patent Citations (1)
Title |
---|
李弈远: "基于特征的软件产品线开发关键技术研究", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111160564A (en) * | 2019-12-17 | 2020-05-15 | 电子科技大学 | Chinese knowledge graph representation learning method based on feature tensor |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8527943B1 (en) | System and method of application development | |
Bakshi et al. | The abstract task graph: a methodology for architecture-independent programming of networked sensor systems | |
Hilaire et al. | Formal specification and prototyping of multi-agent systems | |
US6976020B2 (en) | Software composition using graph types, graph, and agents | |
US7827527B1 (en) | System and method of application development | |
Breu et al. | Systems, views and models of UML | |
CN105210054B (en) | Binding of the data source to composite control | |
Doroshenko et al. | A rewriting framework for rule-based programming dynamic applications | |
Aliyu et al. | The high level language for system specification: A model-driven approach to systems engineering | |
Jézéquel | Model-driven engineering for software product lines | |
CN107851001A (en) | For the method for the computer application that software-type application program is shown based on design specification | |
Ermel et al. | Visual modeling of controlled emf model transformation using henshin | |
Rouhi et al. | Towards a formal model of patterns and pattern languages | |
CN109933307A (en) | A kind of intelligent controller machine learning algorithm modular form description and packaging method based on ontology | |
Shaalan et al. | KROL: a knowledge representation object language on top of Prolog | |
Lochmann et al. | An integrated view on modeling with multiple domain-specific languages | |
Thalheim | From Models_For_Programming to Modelling_To_Program and Towards Models_As_A_Program | |
Krijt et al. | Automated dynamic formation of component ensembles-taking advantage of component cooperation locality | |
Bezivin | Model Engineering for Software Modernization. | |
Srbljić et al. | Programming languages for end-user personalization of cyber-physical systems | |
Borland | Transforming statechart models to DEVS | |
Loja et al. | A business process metamodel for enterprise information systems automatic generation | |
Bakshi et al. | Architecture-independent programming for wireless sensor networks | |
Argente et al. | Describing agent organisations | |
Barbier et al. | Business Components |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190625 |