CN111857680A - Intelligent model construction method based on meta-characteristics - Google Patents

Intelligent model construction method based on meta-characteristics Download PDF

Info

Publication number
CN111857680A
CN111857680A CN202010516304.9A CN202010516304A CN111857680A CN 111857680 A CN111857680 A CN 111857680A CN 202010516304 A CN202010516304 A CN 202010516304A CN 111857680 A CN111857680 A CN 111857680A
Authority
CN
China
Prior art keywords
model
intelligent
description file
meta
source code
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
CN202010516304.9A
Other languages
Chinese (zh)
Inventor
温研
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.)
Beijing Linzhuo Information Technology Co Ltd
Original Assignee
Beijing Linzhuo Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Linzhuo Information Technology Co Ltd filed Critical Beijing Linzhuo Information Technology Co Ltd
Priority to CN202010516304.9A priority Critical patent/CN111857680A/en
Publication of CN111857680A publication Critical patent/CN111857680A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F8/00Arrangements for software engineering
    • G06F8/30Creation or generation of source code
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Stored Programmes (AREA)

Abstract

The invention discloses an intelligent model construction method based on meta-characteristics, which adopts an intelligent model description method based on meta-characteristics to describe and abstract various key characteristics of an intelligent model, and realizes the reconstruction of the intelligent model by setting parameter values in an established model description file.

Description

Intelligent model construction method based on meta-characteristics
Technical Field
The invention belongs to the technical field of machine learning, and particularly relates to an intelligent model construction method based on meta-characteristics.
Background
At present, artificial intelligence technology is widely applied to various fields of life. The key to the application of artificial intelligence technology is the design and development of machine learning models. Most of the existing technical methods for supporting the development of machine learning models define a set of design framework, and developers call interfaces in the framework to realize the internal calculation of the models. The mainstream design framework includes Caffe, tenserflow, pitorch and thano, and the most similar to the present invention is Caffe design framework. Caffe adopts a method of representing and realizing separation to realize the design and operation of a deep neural network model, uses a Protocl Buffer to define a model file, uses a special text file prototxt to represent a network structure, and constructs a learning network in a directed acyclic graph mode. The basic flow of Caffe includes: preparing data, compiling a network structure file, compiling a network solution file, starting training and predicting by using a trained model.
The problems of various design frame interfaces, complex interface parameters, high learning starting point of the frame and the like can be faced when the prior art is used for intelligent model design. Generally, in order to realize an intelligent model with a certain function, a complete set of design framework needs to be mastered, so that the development cost of the intelligent model is high, the learning threshold is high, the development efficiency is low, and common program developers are forbidden to machine learning technology.
Disclosure of Invention
In view of this, the invention provides an intelligent model construction method based on meta-characteristics, which can simply and efficiently complete the construction of an intelligent model.
The invention provides an intelligent model construction method based on meta-characteristics, which comprises the following steps:
step 1, extracting meta-features from a model source code of an intelligent model, wherein the meta-features are feature information of the intelligent model; generating a model description file of the intelligent model based on the meta-features; adding corresponding parameter marks in the model source codes according to the algorithm parameter characteristic information recorded in the model description file to form marked model source codes;
step 2, when in use, selecting a marking model source code and a model description file of an intelligent model to be constructed from the series of marking model source codes and model description files formed in the step 1, and setting algorithm parameters in the model description file to complete the configuration of the model description file; and modifying the marking model source code according to the configured model description file to obtain the final marking model source code of the intelligent model to be constructed, and finishing the construction of the intelligent model to be constructed.
Furthermore, the generation and configuration processes of the model description file are realized in a visualization mode.
Has the advantages that:
1. according to the intelligent model design method based on the meta-characteristics, various key characteristics of the intelligent model are described and abstracted, and the reconstruction of the intelligent model is realized by setting parameter values in the established model description file, under the framework, developers do not need to understand complex knowledge background, and only need to configure proper meta-characteristic state based on the understanding of the model description file, so that the intelligent model design and development meeting different business requirements can be completed, the development efficiency of intelligent application is effectively improved, the learning cost in development is reduced, and the optional range of the intelligent model in intelligent application development is expanded;
2. the invention adopts a visual mode to realize the generation and the configuration of the model description file, effectively improves the efficiency and the usability of the intelligent model construction method, and lays a foundation for the rapid development of the intelligent model.
Drawings
FIG. 1 is a schematic diagram of a model description file of an intelligent model construction method based on meta-features according to the present invention.
FIG. 2 is a schematic diagram of a source code marking method of the intelligent model construction method based on meta-features provided by the invention.
Detailed Description
The invention is described in detail below by way of example with reference to the accompanying drawings.
The invention provides an intelligent model construction method based on meta-characteristics, which has the basic idea that: extracting meta-features in the source code of the intelligent model to form a model description file, marking the source code of the intelligent model according to the model description file to form a marked model source code of the intelligent model, selecting the marked model source code and the model description file as required, configuring the model description file, modifying the corresponding marked model source code according to the configured model description file, and finally forming the required intelligent model.
The invention provides an intelligent model construction method based on meta-characteristics, which specifically comprises the following steps:
step 1, extracting meta-features from a model source code of an intelligent model; generating a model description file of the intelligent model based on the meta-features; and adding corresponding parameter marks in the model source codes according to the algorithm parameter characteristic information recorded in the model description file to form marked model source codes.
The model source code of the intelligent model can be written by itself or the existing source code can be selected. The meta-feature in the invention refers to feature information of the intelligent model, is a feature abstraction of the intelligent model, and covers various kinds of feature information of algorithm parameters, development languages, development tools, hardware systems, operating systems and the like of the related intelligent model. Then, a model description file is generated according to the meta-features extracted from the model source code. And finally, adding parameter marks in the model source codes according to the algorithm parameter characteristic information recorded in the model description file, wherein the parameter marks are used for matching the accurate positions of the parameters in the source code file when the parameter values are replaced.
In addition, in order to further improve the efficiency and the usability of the intelligent model construction method, the generation process of the model description file can be realized in a visual mode, namely, the definition interface of the intelligent model description file is constructed in combination with the definition of the system UI library and the meta-characteristics to complete the generation process of the intelligent model description file. In a specific application, the model description file and the markup model source code can be packaged and placed in the same computer directory.
Step 2, when in use, selecting a marking model source code and a model description file of an intelligent model to be constructed from the series of marking model source codes and model description files formed in the step 1, and setting algorithm parameters in the model description file to complete the configuration of the model description file; and modifying the marking model source code according to the configured model description file to obtain the final marking model source code of the intelligent model to be constructed, and finishing the construction of the intelligent model to be constructed.
Specifically, the developer selects the model to be used from the computer directory set in step 1, analyzes the description file of the model, generates a model parameter configuration interface, and after configuration, the developer configures the model in the configuration interface, and the model configuration interface replaces the configuration value into the marked model source code corresponding to the model description file, so as to complete the design of the intelligent model.
Example (b):
in this embodiment, the method for constructing an intelligent model based on meta-features provided by the present invention is used to implement rapid development of a model, and the specific process includes the following steps:
step 1.1, writing the realization source code of the machine learning algorithm, wherein the realization source code of the DBSCA algorithm is realized by using python language in the example. Model parameters are collected from source codes of the DBSCA algorithm according to the meta-features, and a model description file is generated, wherein the model description file is recorded by using an xml file in the embodiment, and the file content is as shown in fig. 1 below. The model description file in this embodiment includes the following aspects:
model name: name of model, displayed model selection list.
The model author: name of the author of the model.
Model support: the supporting field of the model.
Using the frame: and the intelligent framework used in the model comprises a development language, a development tool, a hardware system and an operating system.
Model description: and describing functions and other aspects of the model, including algorithm parameters.
Visualization: whether the model supports visualization.
Distributed: whether the model supports distributed deployment.
Parameter tag names to match tags in the model source code to facilitate the replacement of parameter values into the model source code.
Parameter display name, the name to display onto the model configuration interface.
Whether or not it is necessary: indicating whether the parameter must be filled in at model configuration
Description of the parameters: the text used for describing the function of the parameter.
Parameter values: and a plurality of legal parameter values are provided for the user to select. If the parameter value for selection does not exist, the user can freely input the parameter value.
Then, the source codes of the BDSCAN algorithm are marked, the marking mode used in this embodiment is as shown in fig. 2, and the '<% >' mark is distinguished from the normal source codes, and the content in the '<% >' mark is the name of the parameter mark and is used for replacing the parameter value.
The model description file and the markup model source code are packaged and placed in a computer directory, and the directory structure of the model package in this example is as follows:
model warehouse inventory
Model name | - - - - - - - - - -, model name
Model (catalog of model template source code)
Properties (model description file)
| | - - - - - - - -param.md (model parameter description file)
And step 1.2, in use, a developer selects a model to be used from the catalog generated in the step 1.1, analyzes a parameter file of the model and generates a model parameter configuration interface. In this embodiment, a jdom library is used to analyze the model parameter file, and an swtu library is used to generate a parameter configuration interface according to the analyzed model parameters. And after the developer finishes the configuration of the parameter configuration interface, the configuration interface replaces the parameter values into the model source code file to complete the design of the intelligent model.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (2)

1. The intelligent model construction method based on the meta-characteristics is characterized by comprising the following steps:
step 1, extracting meta-features from a model source code of an intelligent model, wherein the meta-features are feature information of the intelligent model; generating a model description file of the intelligent model based on the meta-features; adding corresponding parameter marks in the model source codes according to the algorithm parameter characteristic information recorded in the model description file to form marked model source codes;
step 2, when in use, selecting a marking model source code and a model description file of an intelligent model to be constructed from the series of marking model source codes and model description files formed in the step 1, and setting algorithm parameters in the model description file to complete the configuration of the model description file; and modifying the marking model source code according to the configured model description file to obtain the final marking model source code of the intelligent model to be constructed, and finishing the construction of the intelligent model to be constructed.
2. The method of claim 1, wherein the generation and configuration of the model description file are performed visually.
CN202010516304.9A 2020-06-09 2020-06-09 Intelligent model construction method based on meta-characteristics Pending CN111857680A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010516304.9A CN111857680A (en) 2020-06-09 2020-06-09 Intelligent model construction method based on meta-characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010516304.9A CN111857680A (en) 2020-06-09 2020-06-09 Intelligent model construction method based on meta-characteristics

Publications (1)

Publication Number Publication Date
CN111857680A true CN111857680A (en) 2020-10-30

Family

ID=72987497

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010516304.9A Pending CN111857680A (en) 2020-06-09 2020-06-09 Intelligent model construction method based on meta-characteristics

Country Status (1)

Country Link
CN (1) CN111857680A (en)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1176482A1 (en) * 2000-07-27 2002-01-30 Abb Research Ltd. Method and computer program for generating a regulation or control system
CN101004680A (en) * 2006-11-23 2007-07-25 福建顶点软件股份有限公司 Flexible, fast software development method and support system by using kernels of direct operation object model definition
US20080077377A1 (en) * 2004-07-29 2008-03-27 Wolfgang Roesner Method, system and program product supporting presentation of a simulated or hardware system including configuration entities
US20090125878A1 (en) * 2007-11-05 2009-05-14 Cullum Owen H G System and Method for Generating Modified Source Code Based on Change-Models
CN104978170A (en) * 2014-04-03 2015-10-14 中国科学院软件研究所 Multi-agent system generating method based on graphical expression
CN106250164A (en) * 2016-08-16 2016-12-21 广州仕邦人力资源有限公司 A kind of code generating method based on requirement documents and device
US20170351511A1 (en) * 2015-12-22 2017-12-07 Opera Solutions Usa, Llc System and Method for Code and Data Versioning in Computerized Data Modeling and Analysis
CN110190984A (en) * 2019-04-29 2019-08-30 中国电力科学研究院有限公司 A kind of efficient power communication system large scale emulation scene modeling method and system
CN110795077A (en) * 2019-09-26 2020-02-14 北京你财富计算机科技有限公司 Software development method and device based on artificial intelligence and electronic equipment
EP3629245A1 (en) * 2018-09-28 2020-04-01 Koninklijke Philips N.V. Intelligent agent framework

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1176482A1 (en) * 2000-07-27 2002-01-30 Abb Research Ltd. Method and computer program for generating a regulation or control system
US20080077377A1 (en) * 2004-07-29 2008-03-27 Wolfgang Roesner Method, system and program product supporting presentation of a simulated or hardware system including configuration entities
CN101004680A (en) * 2006-11-23 2007-07-25 福建顶点软件股份有限公司 Flexible, fast software development method and support system by using kernels of direct operation object model definition
US20090125878A1 (en) * 2007-11-05 2009-05-14 Cullum Owen H G System and Method for Generating Modified Source Code Based on Change-Models
CN104978170A (en) * 2014-04-03 2015-10-14 中国科学院软件研究所 Multi-agent system generating method based on graphical expression
US20170351511A1 (en) * 2015-12-22 2017-12-07 Opera Solutions Usa, Llc System and Method for Code and Data Versioning in Computerized Data Modeling and Analysis
CN106250164A (en) * 2016-08-16 2016-12-21 广州仕邦人力资源有限公司 A kind of code generating method based on requirement documents and device
EP3629245A1 (en) * 2018-09-28 2020-04-01 Koninklijke Philips N.V. Intelligent agent framework
CN110190984A (en) * 2019-04-29 2019-08-30 中国电力科学研究院有限公司 A kind of efficient power communication system large scale emulation scene modeling method and system
CN110795077A (en) * 2019-09-26 2020-02-14 北京你财富计算机科技有限公司 Software development method and device based on artificial intelligence and electronic equipment

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
宋春磊;高博;李婉婉;赵智杰;谢宗甫;: "软件无线电平台可视化应用管理系统设计与实现", 信息工程大学学报, no. 01, pages 66 - 69 *
晏华;陈昊;郭宣佑;: "一种面向汽车电子的配置界面动态生成方法", 计算机科学, no. 08, pages 178 - 181 *

Similar Documents

Publication Publication Date Title
US7814427B2 (en) Object model tree diagram
Vanderdonckt et al. Flexible reverse engineering of web pages with vaquista
Ahmed et al. Model-based user interface engineering with design patterns
US20050283758A1 (en) Bi-directional programming system/method for program development
EP2228726A2 (en) A method and system for task modeling of mobile phone applications
CN112199086A (en) Automatic programming control system, method, device, electronic device and storage medium
JP2005196291A (en) User interface application development program and development device
US20110083120A1 (en) Modeling and linking documents for packaged software application configuration
CN104731588B (en) Page layout document generating method and page layout file creating apparatus
CN104407863A (en) Abstract control model programming device and method
US8234570B2 (en) Harvesting assets for packaged software application configuration
KR101408280B1 (en) Apparatus and method for logic creation of development tools, and storage medium recording program for implementing method thereof
Lumertz et al. User interfaces metamodel based on graphs
US20190188297A1 (en) Automatic core data service view generator
CN110018827A (en) Method and device for automatically generating code, electronic equipment and readable storage medium
Jácome-Guerrero et al. Software development tools in model-driven engineering
CN108073511A (en) test code generating method and device
Wimmer et al. On the integration of web modeling languages: Preliminary results and future challenges
AU2021263673A1 (en) A system and method for dynamically defining digital forms
CN112631585A (en) XML-based rapid parameter interface configuration method
Pederiva et al. The beautification process in model-driven engineering of user interfaces
JP4836412B2 (en) Method and apparatus for generating a graphical user interface
CN111857680A (en) Intelligent model construction method based on meta-characteristics
CN112764637B (en) Intelligent agriculture Internet of things equipment description file generation method
Beaudoux et al. Specifying and running rich graphical components with loa

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: 20201030