CN109359701A - A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification - Google Patents
A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification Download PDFInfo
- Publication number
- CN109359701A CN109359701A CN201811434379.1A CN201811434379A CN109359701A CN 109359701 A CN109359701 A CN 109359701A CN 201811434379 A CN201811434379 A CN 201811434379A CN 109359701 A CN109359701 A CN 109359701A
- Authority
- CN
- China
- Prior art keywords
- node
- label
- data
- dimensional modeling
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The three-dimensional modeling data analytic method of a kind of extracted with high accuracy and Fast Classification is claimed in the present invention, belongs to technical field of computer information processing.This method comprises: being directed to threedimensional model, a kind of data parsing scheme based on binary tree is designed, the rubidium marking of search model analyzes the nest relation of data, forms most simple document.When receiving the model of a three-dimensional dae format, reading model data, a kind of parser process is provided, pass through each rubidium marking of search model, gradually analyze the nest relation of data, repeated and redundant data are rejected, improvement extreme learning machine is reused and classifies, ultimately form most simple document.The present invention is by binary tree and improves extreme learning machine simultaneously using parsing three-dimensional modeling data, reduces the stacking of redundant data, accelerates the processing speed of data, improves analyzing efficiency.
Description
Technical field
The invention belongs to technical field of computer information processing more particularly to three-dimensional modeling data analytic methods.
Background technique
Three-dimensional data model is a kind of partly-structured data, and user can arbitrarily set rubidium marking and embedding between them
Set relationship, by taking dae as an example, the data pattern of dae file is a very important information, only obtains a dae file
Data pattern, be just conducive to the real meaning for identifying file, mentioned for the work such as subsequent data classification, cluster and data mining
Good basis is supplied.
Up to the present, many researchs propose various algorithms for the extraction of three-dimensional modeling data,
Important thinking is provided for correlative study.Jun-Ki Min et al. proposes a kind of extracting method based on element content model,
This method limits element content model, and in the content model of element, daughter element can only occur once.Make in this method
With bottom-up method, the mode of daughter element is first obtained, the mode of father's element is proposed further according to the mode of daughter element.
SvetlozarNestorov et al. proposes the method that pattern information is extracted from semi-structured data.The oriented label of this method
Figure describes semi-structured data, uses the semantical definition semi-structured data of the greatest fixpoint of unitary datalog program
Data type.But these algorithms are there is also corresponding disadvantage, major embodiment both ways: it is 1. inaccurate.2. algorithm is complicated
Degree is high.Using fancy grade regular expression no doubt available very accurate mode, but regular expression generates and gives system band
Carry out very big expense.This is largely because the semi-structured feature of its language, lacks its mode mandatory, gives data
Extraction work bring very big difficulty.And binary tree is a kind of mode of compromise, even when binary tree step number is larger,
It still can accurately obtain gross data.Therefore data are extracted using a kind of algorithm of binary tree herein, it is very big for complexity
Threedimensional model file, its effective data can be obtained rapidly；Meanwhile for there is situations such as optional and repeating label in document,
Also it can provide and accurately differentiate and handle.
And on the inducing classification that model proposes data, also there is numerous research, traditional machine learning classification model has
Decision tree, Bayes, artificial neural network, support vector machines etc., but there are dimension disaster and training time are longer for these models
Etc. drawbacks.And extreme learning machine algorithm, the connection weight of input layer and implicit interlayer and the threshold of hidden layer neuron is randomly generated
Value, and the number of hidden layer neuron need to be only set, unique optimal solution can be obtained without adjustment in the training process, with
Traditional neural network algorithm is compared, and extreme learning machine method pace of learning is fast, Generalization Capability is good.Finally, y-bend is used herein
The method that tree is used together with improvement extreme learning machine, parses threedimensional model file, exports document.
Summary of the invention
Present invention seek to address that the above problem of the prior art.A kind of stacking for reducing redundant data is proposed, is accelerated
The processing speed of data, improves the extracted with high accuracy of analyzing efficiency and the three-dimensional modeling data analytic method of Fast Classification.This
The technical solution of invention is as follows:
A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification comprising following steps:
1) three-dimensional modeling data, is read first, extracts all elements label in three-dimensional modeling data, the element mark
Note includes beginning label and terminates label, and is indicated with a label node；Will label node in order one by one number to
Generate label chain；
2) binary tree node model, is established, the label chain obtained by step 1) and the number for wherein marking node will
The beginning label and end label of same name match, while their corresponding label nodes are paired into one group, then root
The nest relation of each element label is analyzed according to the number numerical values recited for comparing label node；
3), extreme learning machine is neural networks with single hidden layer, improves extreme learning machine in this algorithm and uses bilayer model, accelerates
Training speed, then use this improvements extreme learning machine method, training step 2) in binary tree node model judge and delete
Except the redundant data in label binary tree, to obtain the most simple document of threedimensional model file.
Further, the step 1) is that element mark all in three-dimensional modeling data is extracted by automatic state machine
Note, the rubidium marking includes beginning label and terminates label, and resulting rubidium marking is indicated by chained list node.
Further, described to indicate resulting rubidium marking by chained list node, it specifically includes:
Beginning and end labels all in document are extracted first, and with a label node Element Node
It indicates, the format of the label node are as follows: (char Elemstring, intNum, int Match Num, Element Node*
Next), wherein Elem String is used to recording mark character string；Num indicates number of the current markers node in chained list；
Match Num is indicated: if current node is the beginning label of element, recording the end mark to match in chained list with current node
Remember the number of node；If node itself is to terminate label, this is set to sky；Next is directed to the pointer of next node.
Further, the step 2) establish the step of binary tree node model include: the binary tree node model by
Node String, Tree Node*Left Child, Tree-Node*Right Child three classes form, wherein Node
String indicates the tab character string that current beginning label node is included；Left Child: left subtree node pointer, if worked as
The Left Child of preceding node is not empty, then it represents that the node that Left Child is directed toward is nested in current node, their fathers each other
Subrelation；Right Child: right subtree node pointer, if the Right Child of current node is not empty, then it represents that Right
The node and current node that Child is directed toward are nested in the same label, their brotherhoods each other.
Further, the nest relation of step 2) the analysis each element label, rule have following two o'clock: 1. identifying root
Element；Maximum nodal value in label chain is found out, identifies that maximum nodal value is root element node；2. leading to non-root element
The mobile pointer for crossing traversal tracking chained list, to verify the nest relation of model data.
Further, the step 3) rejects the repeated and redundant of data using the method for node comparison, specifically includes: passing through
All nodes and the right child node of its non-empty are compared, without duplicate label if label is not identical, if identical
Can also be there are two types of different situations: (1) tree using current node as root be identical with using its right child node as the tree of root: indicating two
A rubidium marking is identical, the right child node of current node directly can be set to sky to delete this redundant marks；And it will work as
Preceding node identification is repeatable label；(2) the tree ratio using current node as the tree of root and using its right child node as root is relatively incompletely
It is identical, it indicates that different optional labels are nested in current node label or its right child node label, is then deleting in this case
Before the right child node of current node, the label merging for being nested in right child node but be not nested in current node is tied to current
In the child node of point, it is made to be nested in current node.
Further, the step 3) obtains the most simple document of threedimensional model file using extreme learning machine method is improved,
Including being divided into training stage and forecast period, this purpose of model is to utilize training sample set between input variable and classification results
Establish a kind of mapping relations；
Corresponding structured vectors is defined as: dsv=< d1,d2,…,dn>, n is characterized of the corresponding feature vector of term
Number；Wherein diFor the corresponding feature vector of featured terms of i-th of threedimensional model file:Wherein TF is word frequency, and IDF is inverse document frequency, ωi
For i-th of term, Doc is threedimensional model file, and m is node number, εjFor node, ejIndicate corresponding unit vector；
Specific step is as follows:
Input: training set FD, test set FT, activation primitive g (x)；Concealed nodes number L (L≤T)；Kernel function K, parameter C and
γ randomly chooses input weight WiWith offset bi。
Output: classification results XTAnd YT。
Step 2.1: then random selection training sample first respectively pre-processes corresponding threedimensional model file,
The vector for obtaining threedimensional model file indicates.Finally, quickly establishing disaggregated model by the improvement extreme learning machine method proposed.
Step 2.2: conversion FD to vector model MS-VSM obtains original training data collection D.
Step 2.3: conversion FT to vector model MS-VSM obtains original training data collection T.
Step 2.4: using D as new training set, to reduce trained mistake, carrying out feature extraction, obtain eigenmatrix
X is calculated classification results X by XT。
Step 2.5: repeating the above training step and obtain parameter L, the optimal value of C, γ.
Step 2.6: using T as new training set, to reduce trained mistake, carrying out feature extraction, obtain eigenmatrix
Y is calculated classification results Y by YT。
Final integrated data extracts and inducing classification, provides final document.
It advantages of the present invention and has the beneficial effect that:
The present invention extracts data using a kind of algorithm of binary tree, the threedimensional model file very big for complexity, energy
Its effective data is obtained rapidly；Meanwhile for there is situations such as optional and repeating label in document, it can also provide and accurately sentence
Not and handle.
Then extreme learning machine algorithm is used, extreme learning machine method pace of learning is fast, Generalization Capability is good, is randomly generated defeated
Enter the connection weight of layer and implicit interlayer and the threshold value of hidden layer neuron, in the training process without adjustment, hidden layer is set
The number of neuron obtains unique optimal solution.
Finally, the method that the present invention is used together using binary tree and extreme learning machine, and extreme learning machine is improved, it adopts
With bilayer model, accelerate trained speed.Data extraction is carried out to threedimensional model file, inducing classification improves extraction data
Accuracy rate, and accelerate data classification, be finally completed parsing, export document.
Detailed description of the invention
Fig. 1 is that the present invention provides the overall realization frame of preferred embodiment；
Fig. 2, which gives, a kind of under the present invention to be extracted three-dimensional modeling data using binary tree and rejects repeated and redundant data
Flow chart；
Fig. 3 give under the present invention it is a kind of using improve extreme learning machine model propose Data induction classification method.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, detailed
Carefully describe.Described embodiment is only a part of the embodiments of the present invention.
The technical solution that the present invention solves above-mentioned technical problem is:
The content of Fig. 1 is exactly overall realization frame, and particular content is divided into data and extracts and data classification two.
When receiving a threedimensional model file, uses first and model file is changed into readable textual form.Using two
The method of fork tree extracts non-duplicate data, that is, the content of Fig. 2 from model.Steps are as follows
Step 1.1: identifying rubidium marking all in threedimensional model file.It is taken out by a simple automatic state machine
The all elements label in three-dimensional modeling data is taken out, the rubidium marking includes beginning label and terminates label, and with one
Node is marked to indicate；It numbers label node to generate label chain one by one in order.First by beginning all in document
It is extracted with label is terminated, and with label node Element Node (char an Elemstring, intNum, int
Match Num, Element Node*next) it indicates.Wherein, Elem String is used to recording mark character string；Num is indicated
Number of the current markers node in chained list；Match Num is indicated: if current node is the beginning label of element, recording chain
The number of the end label node to match in table with current node；If node itself is to terminate label, this is set to sky；
Next is directed to the pointer of next node.Nest relation in evaluation of markers chain between node establishes binary tree node model,
Model is made of Node String, Tree Node*Left Child, Tree-Node*Right Child three classes, wherein Node
String indicates the tab character string that current beginning label node is included；Left Child: left subtree node pointer, if worked as
The Left Child of preceding node is not empty, then it represents that the node that Left Child is directed toward is nested in current node, their fathers each other
Subrelation；Right Child: right subtree node pointer, if the Right Child of current node is not empty, then it represents that Right
The node and current node that Child is directed toward are nested in the same label, their brotherhoods each other.
Step 1.2: the nest relation between analysis gained label.It is compiled in order by label chain obtained in the previous step
Number, the nest relation of each element label is analyzed further according to the marker number with each label node pairing.Analysis model it is embedding
Set relationship, rule have following two o'clock: 1. identifying root element；Maximum nodal value in label chain is found out, identifies maximum section
Point value is root element node 2. to non-root element, by the mobile pointer of traversal tracking chained list, to verify the embedding of model data
Set relationship.
Step 1.3: rejecting repeated and redundant.The repeated and redundant of data is rejected using the method for node comparison.Can pass through by
All nodes and its right child node (if being not sky) are compared, without duplicate label if label is not identical, if
It is identical can yet be there are two types of different situations: (1) tree using current node as root is identical with using its right child node as the tree of root:
It indicates that two rubidium markings are identical, the right child node of current node directly can be set to sky to delete this redundant marks；
And current node is identified as repeatable label.(2) compare using current node as the tree of root and using its right child node as the tree of root
It is not exactly the same, indicate that different optional labels are nested in current node label (or its right child node label).Such case
Under then before the right child node for deleting current node, right child node will be nested in but is not nested in the label merging of current node
Into the child node of current node, it is made to be nested in current node.
After taking rejecting repeated and redundant data, Data induction is classified, here using the side for improving extreme learning machine
Method, that is, the content of Fig. 3.The implementation process of entire model can be divided into training stage and forecast period.This purpose of model
It is to establish a kind of mapping relations between input variable and classification results using training sample set.
Corresponding structured vectors is defined as: dsv=< d1,d2,…,dn>, n is characterized of the corresponding feature vector of term
Number；Wherein diFor the corresponding feature vector of featured terms of i-th of threedimensional model file:Wherein TF is word frequency, and IDF is inverse document frequency, ωi
For i-th of term, Doc is threedimensional model file, and m is node number, εjFor node, ejIndicate corresponding unit vector；
Specific step is as follows:
Input: training set FD, test set FT, activation primitive g (x)；Concealed nodes number L (L≤T)；Kernel function K, parameter C and
γ randomly chooses input weight WiWith offset bi。
Output: classification results XTAnd YT。
Step 2.1: then random selection training sample first respectively pre-processes corresponding threedimensional model file,
The vector for obtaining threedimensional model file indicates.Finally, quickly establishing disaggregated model by the improvement extreme learning machine method proposed.
Step 2.2: conversion FD to vector model MS-VSM obtains original training data collection D.
Step 2.3: conversion FT to vector model MS-VSM obtains original training data collection T.
Step 2.4: using D as new training set, to reduce trained mistake, carrying out feature extraction, obtain eigenmatrix
X is calculated classification results X by XT。
Step 2.5: repeating the above training step and obtain parameter L, the optimal value of C, γ.
Step 2.6: using T as new training set, to reduce trained mistake, carrying out feature extraction, obtain eigenmatrix
Y is calculated classification results Y by YT。
Final integrated data extracts and inducing classification, provides final document.
The above embodiment is interpreted as being merely to illustrate the present invention rather than limit the scope of the invention.?
After the content for having read record of the invention, technical staff can be made various changes or modifications the present invention, these equivalent changes
Change and modification equally falls into the scope of the claims in the present invention.
Claims (7)
1. the three-dimensional modeling data analytic method of a kind of extracted with high accuracy and Fast Classification, which comprises the following steps:
1) three-dimensional modeling data, is read first, extracts all elements label in three-dimensional modeling data, the rubidium marking packet
It includes beginning label and terminates label, each label is indicated with a label node；Label node is numbered one by one in order
To generate label chain；
2) binary tree node model, is established, the label chain obtained by step 1) and the number for wherein marking node will be same
Two subtrees that the beginning label and end label of title are made into binary tree, while their corresponding label nodes being paired into
One group, the nest relation of each element label is analyzed further according to the number numerical values recited for comparing label node；
3), extreme learning machine is neural networks with single hidden layer, improves extreme learning machine in this algorithm and uses bilayer model, accelerates training
Speed, then use this improvements extreme learning machine method, training step 2) in binary tree node model judgement and delete mark
The redundant data in binary tree is remembered, to obtain the most simple document of threedimensional model file.
2. the three-dimensional modeling data analytic method of extracted with high accuracy according to claim 1 and Fast Classification, feature exist
In the step 1) is to extract rubidium marking all in three-dimensional modeling data by automatic state machine, the rubidium marking
Including beginning label and terminate label, and resulting rubidium marking is indicated by chained list node.
3. the three-dimensional modeling data analytic method of extracted with high accuracy according to claim 2 and Fast Classification, feature exist
In, it is described to indicate resulting rubidium marking by chained list node, it specifically includes:
Beginning and end labels all in document are extracted first, and with a label node Element Node table
Show, the format of the label node are as follows: (char Elemstring, intNum, int Match Num, Element Node*
Next), wherein Elem String is used to recording mark character string；Num indicates number of the current markers node in chained list；
Match Num is indicated: if current node is the beginning label of element, recording the end mark to match in chained list with current node
Remember the number of node；If node itself is to terminate label, this is set to sky；Next is directed to the pointer of next node.
4. the three-dimensional modeling data analytic method of extracted with high accuracy according to claim 3 and Fast Classification, feature exist
It include: the binary tree node model in, the step of step 2) establishes binary tree node model by Node String,
Tree Node*Left Child, Tree-Node*Right Child three classes composition, wherein Node String expression are currently opened
Begin the tab character string for marking node to be included；Left Child: left subtree node pointer, if the Left of current node
Child is not empty, then it represents that the node that Left Child is directed toward is nested in current node, their set memberships each other；Right
Child: right subtree node pointer, if the Right Child of current node is not empty, then it represents that Right Child was directed toward
Node and current node are nested in the same label, their brotherhoods each other.
5. the three-dimensional modeling data analytic method of extracted with high accuracy according to claim 4 and Fast Classification, feature exist
In the nest relation of step 2) the analysis each element label, rule has following two o'clock: 1. identifying root element；Find out label
Maximum nodal value in chain identifies that maximum nodal value is root element node；2. tracking chained list by traversal to non-root element
Mobile pointer, to verify the nest relation of model data.
6. the three-dimensional modeling data analytic method of extracted with high accuracy according to claim 5 and Fast Classification, feature exist
The repeated and redundant for being rejected data using the method for node comparison in, the step 3), is specifically included: by by all nodes and its
The right child node of non-empty is compared, if mark it is not identical if without duplicate label, if identical can yet there are two types of different
Situation: (1) tree using current node as root is identical with using its right child node as the tree of root: indicating that two rubidium markings are complete
It is identical, the right child node of current node directly can be set to sky to delete this redundant marks；And be identified as current node can
Repeating label；(2) more not exactly the same using current node as the tree of root and using its right child node as the tree of root, indicate difference
Optional label is nested in current node label or its right child node label, in this case then in the right son for deleting current node
Before node, by the label merging for being nested in right child node but not being nested in current node into the child node of current node, make
It is nested in current node.
7. the three-dimensional modeling data analytic method of extracted with high accuracy according to claim 5 and Fast Classification, feature exist
In the step 3) obtains the most simple document of threedimensional model file using extreme learning machine method is improved, including is divided into trained rank
Section and forecast period, this purpose of model are to establish a kind of mapping pass between input variable and classification results using training sample set
System；
Corresponding structured vectors is defined as: dsv=< d1,d2,…,dn>, n are characterized of the corresponding feature vector of term
Number；Wherein diFor the corresponding feature vector of featured terms of i-th of threedimensional model file:Wherein TF is word frequency, and IDF is inverse document frequency, ωi
For i-th of term, Doc is threedimensional model file, and m is node number, εjFor node, ejIndicate corresponding unit specific steps such as
Under:
Input: training set FD, test set FT, activation primitive g (x)；Concealed nodes number L (L≤T)；Kernel function K, parameter C and γ,
Randomly choose input weight WiWith offset bi；
Output: classification results XTAnd YT；
Step 2.1: then random selection training sample first respectively pre-processes corresponding threedimensional model file, obtains
The vector of threedimensional model file indicates；Finally, quickly establishing disaggregated model by the improvement extreme learning machine method proposed；
Step 2.2: conversion FD to vector model MS-VSM obtains original training data collection D；
Step 2.3: conversion FT to vector model MS-VSM obtains original training data collection T；
Step 2.4: using D as new training set, to reduce trained mistake, carrying out feature extraction, eigenmatrix X is obtained, by X
Classification results X is calculated as next layer of inputT；
Step 2.5: repeating the above training step and obtain parameter L, the optimal value of C, γ；
Step 2.6: using T as new training set, to reduce trained mistake, carrying out feature extraction, eigenmatrix Y is obtained, by Y
Classification results Y is calculated as next layer of inputT；
Final integrated data extracts and inducing classification, provides final document.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811434379.1A CN109359701A (en) | 2018-11-28 | 2018-11-28 | A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811434379.1A CN109359701A (en) | 2018-11-28 | 2018-11-28 | A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109359701A true CN109359701A (en) | 2019-02-19 |
Family
ID=65343154
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811434379.1A Pending CN109359701A (en) | 2018-11-28 | 2018-11-28 | A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109359701A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110245347A (en) * | 2019-05-08 | 2019-09-17 | 平安科技（深圳）有限公司 | A-not-A question automatic generation method, device and storage medium |
| CN111123888A (en) * | 2019-12-19 | 2020-05-08 | 江苏中天科技软件技术有限公司 | Industrial control protocol testing method and system, electronic equipment and storage medium |
| CN113781658A (en) * | 2021-08-16 | 2021-12-10 | 长沙眸瑞网络科技有限公司 | Method and device for processing 3D model data in streaming mode |
| CN114637730A (en) * | 2022-03-23 | 2022-06-17 | 清华大学 | Method, device and system for compressing model file and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104462582A (en) * | 2014-12-30 | 2015-03-25 | 武汉大学 | Web data similarity detection method based on two-stage filtration of structure and content |
| US20150199405A1 (en) * | 2007-01-05 | 2015-07-16 | Digital Doors, Inc. | Information Infrastructure Management Data Processing Tools for Processing Data Flow With Distribution Controls |
| CN105205491A (en) * | 2015-08-19 | 2015-12-30 | 西安电子科技大学 | Polarized SAR image classification method based on extreme learning machine |
| CN105654187A (en) * | 2015-12-21 | 2016-06-08 | 浙江工业大学 | Grid binary tree method of control system midpoint locating method |
| CN107316294A (en) * | 2017-06-28 | 2017-11-03 | 太原理工大学 | One kind is based on improved depth Boltzmann machine Lung neoplasm feature extraction and good pernicious sorting technique |
-
2018
- 2018-11-28 CN CN201811434379.1A patent/CN109359701A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150199405A1 (en) * | 2007-01-05 | 2015-07-16 | Digital Doors, Inc. | Information Infrastructure Management Data Processing Tools for Processing Data Flow With Distribution Controls |
| CN104462582A (en) * | 2014-12-30 | 2015-03-25 | 武汉大学 | Web data similarity detection method based on two-stage filtration of structure and content |
| CN105205491A (en) * | 2015-08-19 | 2015-12-30 | 西安电子科技大学 | Polarized SAR image classification method based on extreme learning machine |
| CN105654187A (en) * | 2015-12-21 | 2016-06-08 | 浙江工业大学 | Grid binary tree method of control system midpoint locating method |
| CN107316294A (en) * | 2017-06-28 | 2017-11-03 | 太原理工大学 | One kind is based on improved depth Boltzmann machine Lung neoplasm feature extraction and good pernicious sorting technique |
Non-Patent Citations (2)
| Title |
|---|
| 赵震 等: "基于双隐层极限学习机的模糊XML文档分类", 《计算机工程与应用》 * |
| 雷庆，熊汉探: "基于标记二叉树的XML数据模式提取算法", 《计算机工程与设计》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110245347A (en) * | 2019-05-08 | 2019-09-17 | 平安科技（深圳）有限公司 | A-not-A question automatic generation method, device and storage medium |
| CN111123888A (en) * | 2019-12-19 | 2020-05-08 | 江苏中天科技软件技术有限公司 | Industrial control protocol testing method and system, electronic equipment and storage medium |
| CN113781658A (en) * | 2021-08-16 | 2021-12-10 | 长沙眸瑞网络科技有限公司 | Method and device for processing 3D model data in streaming mode |
| CN113781658B (en) * | 2021-08-16 | 2024-05-14 | 长沙眸瑞网络科技有限公司 | Method and device for stream processing 3D model data |
| CN114637730A (en) * | 2022-03-23 | 2022-06-17 | 清华大学 | Method, device and system for compressing model file and storage medium |
| CN114637730B (en) * | 2022-03-23 | 2023-01-10 | 清华大学 | Method, device and system for compressing model file and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109359701A (en) | A kind of three-dimensional modeling data analytic method of extracted with high accuracy and Fast Classification | |
| CN111783394B (en) | Training method of event extraction model, event extraction method, system and equipment | |
| CN110334213B (en) | Method for identifying time sequence relation of Hanyue news events based on bidirectional cross attention mechanism | |
| CN104991905B (en) | A kind of mathematic(al) representation search method based on level index | |
| CN105512209A (en) | Biomedicine event trigger word identification method based on characteristic automatic learning | |
| CN107679110A (en) | The method and device of knowledge mapping is improved with reference to text classification and picture attribute extraction | |
| CN103186538A (en) | Image classification method, image classification device, image retrieval method and image retrieval device | |
| CN110390018A (en) | A kind of social networks comment generation method based on LSTM | |
| CN105654144B (en) | A kind of social network ontologies construction method based on machine learning | |
| CN102289522A (en) | Method of intelligently classifying texts | |
| CN113298151A (en) | Remote sensing image semantic description method based on multi-level feature fusion | |
| CN102662923A (en) | Entity instance leading method based on machine learning | |
| KR102015218B1 (en) | Method and apparatus for text classification using machine learning | |
| CN106055560A (en) | Method for collecting data of word segmentation dictionary based on statistical machine learning method | |
| CN110377727A (en) | A kind of multi-tag file classification method and device based on multi-task learning | |
| CN105930792A (en) | Human action classification method based on video local feature dictionary | |
| CN108090223A (en) | A kind of opening scholar portrait method based on internet information | |
| CN111597328A (en) | New event theme extraction method | |
| CN110910175A (en) | Tourist ticket product portrait generation method | |
| CN114443855A (en) | Knowledge graph cross-language alignment method based on graph representation learning | |
| CN106934055A (en) | A kind of semi-supervised automatic webpage classification method based on insufficient modal information | |
| CN112966117A (en) | Entity linking method | |
| CN101655911B (en) | Mode identification method based on immune antibody network | |
| CN107590119A (en) | Character attribute information extraction method and device | |
| CN109858008A (en) | The tendentious method and device of document court verdict based on deep learning |
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: 20190219 |