CN103279761A - On-line vector selecting method for support vector machine - Google Patents
On-line vector selecting method for support vector machine Download PDFInfo
- Publication number
- CN103279761A CN103279761A CN201310186771XA CN201310186771A CN103279761A CN 103279761 A CN103279761 A CN 103279761A CN 201310186771X A CN201310186771X A CN 201310186771XA CN 201310186771 A CN201310186771 A CN 201310186771A CN 103279761 A CN103279761 A CN 103279761A
- Authority
- CN
- China
- Prior art keywords
- vector
- support
- border
- vectors
- support vector
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an on-line vector selecting method for a support vector machine, which comprises the concrete steps of computing a sector set for solving boundary vectors when new training vectors are obtained; comparing the obtained boundary vectors with original support vectors to remove the boundary vectors coincided with the support vectors; judging the processed boundary vectors by the support vector machine to preserve a misjudged boundary vector set; and combining the preserved boundary vectors and original support vectors to finally obtain the training vectors supporting the training of the vector machine. The on-line vector selecting method for the support vector machine is capable of greatly reducing the number of training vectors, thus reducing the training time of the support vector machine without affecting the sorting precision of the support vector machine.
Description
Technical field
The present invention relates to the choosing method of online vector, be applied to choosing of the preceding online vector of training of support vector machine on-line study.
Background technology
In machine learning, support vector machine is owing to the classification that can solve non-linear sample has efficiently obtained using widely.Many times, the disaggregated model of support vector machine is not disposablely just can finish training, and its training vector can increase and constantly increase new training sample along with the time, and we call online vector to newly-increased training sample vector usually.For support vector machine, not all online vector is all to the meaning that all has no less important of finding the solution of disaggregated model, and in fact a lot of vectors do not influence finding the solution of disaggregated model except the time of improving training.For the application of machine learning, reduce the training time and guarantee that training precision is important performance requirement.Therefore, the choosing method of online vector just has important effect
On-line study vector processing method commonly used is to select by time window, sorts by the time sequencing that adds for online vector set exactly.After this, along with the increase of online vector, constantly remove the sample vector that adds the earliest, add new sample vector, form the online vectorial sample set after upgrading.
And in the application of on-line study, this simple on-line study vector processing method, it is the mechanical alignment amount that is chosen at, the sample that has no basis sorts to the importance of support vector machine disaggregated model, so the online vector set of choosing does not guarantee for the classification accuracy rate of support vector machine.In order to improve performance, need there be new choosing method to solve problem.
Therefore, the online vector of support vector machine is chosen and is also had a lot of improvements.
Summary of the invention
The present invention is directed to the deficiencies in the prior art, a kind of online vectorial choosing method for support vector machine is provided.
The present invention finds the solution vector space border vector, the border vector compared, and the wrongheaded border of record support vector machine vector, former support vector and wrongheaded support vector are merged into new training vector, specifically:
(1) when obtaining new training vector, the subtend duration set calculates finds the solution the border vector.
(2) the border vector that obtains is compared with former support vector, remove the border vector that overlaps with support vector.
(3) the border vector after will handling is judged by support vector machine, keeps the vector set of wrongheaded border.
(4) the border vector that keeps is just obtained the final training vector that support vector machine is trained that is used for former support vector merging.
The present invention can significantly reduce the quantity of training vector, therefore reduces the training time of support vector machine, and can not influence the nicety of grading of support vector machine.
Embodiment
Describing step of the present invention below in detail forms: find the solution vector space border vector, the border vector is compared, and the wrongheaded border of record support vector machine vector, former support vector and wrongheaded support vector are merged into new training vector.Concrete implementation procedure is as follows.
(1) vector space border vector derivation algorithm:
1, will contain given training vector collection
Point set on interior minimum hypersphere initially is made as
, that is:
It is the training vector collection
Mapping on minimum hypersphere.
2, according to the distance to the hypersphere center, to the training vector collection
Carrying out descending handles and to arrange
,
I.e. expression is through the training vector collection after sorting.
3, add fashionablely as new training vector, according to the centre coordinate of former hypersphere, calculate its centre distance to hypersphere, and upgrade original
4, surpass when new training vector
Scope the time, the centre coordinate of iterative computation hypersphere again.In a word,
In the coordinate the first five/one part vector is exactly the border vector of asking.
(2) the border vector is compared
The border of finding the solution vector set and former support vector set are compared, because may there be the vector of coincidence in two kinds of set, therefore deletion and support vector superposed part in the vector set of border only keeps the border vector set that does not overlap.
(3) the wrongheaded border of record support vector machine vector
The border that obtains after above-mentioned processing vector is utilized the support vector machine that has trained judgements of classifying, note the wrongheaded border of support vector machine classifier vector and gather.
(4) merge into new training vector
The border vector set of above-mentioned error in judgement is merged with former support vector, just constituted new online vector set.So far, just finished choosing of online vector set.
The above only is the specific embodiment of the present invention, not in order to limiting the present invention, and those of skill in the art under any the present invention, in the technical scope that the present invention discloses, the modification of doing or replacement all should be encompassed within protection scope of the present invention.
Claims (1)
1. online vectorial choosing method that is used for support vector machine, it is characterized in that: this method is found the solution vector space border vector, the border vector is compared, the wrongheaded border of record support vector machine vector, former support vector and wrongheaded support vector are merged into new training vector, specifically:
(1) when obtaining new training vector, the subtend duration set calculates finds the solution the border vector;
(2) the border vector that obtains is compared with former support vector, remove the border vector that overlaps with support vector;
(3) the border vector after will handling is judged by support vector machine, keeps the vector set of wrongheaded border;
(4) the border vector that keeps is just obtained the final training vector that support vector machine is trained that is used for former support vector merging.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310186771XA CN103279761A (en) | 2013-05-20 | 2013-05-20 | On-line vector selecting method for support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310186771XA CN103279761A (en) | 2013-05-20 | 2013-05-20 | On-line vector selecting method for support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103279761A true CN103279761A (en) | 2013-09-04 |
Family
ID=49062276
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310186771XA Pending CN103279761A (en) | 2013-05-20 | 2013-05-20 | On-line vector selecting method for support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103279761A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104750867A (en) * | 2015-04-16 | 2015-07-01 | 南通理工学院 | Adaptive vector projecting type linear supporting vector selecting method |
| CN104750857A (en) * | 2015-04-16 | 2015-07-01 | 南通理工学院 | Adaptive vector projecting type nonlinear supporting vector selecting method |
| CN108763448A (en) * | 2018-05-28 | 2018-11-06 | 重庆工业职业技术学院 | A kind of dissemination method of the electronic information based on Internet of Things |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1460947A (en) * | 2003-06-13 | 2003-12-10 | 北京大学计算机科学技术研究所 | Text classification incremental training learning method supporting vector machine by compromising key words |
| US20070122009A1 (en) * | 2005-11-26 | 2007-05-31 | Hyung Keun Jee | Face recognition method and apparatus |
| CN101944122A (en) * | 2010-09-17 | 2011-01-12 | 浙江工商大学 | Incremental learning-fused support vector machine multi-class classification method |
| CN102982392A (en) * | 2012-11-07 | 2013-03-20 | 中国科学院亚热带农业生态研究所 | Index of agricultural rodent pest outbreak risk estimation method based on geographical information system |
-
2013
- 2013-05-20 CN CN201310186771XA patent/CN103279761A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1460947A (en) * | 2003-06-13 | 2003-12-10 | 北京大学计算机科学技术研究所 | Text classification incremental training learning method supporting vector machine by compromising key words |
| US20070122009A1 (en) * | 2005-11-26 | 2007-05-31 | Hyung Keun Jee | Face recognition method and apparatus |
| CN101944122A (en) * | 2010-09-17 | 2011-01-12 | 浙江工商大学 | Incremental learning-fused support vector machine multi-class classification method |
| CN102982392A (en) * | 2012-11-07 | 2013-03-20 | 中国科学院亚热带农业生态研究所 | Index of agricultural rodent pest outbreak risk estimation method based on geographical information system |
Non-Patent Citations (2)
| Title |
|---|
| 李金凤: ""支持向量机增量学习算法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 覃俊: ""基于壳向量的支持向量机渐进式增量学习算法"", 《中南民族大学学报（自然科学版）》, vol. 30, no. 3, 28 April 2012 (2012-04-28) * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104750867A (en) * | 2015-04-16 | 2015-07-01 | 南通理工学院 | Adaptive vector projecting type linear supporting vector selecting method |
| CN104750857A (en) * | 2015-04-16 | 2015-07-01 | 南通理工学院 | Adaptive vector projecting type nonlinear supporting vector selecting method |
| CN108763448A (en) * | 2018-05-28 | 2018-11-06 | 重庆工业职业技术学院 | A kind of dissemination method of the electronic information based on Internet of Things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107563446B (en) | Target detection method for micro-operation system | |
| CN106980623B (en) | Data model determination method and device | |
| KR102424803B1 (en) | Touch classification | |
| TWI752455B (en) | Image classification model training method, image processing method, data classification model training method, data processing method, computer device, and storage medium | |
| US20200065710A1 (en) | Normalizing text attributes for machine learning models | |
| CN106294350A (en) | A kind of text polymerization and device | |
| CN104915327A (en) | Text information processing method and device | |
| CN108022262A (en) | A kind of point cloud registration method based on neighborhood of a point center of gravity vector characteristics | |
| CN108733644B (en) | A kind of text emotion analysis method, computer readable storage medium and terminal device | |
| CN111325223B (en) | Training method and device for deep learning model and computer readable storage medium | |
| CN109726391B (en) | Method, device and terminal for emotion classification of text | |
| CN103279761A (en) | On-line vector selecting method for support vector machine | |
| CN105930859B (en) | Radar Signal Sorting Method based on linear manifold cluster | |
| CN110348540B (en) | Clustering-based method and device for screening transient power angle stability faults of power system | |
| CN108428234B (en) | Interactive segmentation performance optimization method based on image segmentation result evaluation | |
| CN110428438B (en) | Single-tree modeling method and device and storage medium | |
| CN111597336A (en) | Processing method and device of training text, electronic equipment and readable storage medium | |
| CN105488523A (en) | Data clustering analysis method based on Grassmann manifold | |
| CN109783698B (en) | Industrial production data entity identification method based on Merkle-tree | |
| CN104331507B (en) | Machine data classification is found automatically and the method and device of classification | |
| CN111274123A (en) | Automatic generation method and framework of safety protection software test set based on software genes | |
| Hong et al. | Semi-supervised learning for sentiment analysis in mass social media | |
| CN105654106A (en) | Decision tree generation method and system thereof | |
| CN114139636B (en) | Abnormal operation processing method and device | |
| CN104765776A (en) | Data sample clustering method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130904 |
|
| WD01 | Invention patent application deemed withdrawn after publication |