CN111582373A - Radiation source identification method based on weighted migration extreme learning machine algorithm - Google Patents

Radiation source identification method based on weighted migration extreme learning machine algorithm Download PDF

Info

Publication number
CN111582373A
CN111582373A CN202010385735.6A CN202010385735A CN111582373A CN 111582373 A CN111582373 A CN 111582373A CN 202010385735 A CN202010385735 A CN 202010385735A CN 111582373 A CN111582373 A CN 111582373A
Authority
CN
China
Prior art keywords
sample
representing
samples
domain
learning machine
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
CN202010385735.6A
Other languages
Chinese (zh)
Inventor
苟嫣
邵怀宗
王沙飞
林静然
利强
潘晔
胡全
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.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
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 University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN202010385735.6A priority Critical patent/CN111582373A/en
Publication of CN111582373A publication Critical patent/CN111582373A/en
Pending legal-status Critical Current

Links

Images

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a radiation source identification method based on a weighted transfer extreme learning machine algorithm, which comprises the following steps: collecting a plurality of source domain samples and target domain samples, marking a part of the source domain samples and the target domain samples, and taking the rest of the target domain samples as a sample set to be detected; constructing a training target function of the extreme learning machine model, and updating a model parameter beta of the extreme learning machine model according to the training target function, the source domain mark sample and the target domain mark sample to obtain a trained extreme learning machine model; and inputting the sample set to be tested into the trained extreme learning machine model to obtain a newly added radiation source identification result. The invention improves the knowledge transfer efficiency and the accuracy and stability of the classification and identification of the newly added radiation source individuals under the condition of small samples.

Description

Radiation source identification method based on weighted migration extreme learning machine algorithm
Technical Field
The invention belongs to the field of radiation source identification, and particularly relates to a radiation source identification method based on a weighted migration extreme learning machine algorithm.
Background
In the modern electronic warfare application environment, under the influence of factors such as time, space, combat demand and the like, the rapidly changing battlefield conditions greatly increase the difficulty of equipment for acquiring radiation source signals, and only a small amount of radiation source sample data with labels can be acquired. Therefore, under the condition of a small sample, the problem of identifying newly added individuals of the radiation source needs to be solved urgently. At present, aiming at the problem of identification of individual radiation sources under the condition of a small sample, two solutions are mainly provided: and (3) migration learning and migration limit learning machine algorithms based on the convolutional neural network. The basic idea of convolutional neural network-based migration learning algorithms is to train a convolutional neural network model on a known data set and then apply the model to another data set by fast and simple adjustment. The method comprises the steps of training a convolutional neural network model by utilizing a large number of known radiation source samples, keeping parameters of convolutional layers in the model unchanged, and then training full-link layer parameters of the model by utilizing a small number of newly added radiation source samples with labels to further identify newly added radiation source individuals. However, the number of layers of the deep neural network is large, and the complexity is high, so that the number of network parameters is huge, a small number of labeled samples are difficult to make model parameters fully trained, the error rate of identification of individuals with newly added radiation sources is high, and the identification performance is difficult to meet the application requirements. The migration extreme learning machine algorithm is that on the basis of extreme learning machine and migration learning, a large number of source domain radiation source marking samples and a small number of newly added radiation source marking samples are used for constructing an extreme learning machine model, so that the model migrates knowledge learned from a source domain to a target domain, and the identification of newly added radiation source individuals in the target domain is realized. However, because different source domain samples have different effectiveness for establishing the target model, directly using each sample in the source domain easily causes the problem of "negative migration" of the model, and affects the recognition performance of the model on the target sample set, so under this algorithm, the recognition performance of the model is unstable, and when a bad sample exists in the source domain, the model directly interferes and affects the construction of the target model.
Disclosure of Invention
Aiming at the defects in the prior art, the radiation source identification method based on the weighted migration limit learning machine algorithm solves the problems that in the prior art, the identification error rate is high, the model identification performance is unstable, and the algorithm is seriously dependent on the number of samples.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that: a radiation source identification method based on a weighted migration extreme learning machine algorithm comprises the following steps:
s1, collecting a plurality of source domain samples and target domain samples, marking a part of the source domain samples and the target domain samples, and taking the rest part of the target domain samples as a sample set to be detected;
s2, constructing a training target function of the extreme learning machine model, and updating a model parameter beta of the extreme learning machine model according to the training target function, the source domain marker sample and the target domain marker sample to obtain a trained extreme learning machine model;
and S3, inputting the sample set to be tested into the trained extreme learning machine model to obtain the newly added radiation source identification result.
Further, the objective function L is trained in the step S2W-TELMComprises the following steps:
Figure BDA0002483855600000021
the above-mentioned
Figure BDA0002483855600000022
And
Figure BDA0002483855600000023
the relationship with β is:
Figure BDA0002483855600000024
wherein s.t. represents a condition limit, β represents a model parameter of the extreme learning machine model, CsA prediction error balance constant, C, representing the source domain stRepresenting the prediction error balance constant of the target domain t, W representing the weight matrix of the source domain samples, NsRepresents the total number of source domain samples, NtRepresenting the number of marked samples in the target domain sample,
Figure BDA0002483855600000025
representing the prediction error of the ith source domain sample,
Figure BDA0002483855600000026
representing the prediction error of the jth target domain sample,
Figure BDA0002483855600000027
representing the hidden layer output corresponding to the source domain samples,
Figure BDA0002483855600000028
representing the hidden layer output corresponding to the target domain sample,
Figure BDA0002483855600000029
label representing the ith source domain sample, Yt jA label representing the jth target domain sample.
Further, in the step S2, the specific method for updating the model parameter β of the extreme learning machine model according to the training objective function, the source domain labeled sample and the target domain labeled sample is as follows:
a1, inputting the source domain marking sample into an extreme learning machine model, and obtaining hidden layer output of the source domain sample;
a2, acquiring a weight matrix of a source domain sample;
a3, inputting the target domain mark sample into an extreme learning machine model, and obtaining hidden layer output of the target domain sample;
and A4, updating the model parameter beta of the extreme learning machine model according to the hidden layer output of the source domain sample, the weight matrix and the hidden layer output of the target domain sample.
Further, the method can be used for preparing a novel materialIn step A1, the source domain labeled sample is
Figure BDA0002483855600000031
Wherein, XsRepresenting a set of source domain samples, YsA set of sample labels representing the source domain,
Figure BDA0002483855600000032
represents the ith source domain sample,
Figure BDA0002483855600000033
represents the ith source domain sample label;
obtaining the hidden layer output H of the source domain sample in the step A1sThe formula of (1) is:
Hs＝g(a·Xs+b)
where g () represents an activation function, a represents an input weight parameter, and b represents a hidden layer bias.
Further, the weight matrix for obtaining the source domain samples in step a2 is W ═ diag {iRepresents a diagonal matrix, the weights WiComprises the following steps:
Figure BDA0002483855600000034
wherein d isiDenotes the euclidean distance between the ith source domain sample and the target domain sample mean point, i ═ 1,2s，NsRepresenting the total number of source domain samples.
Further, the target domain mark sample in the step A3 is
Figure BDA0002483855600000035
Wherein, XtRepresenting a set of target domain samples, YtA set of sample labels representing the target domain,
Figure BDA0002483855600000036
denotes the jth target domain sample, Yt jDenotes the jth target field sample label, j 1,2t，NtRepresenting the total number of target domain marker samples;
obtaining the hidden layer output H of the target domain sample in the step A3tComprises the following steps:
Ht＝g(a·Xt+b)
where g () represents an activation function, a represents an input weight parameter, and b represents a hidden layer bias.
Further, the step a4 is specifically to set the gradient of the training objective function with respect to the model parameter β to zero, and determine whether the number of training samples is greater than the number of hidden nodes, if so, make the model parameter β equal to the first optimal solution β1Otherwise, let the model parameters β equal the second optimal solution β2The update of the model parameters β is completed.
Further, the training sample numbers include a target-domain marker sample number and a source-domain marker sample number.
Further, the first optimal solution β1Comprises the following steps:
Figure BDA0002483855600000041
wherein, CsA prediction error balance constant, H, representing the source domain ssRepresenting the hidden layer output of the source domain samples, T representing the transpose, W representing the weight matrix of the source domain samples, CtA prediction error balance constant, H, representing the target field ttHidden layer output, Y, representing a target domain samplesRepresenting a set of source domain sample labels, YtRepresenting a target domain sample label set.
Further, the second optimal solution β2Comprises the following steps:
Figure BDA0002483855600000042
wherein A represents a first calculation parameter,
Figure BDA0002483855600000043
b denotes a second calculation parameter which is,
Figure BDA0002483855600000044
i denotes an identity matrix, C denotes a third calculation parameter,
Figure BDA0002483855600000045
d denotes a fourth calculation parameter which is,
Figure BDA0002483855600000046
the invention has the beneficial effects that:
(1) according to the invention, aiming at the problem that the source field samples have difference in help of the target tasks, the source field samples are weighted, samples which are beneficial to the training of the target model are selected from the source field, and the weight of the samples is increased, so that the classifier model can be more suitable for the target classification tasks, and the problem of negative migration is solved.
(2) The invention improves the knowledge transfer efficiency and the accuracy and stability of the classification and identification of the newly added radiation source individuals under the condition of small samples.
(3) According to the method, the extreme learning machine is enabled to have knowledge transfer capability by combining the extreme learning machine theory and establishing the target classifier model by utilizing the source domain samples and the limited labeled samples in the target domain, repeated iterative training is not required to be carried out by depending on a large number of samples, and the condition limit of small samples is effectively met.
Drawings
Fig. 1 is a flowchart of a radiation source identification method based on a weighted transfer extreme learning machine algorithm according to the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
As shown in fig. 1, a radiation source identification method based on a weighted migration extreme learning algorithm includes the following steps:
s1, collecting a plurality of source domain samples and target domain samples, marking a part of the source domain samples and the target domain samples, and taking the rest part of the target domain samples as a sample set to be detected.
In this embodiment, the target domain marked samples are 2% of the total number of samples, i.e. the number of samples to be detected accounts for 98% of the total number of samples in the target domain.
S2, constructing a training target function of the extreme learning machine model, and updating the model parameter beta of the extreme learning machine model according to the training target function, the source domain mark sample and the target domain mark sample to obtain the trained extreme learning machine model.
In the present embodiment, the model parameters β are initialized randomly before being updated.
And S3, inputting the sample set to be tested into the trained extreme learning machine model to obtain the newly added radiation source identification result.
In this embodiment, the known radiation source is a source domain s, the newly added radiation source is a target domain t, and the radiation source is a communication station.
Training the objective function L in the step S2W-TELMComprises the following steps:
Figure BDA0002483855600000061
the above-mentioned
Figure BDA0002483855600000062
And
Figure BDA0002483855600000063
the relationship with β is:
Figure BDA0002483855600000064
wherein s.t. represents a condition limit, β representsModel parameters of extreme learning machine model, CsA prediction error balance constant, C, representing the source domain stRepresenting the prediction error balance constant of the target domain t, W representing the weight matrix of the source domain samples, NsRepresents the total number of source domain samples, NtRepresenting the number of marked samples in the target domain sample,
Figure BDA0002483855600000065
representing the prediction error of the ith source domain sample,
Figure BDA0002483855600000066
representing the prediction error of the jth target domain sample,
Figure BDA0002483855600000067
representing the hidden layer output corresponding to the source domain samples,
Figure BDA0002483855600000068
representing the hidden layer output corresponding to the target domain sample,
Figure BDA0002483855600000069
label representing the ith source domain sample, Yt jA label representing the jth target domain sample.
The specific method for updating the model parameter β of the extreme learning machine model according to the training objective function, the source domain marker sample and the target domain marker sample in step S2 is as follows:
a1, inputting the source domain marking sample into an extreme learning machine model, and obtaining hidden layer output of the source domain sample;
a2, acquiring a weight matrix of a source domain sample;
a3, inputting the target domain mark sample into an extreme learning machine model, and obtaining hidden layer output of the target domain sample;
and A4, updating the model parameter beta of the extreme learning machine model according to the hidden layer output of the source domain sample, the weight matrix and the hidden layer output of the target domain sample.
In this embodiment, setting the gradient of the training objective function with respect to the model parameter β to zero may result:
Figure BDA0002483855600000071
in the present embodiment, the number of hidden nodes is set to 1e + 3.
The source domain marker sample in the step A1 is
Figure BDA0002483855600000072
Wherein, XsRepresenting a set of source domain samples, YsA set of sample labels representing the source domain,
Figure BDA0002483855600000073
represents the ith source domain sample,
Figure BDA0002483855600000074
represents the ith source domain sample label;
obtaining the hidden layer output H of the source domain sample in the step A1sThe formula of (1) is:
Hs＝g(a·Xs+b)
where g () represents an activation function, a represents an input weight parameter, and b represents a hidden layer bias.
The weight matrix for obtaining the source domain sample in step a2 is W ═ diag { (W } { (W })iRepresents a diagonal matrix, the weights WiComprises the following steps:
Figure BDA0002483855600000075
wherein d isiDenotes the euclidean distance between the ith source domain sample and the target domain sample mean point, i ═ 1,2s，NsRepresenting the total number of source domain samples.
The target domain mark sample in the step A3 is
Figure BDA0002483855600000076
Wherein, XtRepresenting a set of target domain samples, YtRepresenting a target domain sample label set，
Figure BDA0002483855600000077
Denotes the jth target domain sample, Yt jDenotes the jth target field sample label, j 1,2t，NtRepresenting the total number of target domain marker samples;
obtaining the hidden layer output H of the target domain sample in the step A3tComprises the following steps:
Ht＝g(a·Xt+b)
where g () represents an activation function, a represents an input weight parameter, and b represents a hidden layer bias.
The step A4 is specifically to set the gradient of the training objective function relative to the model parameter β to zero, and determine whether the number of training samples is greater than the number of hidden nodes, if so, make the model parameter β equal to the first optimal solution β1Otherwise, let the model parameters β equal the second optimal solution β2The update of the model parameters β is completed.
The training sample numbers include a target-domain marker sample number and a source-domain marker sample number.
The first optimal solution β1Comprises the following steps:
Figure BDA0002483855600000081
wherein, CsA prediction error balance constant, H, representing the source domain ssRepresenting the hidden layer output of the source domain samples, T representing the transpose, W representing the weight matrix of the source domain samples, CtA prediction error balance constant, H, representing the target field ttHidden layer output, Y, representing a target domain samplesRepresenting a set of source domain sample labels, YtRepresenting a target domain sample label set.
The second optimal solution β2Comprises the following steps:
Figure BDA0002483855600000082
wherein A represents a first calculation parameter,
Figure BDA0002483855600000083
b denotes a second calculation parameter which is,
Figure BDA0002483855600000084
i denotes an identity matrix, C denotes a third calculation parameter,
Figure BDA0002483855600000085
d denotes a fourth calculation parameter which is,
Figure BDA0002483855600000086
the invention has the beneficial effects that:
(1) according to the invention, aiming at the problem that the source field samples have difference in help of the target tasks, the source field samples are weighted, samples which are beneficial to the training of the target model are selected from the source field, and the weight of the samples is increased, so that the classifier model can be more suitable for the target classification tasks, and the problem of negative migration is solved.
(2) The invention improves the knowledge transfer efficiency and the accuracy and stability of the classification and identification of the newly added radiation source individuals under the condition of small samples.
(3) According to the method, the extreme learning machine is enabled to have knowledge transfer capability by combining the extreme learning machine theory and establishing the target classifier model by utilizing the source domain samples and the limited labeled samples in the target domain, repeated iterative training is not required to be carried out by depending on a large number of samples, and the condition limit of small samples is effectively met.

Claims (10)

1. A radiation source identification method based on a weighted migration extreme learning machine algorithm is characterized by comprising the following steps:
s1, collecting a plurality of source domain samples and target domain samples, marking a part of the source domain samples and the target domain samples, and taking the rest part of the target domain samples as a sample set to be detected;
s2, constructing a training target function of the extreme learning machine model, and updating a model parameter beta of the extreme learning machine model according to the training target function, the source domain marker sample and the target domain marker sample to obtain a trained extreme learning machine model;
and S3, inputting the sample set to be tested into the trained extreme learning machine model to obtain the newly added radiation source identification result.
2. The radiation source identification method based on the weighted transfer extreme learning machine algorithm as claimed in claim 1, wherein the step S2 is to train an objective function LW-TELMComprises the following steps:
Figure FDA0002483855590000011
the above-mentioned
Figure FDA0002483855590000012
And
Figure FDA0002483855590000013
the relationship with β is:
Figure FDA0002483855590000014
wherein s.t. represents a condition limit, β represents a model parameter of the extreme learning machine model, CsA prediction error balance constant, C, representing the source domain stRepresenting the prediction error balance constant of the target domain t, W representing the weight matrix of the source domain samples, NsRepresents the total number of source domain samples, NtRepresenting the number of marked samples in the target domain sample,
Figure FDA0002483855590000015
representing the prediction error of the ith source domain sample,
Figure FDA0002483855590000016
representing the prediction error of the jth target domain sample,
Figure FDA0002483855590000017
representing the hidden layer output corresponding to the source domain samples,
Figure FDA0002483855590000018
representing the hidden layer output corresponding to the target domain sample,
Figure FDA0002483855590000019
label representing the ith source domain sample, Yt jA label representing the jth target domain sample.
3. The radiation source identification method based on the weighted migration extreme learning machine algorithm according to claim 2, wherein the specific method for updating the model parameter β of the extreme learning machine model according to the training target function, the source domain label samples and the target domain label samples in step S2 is as follows:
a1, inputting the source domain marking sample into an extreme learning machine model, and obtaining hidden layer output of the source domain sample;
a2, acquiring a weight matrix of a source domain sample;
a3, inputting the target domain mark sample into an extreme learning machine model, and obtaining hidden layer output of the target domain sample;
and A4, updating the model parameter beta of the extreme learning machine model according to the hidden layer output of the source domain sample, the weight matrix and the hidden layer output of the target domain sample.
4. The radiation source identification method based on the weighted transfer extreme learning machine algorithm as claimed in claim 3, wherein the source domain label samples in the step A1 are
Figure FDA0002483855590000021
Wherein, XsRepresenting a set of source domain samples, YsA set of sample labels representing the source domain,
Figure FDA0002483855590000022
representing the ith source domainThe sample is taken from the sample container,
Figure FDA0002483855590000023
represents the ith source domain sample label;
obtaining the hidden layer output H of the source domain sample in the step A1sThe formula of (1) is:
Hs＝g(a·Xs+b)
where g () represents an activation function, a represents an input weight parameter, and b represents a hidden layer bias.
5. The radiation source identification method based on the weighted transfer extreme learning machine algorithm as claimed in claim 3, wherein the weight matrix of the source domain samples obtained in the step A2 is W ═ diag { W ═ W ═iRepresents a diagonal matrix, the weights WiComprises the following steps:
Figure FDA0002483855590000024
wherein d isiDenotes the euclidean distance between the ith source domain sample and the target domain sample mean point, i ═ 1,2s，NsRepresenting the total number of source domain samples.
6. The radiation source identification method based on the weighted transfer extreme learning machine algorithm as claimed in claim 3, wherein the target domain mark samples in the step A3 are
Figure FDA0002483855590000025
Wherein, XtRepresenting a set of target domain samples, YtA set of sample labels representing the target domain,
Figure FDA0002483855590000026
denotes the jth target domain sample, Yt jDenotes the jth target field sample label, j 1,2t，NtRepresenting the total number of target domain marker samples;
said step (c) isObtaining hidden layer output H of target domain sample in A3tComprises the following steps:
Ht＝g(a·Xt+b)
where g () represents an activation function, a represents an input weight parameter, and b represents a hidden layer bias.
7. The method as claimed in claim 3, wherein the step A4 is to set the gradient of the training objective function with respect to the model parameter β to zero, and determine whether the number of training samples is greater than the number of hidden nodes, if so, make the model parameter β equal to the first optimal solution β1Otherwise, let the model parameters β equal the second optimal solution β2The update of the model parameters β is completed.
8. The method of claim 7, wherein the training sample numbers comprise target-domain marker sample numbers and source-domain marker sample numbers.
9. The radiation source identification method based on the weighted transfer extreme learning machine algorithm of claim 8, wherein the first optimal solution β is1Comprises the following steps:
Figure FDA0002483855590000031
wherein, CsA prediction error balance constant, H, representing the source domain ssRepresenting the hidden layer output of the source domain samples, T representing the transpose, W representing the weight matrix of the source domain samples, CtA prediction error balance constant, H, representing the target field ttHidden layer output, Y, representing a target domain samplesRepresenting a set of source domain sample labels, YtRepresenting a target domain sample label set.
10. The radiation source identification method based on the weighted transfer extreme learning machine algorithm according to claim 9, whichCharacterized in that said second optimal solution β2Comprises the following steps:
Figure FDA0002483855590000032
wherein A represents a first calculation parameter,
Figure FDA0002483855590000033
b denotes a second calculation parameter which is,
Figure FDA0002483855590000034
i denotes an identity matrix, C denotes a third calculation parameter,
Figure FDA0002483855590000035
d denotes a fourth calculation parameter which is,
Figure FDA0002483855590000036
CN202010385735.6A 2020-05-09 2020-05-09 Radiation source identification method based on weighted migration extreme learning machine algorithm Pending CN111582373A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010385735.6A CN111582373A (en) 2020-05-09 2020-05-09 Radiation source identification method based on weighted migration extreme learning machine algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010385735.6A CN111582373A (en) 2020-05-09 2020-05-09 Radiation source identification method based on weighted migration extreme learning machine algorithm

Publications (1)

Publication Number Publication Date
CN111582373A true CN111582373A (en) 2020-08-25

Family

ID=72126390

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010385735.6A Pending CN111582373A (en) 2020-05-09 2020-05-09 Radiation source identification method based on weighted migration extreme learning machine algorithm

Country Status (1)

Country Link
CN (1) CN111582373A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070236A (en) * 2020-09-11 2020-12-11 福州大学 Sparse feature learning method for solving online complex optimization calculation based on transfer learning
CN113029559A (en) * 2021-03-30 2021-06-25 山东大学 Gear box fault diagnosis method and system
CN113177520A (en) * 2021-05-26 2021-07-27 电子科技大学 Intelligent radiation source identification method based on ensemble learning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107301381A (en) * 2017-06-01 2017-10-27 西安电子科技大学昆山创新研究院 Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy
CN107832787A (en) * 2017-10-31 2018-03-23 杭州电子科技大学 Recognition Method of Radar Emitters based on bispectrum own coding feature
AU2018101317A4 (en) * 2018-09-07 2018-10-11 Chen, Guoyi Mr A Deep Learning Based System for Animal Species Classification
CN109446877A (en) * 2018-09-01 2019-03-08 哈尔滨工程大学 A kind of radar emitter signal Modulation Identification method of joint multidimensional characteristic migration fusion

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107301381A (en) * 2017-06-01 2017-10-27 西安电子科技大学昆山创新研究院 Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy
CN107832787A (en) * 2017-10-31 2018-03-23 杭州电子科技大学 Recognition Method of Radar Emitters based on bispectrum own coding feature
CN109446877A (en) * 2018-09-01 2019-03-08 哈尔滨工程大学 A kind of radar emitter signal Modulation Identification method of joint multidimensional characteristic migration fusion
AU2018101317A4 (en) * 2018-09-07 2018-10-11 Chen, Guoyi Mr A Deep Learning Based System for Animal Species Classification

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
X.LI等: ""Extreme learning machine based transfer learning for data classification\"" *
余欢欢;魏文戈;: "模型参数自适应迁移的多源域适应" *
卢忱;王晶;: "一种基于自动标签机制的行为识别模型迁移方法" *
孟凡杰等: "基于多特征融合的雷达辐射源信号识别" *
舒醒;于慧敏;郑伟伟;谢奕;胡浩基;唐慧明;: "基于边际Fisher准则和迁移学习的小样本集分类器设计算法" *
董莹莹;邓万宇;刘光达;: "基于score样本选择的同构域适应迁移学习" *
黄健航;雷迎科;: "基于深度学习的通信辐射源指纹特征提取算法" *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070236A (en) * 2020-09-11 2020-12-11 福州大学 Sparse feature learning method for solving online complex optimization calculation based on transfer learning
CN113029559A (en) * 2021-03-30 2021-06-25 山东大学 Gear box fault diagnosis method and system
CN113029559B (en) * 2021-03-30 2022-03-18 山东大学 Gear box fault diagnosis method and system
CN113177520A (en) * 2021-05-26 2021-07-27 电子科技大学 Intelligent radiation source identification method based on ensemble learning
CN113177520B (en) * 2021-05-26 2022-06-28 电子科技大学 Intelligent radiation source identification method based on ensemble learning

Similar Documents

Publication Publication Date Title
CN111738315B (en) Image classification method based on countermeasure fusion multi-source transfer learning
CN111582373A (en) Radiation source identification method based on weighted migration extreme learning machine algorithm
CN111191732A (en) Target detection method based on full-automatic learning
CN108875933B (en) Over-limit learning machine classification method and system for unsupervised sparse parameter learning
CN110175615B (en) Model training method, domain-adaptive visual position identification method and device
CN112347970B (en) Remote sensing image ground object identification method based on graph convolution neural network
CN105701509B (en) A kind of image classification method based on across classification migration Active Learning
CN113112020B (en) Model network extraction and compression method based on generation network and knowledge distillation
CN111079847A (en) Remote sensing image automatic labeling method based on deep learning
CN103886330A (en) Classification method based on semi-supervised SVM ensemble learning
CN107909053B (en) Face detection method based on hierarchical learning cascade convolution neural network
CN111239137B (en) Grain quality detection method based on transfer learning and adaptive deep convolution neural network
CN110222785B (en) Self-adaptive confidence active learning method for drift correction of gas sensor
CN110619059A (en) Building marking method based on transfer learning
CN112685504A (en) Production process-oriented distributed migration chart learning method
CN114726692B (en) SERESESESENet-LSTM-based radiation source modulation mode identification method
CN116522118A (en) Fault diagnosis method based on improved unsupervised domain self-adaptive network
CN116129219A (en) SAR target class increment recognition method based on knowledge robust-rebalancing network
CN112381730B (en) Remote sensing image data amplification method
CN113837046A (en) Small sample remote sensing image scene classification method based on iterative feature distribution learning
CN115294381B (en) Small sample image classification method and device based on feature migration and orthogonal prior
CN116089812A (en) Fault diagnosis method based on semi-supervised contrast domain generalization intelligent model
CN114494819B (en) Anti-interference infrared target identification method based on dynamic Bayesian network
CN109635237A (en) It is a kind of based on the target platform method for quickly identifying for dynamically cutting out Configuration knowledge
CN109977797A (en) The optimization method of single order object detector based on sequence loss function

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: 20200825