CN109858509A - Based on multilayer stochastic neural net single classifier method for detecting abnormality - Google Patents
Based on multilayer stochastic neural net single classifier method for detecting abnormality Download PDFInfo
- Publication number
- CN109858509A CN109858509A CN201811305622.XA CN201811305622A CN109858509A CN 109858509 A CN109858509 A CN 109858509A CN 201811305622 A CN201811305622 A CN 201811305622A CN 109858509 A CN109858509 A CN 109858509A
- Authority
- CN
- China
- Prior art keywords
- elm
- input
- output
- multilayer
- sample
- 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
- Image Analysis (AREA)
Abstract
The invention discloses one kind to be based on multilayer stochastic neural net single classifier method for detecting abnormality.Present invention input only includes the training dataset of normal class；Input sample data are handled by multilayer ELM-AE self-encoding encoder coding and decoding, the characteristic value reconstructed；The characteristic value input the last layer ELM of reconstruct is obtained into reality output；Then the range error vector for obtaining reality output and output label is ranked up from big to small, according to the threshold parameter of setting, determines the threshold value for separating normal class and exception class；Finally, test data is input in the multilayer stochastic neural net list abnormal classification detection model, the recognition effect of the model is tested.The present invention mutually more quickly and efficiently extracts main information and dimensionality reduction, then carries out identification classification.Faster, accuracy rate is higher for speed, and Generalization Capability is more preferable.It is applicable not only to small data set and is applied equally to higher-dimension large data sets, there is universality.Have great importance for practical application from now on.
Description
Technical field
The invention belongs to machine learning and the field of data mining, are related to a kind of based on multilayer stochastic neural net single classifier
Method for detecting abnormality.
Background technique
Abnormality detection is an important branch in machine learning and data mining, is widely used in each field, example
If the credit card fraud in trade financing field detects, the disease detection and chemical toxicity in biologic pharmacological science are detected, meter
The analysis detection etc. of calculation machine image domains.The presence of abnormal data can bring certain harm and loss, seriously threaten people
The security of the lives and property.Therefore having great importance extremely present in data how is detected.
Abnormality detection is to detect data undesirably, behavior, but in reality by analyzing input data
In detection process, it is faced with lot of challenges.One, acquisition is accurate, representational label is highly difficult, especially for abnormal data
For, the data volume of tape label is seldom；Two, certain fields are normal and abnormal data there is no specific boundaries；Three, data sheets
There are noise, noise and exceptions to be difficult to differentiate between for body；Four, are normal and abnormal data is unevenly distributed weighing apparatus；The existing abnormality detection list of five,
Often there is the problems such as characterization ability is inadequate, and discrimination is low when for higher-dimension big data sample in classifier.
In known existing method for detecting abnormality, based on traditional neural network, support vector machine, rule and arest neighbors etc.
It is slow that method carries out abnormality detection not only speed, but also when large data sets unbalanced in face of higher-dimension, not to the characterization abilities of data
It is enough, lead to training effect difference and is unable to satisfy the demand of real-time.Therefore, how to be concentrated through in more generally data a kind of excellent
It is that a core is asked that different algorithm, which obtains the fast abnormality detection model of the good high training speed of discrimination of the strong Generalization Capability of characterization ability,
Topic.There is employed herein a kind of multilayer based on learning machine (Extreme Learning Machine, the ELM) algorithm that transfinites is refreshing at random
Single classifier through network is classified to carry out abnormality detection.
ELM is a kind of easy to use, effective Single hidden layer feedforward neural networks (SLFNs) learning algorithm.It is in abnormal inspection
Surveying in classification has following advantage: (1) ELM is between input layer and hidden layer using random weight.We can repeatedly instruct
Practice identical data set, this output space different to different niceties of grading.(2) ELM is a simpler feed forward neural
Learning Algorithms.Traditional Learning Algorithm (such as BP algorithm) needs artificially to be arranged a large amount of network training parameter,
It can thus be very easy to generate locally optimal solution.And ELM is during determining network parameter, it is only necessary to the hidden of network be arranged
Node layer number does not need the input weight of adjustment network and the biasing of hidden member during the execution of the algorithm, and generates only
One optimal solution.Therefore, ELM pace of learning faster and Generalization Capability is more preferable than traditional artificial neural network, can be quickly real
The training and test of existing model.
The output of ELM isWherein: βiIt is between hidden node and output node
Weight, G (ai,bi, x) and it is hidden layer output function.H (x)=[G (a1,b1,x),...,G(aL,bL,x)]TIt is hidden layer relative to defeated
Enter the output vector of x.The key of ELM is to minimize training error and exports weight norm.Minimize
And | | β | |.
ELM algorithm is summarized as follows: given training set { (xi,ti)|xi∈Rn,ti∈Rm, i=1,2 ... N }, hidden node
Output function g (w, b, x) and the number of hidden nodes L.
(1) it is randomly assigned parameter (the w of hidden nodei,bi), i=1,2..., L.
(2) hidden layer output matrix Η is calculated.
(3) weight beta=H+T between hidden node and output node is calculated.
H+ is the Moore-Penrose generalized inverse matrix of hidden layer output matrix H, and orthographic projection, orthogonalization can be used
The methods of method and singular value decomposition are calculated.
Summary of the invention
The purpose of the present invention is being directed to existing Outlier Detection Algorithm, provide a kind of self-editing based on ELM-AE
The multilayer stochastic neural net list classification and Detection method of code device, is a kind of faster more efficient method for detecting abnormality.
The main implementation process of this patent algorithm is as follows: firstly, input only includes the training dataset of normal class；Input sample
Data are handled by multilayer ELM-AE self-encoding encoder coding and decoding, the characteristic value reconstructed；The characteristic value of reconstruct is inputted
The last layer ELM (does not include hidden layer), obtains reality output；Then, the range error of reality output and output label is obtained
Vector, and be ranked up from big to small, according to the threshold parameter of setting, determine the threshold value for separating normal class and exception class；Finally,
Test data (including normal class and exception class) is input in the multilayer stochastic neural net list abnormal classification detection model, is surveyed
Try the recognition effect of the model.
The invention mainly comprises the following steps:
Step 1, input training sample carry out feature normalization
1-1, a series of training samples are givenWhereinIndicate i-th of sample
This,It is indicated as target sample (normal data), training set only includes target sample, and N is total training samples number.
Step 2, sample data feature extraction
Training sample after 2-1, normalizationThat is ELM-AE's outputs and inputs matrix.
2-2, the random hidden layer that generates input weight matrixWith orthogonalization bias vector matrixInput data is mapped to same or different data dimension space: hk=g (akxα+bk), (ak)Tak=
I,(bk)Tbk=1, in which: g () indicates activation primitive,
(k=1,2 ..., K) it is ELM-AE number.
2-3, the output weight matrix for solving ELM-AE
Assuming that ELM-AE number is K, input and output layer neuronal quantity is d, and hidden layer neuron quantity isAnd it is every
The regularization parameter of a hidden layerIfOrI.e. for it is sparse and compression feature representation,IfI.e. for etc. dimensions Feature Mapping, βk=H-1XA,(βk)Tβk=I, (k
=1,2 ..., K).
Wherein:Indicate the hidden layer output matrix of ELM-AE.
Step 3, the output weight beta of classification learning calculate
3-1, output X'=[x ' is obtained by multilayer ELM-AE system1,x'2,...,x'N], input layer quantity is
D, ELM classify layer hidden neuron quantity be L (herein) and ELM classification layer regularization parameter C.IfIf
Hidden layer output node matrix isAnd
3-2, reality output is obtained
Step 4 calculates single classifier threshold θ
4-1, the error distance for calculating reality output Y and sample label T,
4-2, the error distance that will be obtainedIt is sorted from large to small, is obtainedIts
InWithRespectively indicate minimum and maximum error distance.
One 4-3, setting threshold parameter μ, obtaining threshold value is θ=εfloor(μ·N)。
Step 5, input test data are tested
5-1, a series of test samples are givenWhereinIndicate i-th of sample
This,Indicate that it is target sample (normal data),It is indicated as non-targeted samples (exceptional sample), P is in total
Test sample quantity.
5-2, be input to multilayer ELM-AE obtain every layer outputEnable the last layer ELM-AE output be
Y'=[y '1,y'2,...,y'N]。
5-3, be input to ELM classification layer obtain Yβ=β Y' calculates the error distance of reality output and sample label T
5-4, the error distance that will be obtainedCompared with single classifier threshold θ
The present invention has the beneficial effect that:
The present invention is mentioned using the feature that learning machine self-encoding encoder (ELM-AE) algorithm that transfinites carries out abnormality detection data
It takes, which is a kind of than commonly from encoding, (AE) algorithm is highly efficient from encryption algorithm, it carries out original self-encoding encoder
The BP gradient descent method of feature extraction optimization is changed to ELM, can quickly handle the input data of higher dimensional, extract its trunk
Partial information, and may be implemented initial data it is high-dimensional, etc. dimensions, low dimensional feature representation.Pass through multiple ELM-AE minds
Ability in feature extraction is enhanced through network superposition, especially for the unbalanced large data sets of higher-dimension in abnormality detection.Finally will
Extract obtained feature and detection identification carried out by ELM single classifier, the fast Generalization Capability of classification speed is more preferable, meet real-time with
Processing more typically property data demand.
The present invention is relative to the Outlier Detection Algorithm based on traditional neural network, support vector machines and arest neighbors, Ke Yigeng
Quickly and efficiently to extract main information and dimensionality reduction, then carry out identification classification.Faster, accuracy rate is higher for speed, and Generalization Capability is more
It is good.It is applicable not only to small data set and is applied equally to higher-dimension large data sets, there is universality.For practical application tool from now on
There is important meaning.
The present invention equally tests the abnormality detection effect of field measurement signal, has stronger anti-interference under complex situations
Ability and real-time, then will have better adaptability and accuracy rate, base for abnormal data set processing ideally
The processing of the abnormal data to every field from now on is provided in the multilayer stochastic neural net algorithm that ELM-AE and ELM are combined
Huge help.
This is a kind of highly effective from encryption algorithm to the present invention, and realization quickly and effectively mentions in higher-dimension and large data sets
Take useful feature.Sample data is handled by coding and decoding, if reconstructed error is sufficiently small, is being limited in range, can recognized
Fixed this coding code is the effective expression to input sample data.Later data are reconstructed by single abnormal classification based on ELM
Detection algorithm obtains output model, the algorithm only single class normal data of training, export the high Generalization Capability of recognition accuracy compared with
Good single disaggregated model.Normal and abnormal data, while the neural network mould of multilayer ELM-AE superposition can preferably be distinguished
Type can preferably be applied in higher-dimension large data sets.
Detailed description of the invention
Fig. 1 is flow diagram of the present invention；
Fig. 2 is Single hidden layer feedforward neural networks schematic diagram；
Fig. 3 is ELM-AE network structure；
Fig. 4 is ML-OCELM network structure.
Specific embodiment
The invention will be further described with example with reference to the accompanying drawing.
As shown in Figure 1, taking training data (only normal data set) to be input to multilayer ELM-AE first carries out feature extraction, then
Classification output actual result is carried out by ELM classification layer (no hidden layer), according to the mistake of obtained reality output and sample label
Difference sequence, obtains threshold value by threshold parameter.Sample to be tested is fed into trained abnormality detection model later, is surveyed
The error for trying data reality output and sample label, is classified as exception class greater than threshold value, is normal class less than or equal to threshold value, and
Calculate accuracy rate.
Fig. 2 shows the basic structure of Single hidden layer feedforward neural networks, that is, the frame of ELM algorithm.Fig. 3 shows list
The network structure of a ELM-AE, that is, self-encoding encoder use ELM algorithm as optimization algorithm, accelerate the instruction of self-encoding encoder
Practice speed, enhances generalization ability.Fig. 4 is the network structure that the present invention uses method, and three layers of ELM-AE heap poststack pass through again
One ELM classification layer without hidden layer.
The purpose of the present invention is being directed to existing Outlier Detection Algorithm, provide a kind of self-editing based on ELM
The multilayer stochastic neural net list classification and Detection algorithm of code device, is a kind of faster more efficient method for detecting abnormality.
The invention mainly comprises the following steps:
Step 1, input training sample carry out feature normalization
1-2, a series of training sample { (xi are givenα,tiα)xiα∈Rn,ti α∈Rm, i=1,2 ... N }, wherein xi αIt indicates
I-th of sample, ti α=1 indicates that it is target sample (normal data), and training set only includes target sample, and N is total training sample
This quantity.
Step 2, sample data feature extraction
Training sample after 2-1, normalizationThat is ELM-AE's outputs and inputs matrix.
2-2, the random hidden layer that generates input weight matrixWith orthogonalization bias vector matrixInput data is mapped to same or different data dimension space: hk=g (akxα+bk), (ak)Tak=
I,(bk)Tbk=1, in which: g () indicates that activation primitive, (k=1,2 ..., K) are ELM-AE number.
2-3, the output weight matrix for solving ELM-AE
Assuming that ELM-AE number is K, input and output layer neuronal quantity is d, and hidden layer neuron quantity isAnd it is every
The regularization parameter of a hidden layerIfOrI.e. for it is sparse and compression feature representation,IfI.e. for etc. dimensions Feature Mapping, βk=H-1XA,(βk)Tβk=I, (k=
1,2,...,K)。
Wherein:Indicate the hidden layer output matrix of ELM-AE.
Step 3, the output weight beta of classification learning calculate
3-1, output X'=[x ' is obtained by multilayer ELM-AE system1,x'2,...,x'N], input layer quantity is
D, ELM classify layer hidden neuron quantity be L (herein) and ELM classification layer regularization parameter C.IfIf
Hidden layer output node matrix isAnd
3-2, reality output is obtained
Step 4 calculates single classifier threshold θ
4-1, the error distance for calculating reality output Y and sample label T,
4-2, the error distance that will be obtainedIt is sorted from large to small, is obtainedIts
InWithRespectively indicate minimum and maximum error distance.
One 4-3, setting threshold parameter μ, obtaining threshold value is θ=εfloor(μ·N)。
Step 5, input test data are tested
5-1, a series of test samples are givenWhereinIndicate i-th of sample
This,Indicate that it is target sample (normal data),It is indicated as non-targeted samples (exceptional sample), P is in total
Test sample quantity.
5-2, be input to multilayer ELM-AE obtain every layer outputEnable the last layer ELM-AE output be
Y'=[y '1,y'2,...,y'N]。
5-3, be input to ELM classification layer obtain Yβ=β Y' calculates the error distance of reality output and sample label T
5-4, the error distance that will be obtainedCompared with single classifier threshold θ
Claims (2)
1. being based on multilayer stochastic neural net single classifier method for detecting abnormality, it is characterised in that input only includes the instruction of normal class
Practice data set；Input sample data are handled by multilayer ELM-AE self-encoding encoder coding and decoding, the characteristic value reconstructed；It will
The characteristic value input the last layer ELM of reconstruct obtains reality output；Then reality output will be obtained to miss at a distance from output label
Difference vector is ranked up from big to small, according to the threshold parameter of setting, determines the threshold value for separating normal class and exception class；Finally,
Test data is input in the multilayer stochastic neural net list abnormal classification detection model, the recognition effect of the model is tested.
2. according to claim 1 be based on multilayer stochastic neural net single classifier method for detecting abnormality, it is characterised in that
Specifically comprise the following steps:
Step 1, input training sample carry out feature normalization
1-1, a series of training samples are givenWhereinIndicate i-th of sample,It is indicated as target sample, training set only includes target sample, and N is total training samples number；
Step 2, sample data feature extraction
Training sample after 2-1, normalizationThat is ELM-AE's outputs and inputs matrix；
2-2, the random hidden layer that generates input weight matrixWith orthogonalization bias vector matrix
Input data is mapped to same or different data dimension space: hk=g (akxα+bk), (ak)Tak=I, (bk)Tbk=1,
Wherein: g () indicates that activation primitive, (k=1,2 ..., K) are ELM-AE number；
2-3, the output weight matrix for solving ELM-AE
Assuming that ELM-AE number is K, input and output layer neuronal quantity is d, and hidden layer neuron quantity isAnd it is each hidden
Regularization parameter containing layerIfOrI.e. for it is sparse and compression feature representation,IfI.e. for etc. dimensions Feature Mapping, βk=H-1XA,(βk)Tβk=I, (k=
1,2,...,K)；
Wherein:Indicate the hidden layer output matrix of ELM-AE；
Step 3, the output weight beta of classification learning calculate
3-1, output X'=[x ' is obtained by multilayer ELM-AE system1,x′2,...,x′N], input layer quantity is d,
ELM classify layer hidden neuron quantity be L (herein) and ELM classification layer regularization parameter C；IfIf
Hidden layer output node matrix isAnd
3-2, reality output is obtained
Step 4 calculates single classifier threshold θ
4-1, the error distance for calculating reality output Y and sample label T,
4-2, the error distance that will be obtainedIt is sorted from large to small, is obtainedWherein
WithRespectively indicate minimum and maximum error distance；
One 4-3, setting threshold parameter μ, obtaining threshold value is θ=εfloor(μ·N)；
Step 5, input test data are tested
5-1, a series of test samples are givenWhereinIndicate i-th of sample,Indicate that it is target sample,It is indicated as non-targeted samples, P is test sample quantity in total；
5-2, be input to multilayer ELM-AE obtain every layer outputEnabling the last layer ELM-AE output is Y'=
[y′1,y′2,...,y′N]；
5-3, be input to ELM classification layer obtain Yβ=β Y' calculates the error distance of reality output and sample label T
5-4, the error distance that will be obtainedCompared with single classifier threshold θ；
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811305622.XA CN109858509A (en) | 2018-11-05 | 2018-11-05 | Based on multilayer stochastic neural net single classifier method for detecting abnormality |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811305622.XA CN109858509A (en) | 2018-11-05 | 2018-11-05 | Based on multilayer stochastic neural net single classifier method for detecting abnormality |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109858509A true CN109858509A (en) | 2019-06-07 |
Family
ID=66889834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811305622.XA Pending CN109858509A (en) | 2018-11-05 | 2018-11-05 | Based on multilayer stochastic neural net single classifier method for detecting abnormality |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858509A (en) |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110381079A (en) * | 2019-07-31 | 2019-10-25 | 福建师范大学 | Network log method for detecting abnormality is carried out in conjunction with GRU and SVDD |
| CN110701087A (en) * | 2019-09-25 | 2020-01-17 | 杭州电子科技大学 | Axial flow compressor pneumatic instability detection method based on single-classification overrun learning machine |
| CN110941542A (en) * | 2019-11-06 | 2020-03-31 | 国网上海市电力公司 | Sequence integration high-dimensional data anomaly detection system and method based on elastic network |
| CN110990835A (en) * | 2019-11-22 | 2020-04-10 | 南京邮电大学 | Neural network Trojan horse detection method based on sample judgment error |
| CN111136509A (en) * | 2020-01-10 | 2020-05-12 | 大连理工大学 | Cutter state monitoring method based on ELM-SDAE algorithm |
| CN111159961A (en) * | 2019-12-13 | 2020-05-15 | 杭州广立微电子有限公司 | Abnormity detection method and system based on curve data |
| CN111415032A (en) * | 2020-03-03 | 2020-07-14 | 东华大学 | Method for predicting production performance of polyester fiber precursor based on E L M-AE of transfer learning |
| CN111428772A (en) * | 2020-03-19 | 2020-07-17 | 南京邮电大学 | Photovoltaic system depth anomaly detection method based on k-nearest neighbor adaptive voting |
| CN111538776A (en) * | 2020-03-19 | 2020-08-14 | 浙江中海达空间信息技术有限公司 | Multilayer cognitive constraint high-dimensional geographic spatial data focusing visualization method |
| CN112037929A (en) * | 2020-09-07 | 2020-12-04 | 重庆大学 | Classification method based on multi-modal machine learning, online new coronary pneumonia early warning model training method and early warning method |
| CN112086100A (en) * | 2020-08-17 | 2020-12-15 | 杭州电子科技大学 | Quantization error entropy based urban noise identification method of multilayer random neural network |
| CN112393860A (en) * | 2020-12-09 | 2021-02-23 | 杭州拓深科技有限公司 | Fire-fighting pipe network water leakage monitoring method based on in-class distance constraint self-encoder |
| CN112598111A (en) * | 2020-12-04 | 2021-04-02 | 光大科技有限公司 | Abnormal data identification method and device |
| CN112614298A (en) * | 2020-12-09 | 2021-04-06 | 杭州拓深科技有限公司 | Composite smoke sensation monitoring method based on intra-class interaction constraint layering single classification |
| CN113112011A (en) * | 2020-01-13 | 2021-07-13 | 中移物联网有限公司 | Data prediction method and device |
| CN113421176A (en) * | 2021-07-16 | 2021-09-21 | 昆明学院 | Intelligent abnormal data screening method |
| CN113470316A (en) * | 2021-05-24 | 2021-10-01 | 电子科技大学 | Debris flow monitoring and early warning method based on self-coding single classification model |
| CN113469257A (en) * | 2021-07-07 | 2021-10-01 | 云南大学 | Distribution transformer fault detection method and system |
| CN113516180A (en) * | 2021-06-25 | 2021-10-19 | 重庆邮电大学 | Method for identifying Z-Wave intelligent equipment |
| CN113643722A (en) * | 2021-08-27 | 2021-11-12 | 杭州电子科技大学 | Urban noise identification method based on multilayer matrix stochastic neural network |
| CN113688036A (en) * | 2021-08-13 | 2021-11-23 | 北京灵汐科技有限公司 | Data processing method, device, equipment and storage medium |
| CN114036973A (en) * | 2021-10-12 | 2022-02-11 | 浙江工业大学 | Series arc fault identification method of extreme learning machine based on dynamic online sequence |
| CN115293244A (en) * | 2022-07-15 | 2022-11-04 | 北京航空航天大学 | Smart grid false data injection attack detection method based on signal processing and data reduction |
| CN115714731A (en) * | 2022-09-27 | 2023-02-24 | 中国人民解放军63921部队 | Deep space measurement and control link abnormity detection method based on deep learning self-encoder |
| CN116383723A (en) * | 2023-06-06 | 2023-07-04 | 中国地质调查局水文地质环境地质调查中心 | Debris flow data anomaly identification method, computer equipment and medium |
| CN116743646A (en) * | 2023-08-15 | 2023-09-12 | 云南省交通规划设计研究院有限公司 | Tunnel network anomaly detection method based on domain self-adaptive depth self-encoder |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104156560A (en) * | 2014-07-12 | 2014-11-19 | 中国矿业大学 | Multi-level coal mine water inrush prediction method based on SaE-ELM (self-adaptive evolutionary extreme learning machine) |
| CN107085704A (en) * | 2017-03-27 | 2017-08-22 | 杭州电子科技大学 | Fast face expression recognition method based on ELM own coding algorithms |
| CN107154923A (en) * | 2016-03-04 | 2017-09-12 | 中国矿业大学 | A kind of network inbreak detection method based on the very fast learning machine of multilayer |
| CN107451278A (en) * | 2017-08-07 | 2017-12-08 | 北京工业大学 | Chinese Text Categorization based on more hidden layer extreme learning machines |
| CN107610692A (en) * | 2017-09-22 | 2018-01-19 | 杭州电子科技大学 | The sound identification method of self-encoding encoder multiple features fusion is stacked based on neutral net |
-
2018
- 2018-11-05 CN CN201811305622.XA patent/CN109858509A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104156560A (en) * | 2014-07-12 | 2014-11-19 | 中国矿业大学 | Multi-level coal mine water inrush prediction method based on SaE-ELM (self-adaptive evolutionary extreme learning machine) |
| CN107154923A (en) * | 2016-03-04 | 2017-09-12 | 中国矿业大学 | A kind of network inbreak detection method based on the very fast learning machine of multilayer |
| CN107085704A (en) * | 2017-03-27 | 2017-08-22 | 杭州电子科技大学 | Fast face expression recognition method based on ELM own coding algorithms |
| CN107451278A (en) * | 2017-08-07 | 2017-12-08 | 北京工业大学 | Chinese Text Categorization based on more hidden layer extreme learning machines |
| CN107610692A (en) * | 2017-09-22 | 2018-01-19 | 杭州电子科技大学 | The sound identification method of self-encoding encoder multiple features fusion is stacked based on neutral net |
Non-Patent Citations (3)
| Title |
|---|
| LIYANAARACHCHI LEKAMALAGE CHAMARA KASUN ET AL: "Representational Learning with Extreme Learning Machine for Big Data", 《IEEE INTELLIGENT SYSTEMS》 * |
| QIAN LENG ET AL: "One-Class Classification with Extreme Learning Machine", 《MATHEMATICAL PROBLEMS IN ENGINEERING》 * |
| VIJAY MANIKANDAN JANAKIRAMAN ET AL: "Anomaly Detection in Aviation Data using Extreme Learning Machines", 《2016 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)》 * |
Cited By (43)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110381079A (en) * | 2019-07-31 | 2019-10-25 | 福建师范大学 | Network log method for detecting abnormality is carried out in conjunction with GRU and SVDD |
| CN110701087A (en) * | 2019-09-25 | 2020-01-17 | 杭州电子科技大学 | Axial flow compressor pneumatic instability detection method based on single-classification overrun learning machine |
| CN110941542A (en) * | 2019-11-06 | 2020-03-31 | 国网上海市电力公司 | Sequence integration high-dimensional data anomaly detection system and method based on elastic network |
| CN110941542B (en) * | 2019-11-06 | 2023-08-25 | 国网上海市电力公司 | Sequence integration high-dimensional data anomaly detection system and method based on elastic network |
| CN110990835A (en) * | 2019-11-22 | 2020-04-10 | 南京邮电大学 | Neural network Trojan horse detection method based on sample judgment error |
| CN110990835B (en) * | 2019-11-22 | 2022-10-14 | 南京邮电大学 | Neural network Trojan horse detection method based on sample judgment error |
| CN111159961B (en) * | 2019-12-13 | 2024-02-02 | 杭州广立微电子股份有限公司 | Anomaly detection method and system based on curve data |
| CN111159961A (en) * | 2019-12-13 | 2020-05-15 | 杭州广立微电子有限公司 | Abnormity detection method and system based on curve data |
| CN111136509A (en) * | 2020-01-10 | 2020-05-12 | 大连理工大学 | Cutter state monitoring method based on ELM-SDAE algorithm |
| CN113112011B (en) * | 2020-01-13 | 2024-02-27 | 中移物联网有限公司 | Data prediction method and device |
| CN113112011A (en) * | 2020-01-13 | 2021-07-13 | 中移物联网有限公司 | Data prediction method and device |
| CN111415032B (en) * | 2020-03-03 | 2022-04-29 | 东华大学 | Method for predicting production performance of polyester fiber protofilament based on ELM-AE of transfer learning |
| CN111415032A (en) * | 2020-03-03 | 2020-07-14 | 东华大学 | Method for predicting production performance of polyester fiber precursor based on E L M-AE of transfer learning |
| CN111538776A (en) * | 2020-03-19 | 2020-08-14 | 浙江中海达空间信息技术有限公司 | Multilayer cognitive constraint high-dimensional geographic spatial data focusing visualization method |
| CN111428772A (en) * | 2020-03-19 | 2020-07-17 | 南京邮电大学 | Photovoltaic system depth anomaly detection method based on k-nearest neighbor adaptive voting |
| CN111428772B (en) * | 2020-03-19 | 2022-08-23 | 南京邮电大学 | Photovoltaic system depth anomaly detection method based on k-nearest neighbor adaptive voting |
| CN112086100A (en) * | 2020-08-17 | 2020-12-15 | 杭州电子科技大学 | Quantization error entropy based urban noise identification method of multilayer random neural network |
| CN112086100B (en) * | 2020-08-17 | 2022-12-02 | 杭州电子科技大学 | Quantization error entropy based urban noise identification method of multilayer random neural network |
| CN112037929B (en) * | 2020-09-07 | 2024-11-08 | 重庆大学 | Classification method based on multi-modal machine learning, online new coronaries pneumonia early warning model training method and early warning method |
| CN112037929A (en) * | 2020-09-07 | 2020-12-04 | 重庆大学 | Classification method based on multi-modal machine learning, online new coronary pneumonia early warning model training method and early warning method |
| CN112598111A (en) * | 2020-12-04 | 2021-04-02 | 光大科技有限公司 | Abnormal data identification method and device |
| CN112598111B (en) * | 2020-12-04 | 2024-09-20 | 光大科技有限公司 | Abnormal data identification method and device |
| CN112393860B (en) * | 2020-12-09 | 2022-12-20 | 杭州拓深科技有限公司 | Fire-fighting pipe network water leakage monitoring method based on in-class distance constraint self-encoder |
| CN112614298A (en) * | 2020-12-09 | 2021-04-06 | 杭州拓深科技有限公司 | Composite smoke sensation monitoring method based on intra-class interaction constraint layering single classification |
| CN112393860A (en) * | 2020-12-09 | 2021-02-23 | 杭州拓深科技有限公司 | Fire-fighting pipe network water leakage monitoring method based on in-class distance constraint self-encoder |
| CN113470316A (en) * | 2021-05-24 | 2021-10-01 | 电子科技大学 | Debris flow monitoring and early warning method based on self-coding single classification model |
| CN113470316B (en) * | 2021-05-24 | 2022-03-25 | 电子科技大学 | Debris flow monitoring and early warning method based on self-coding single classification model |
| CN113516180B (en) * | 2021-06-25 | 2022-07-12 | 重庆邮电大学 | Method for identifying Z-Wave intelligent equipment |
| CN113516180A (en) * | 2021-06-25 | 2021-10-19 | 重庆邮电大学 | Method for identifying Z-Wave intelligent equipment |
| CN113469257A (en) * | 2021-07-07 | 2021-10-01 | 云南大学 | Distribution transformer fault detection method and system |
| CN113421176B (en) * | 2021-07-16 | 2022-11-01 | 昆明学院 | Intelligent screening method for abnormal data in student score scores |
| CN113421176A (en) * | 2021-07-16 | 2021-09-21 | 昆明学院 | Intelligent abnormal data screening method |
| CN113688036A (en) * | 2021-08-13 | 2021-11-23 | 北京灵汐科技有限公司 | Data processing method, device, equipment and storage medium |
| CN113643722B (en) * | 2021-08-27 | 2024-04-19 | 杭州电子科技大学 | Urban noise identification method based on multilayer matrix random neural network |
| CN113643722A (en) * | 2021-08-27 | 2021-11-12 | 杭州电子科技大学 | Urban noise identification method based on multilayer matrix stochastic neural network |
| CN114036973A (en) * | 2021-10-12 | 2022-02-11 | 浙江工业大学 | Series arc fault identification method of extreme learning machine based on dynamic online sequence |
| CN115293244A (en) * | 2022-07-15 | 2022-11-04 | 北京航空航天大学 | Smart grid false data injection attack detection method based on signal processing and data reduction |
| CN115293244B (en) * | 2022-07-15 | 2023-08-15 | 北京航空航天大学 | Smart grid false data injection attack detection method based on signal processing and data reduction |
| CN115714731B (en) * | 2022-09-27 | 2023-06-27 | 中国人民解放军63921部队 | Deep space measurement and control link abnormality detection method based on deep learning self-encoder |
| CN115714731A (en) * | 2022-09-27 | 2023-02-24 | 中国人民解放军63921部队 | Deep space measurement and control link abnormity detection method based on deep learning self-encoder |
| CN116383723A (en) * | 2023-06-06 | 2023-07-04 | 中国地质调查局水文地质环境地质调查中心 | Debris flow data anomaly identification method, computer equipment and medium |
| CN116743646B (en) * | 2023-08-15 | 2023-12-19 | 云南省交通规划设计研究院股份有限公司 | Tunnel network anomaly detection method based on domain self-adaptive depth self-encoder |
| CN116743646A (en) * | 2023-08-15 | 2023-09-12 | 云南省交通规划设计研究院有限公司 | Tunnel network anomaly detection method based on domain self-adaptive depth self-encoder |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109858509A (en) | Based on multilayer stochastic neural net single classifier method for detecting abnormality | |
| McBrearty et al. | Pairwise association of seismic arrivals with convolutional neural networks | |
| Chatterjee et al. | Extraction of binary black hole gravitational wave signals from detector data using deep learning | |
| CN112508105B (en) | Fault detection and retrieval method for oil extraction machine | |
| CN110287983A (en) | Based on maximal correlation entropy deep neural network single classifier method for detecting abnormality | |
| CN112801169B (en) | Camouflage target detection method, system, device and storage medium based on improved YOLO algorithm | |
| WO2019179403A1 (en) | Fraud transaction detection method based on sequence width depth learning | |
| CN107784276B (en) | Microseismic event identification method and device | |
| CN111562612B (en) | Deep learning microseismic event identification method and system based on attention mechanism | |
| CN102158486A (en) | Method for rapidly detecting network invasion | |
| CN109919241A (en) | Hyperspectral unknown class target detection method based on probability model and deep learning | |
| US20190034786A1 (en) | System and Method for Making Nuclear Radiation Detection Decisions and/or Radionuclide Identification Classifications | |
| Shoohi et al. | DCGAN for Handling Imbalanced Malaria Dataset based on Over-Sampling Technique and using CNN. | |
| CN106250918B (en) | A kind of mixed Gauss model matching process based on improved soil-shifting distance | |
| GB2616996A (en) | Method for recognizing type of vortex signal of evaporator of nuclear power plant on basis of LSTM-CNN | |
| CN106941490A (en) | Online network flow abnormal detecting method based on bidirectional two-dimensional principal component analysis | |
| CN112836719A (en) | Indicator diagram similarity detection method fusing two classifications and three groups | |
| CN117709454A (en) | Unbiased scene graph construction method, system and equipment based on generated template | |
| CN104980442A (en) | Network intrusion detection method based on element sample sparse representation | |
| CN111786999A (en) | Intrusion behavior detection method, device, equipment and storage medium | |
| Ghorpade et al. | 3D shape representation with spatial probabilistic distribution of intrinsic shape keypoints | |
| CN117011577A (en) | Image classification method, apparatus, computer device and storage medium | |
| CN110276358A (en) | High similarity wooden unit cross section detection method under intensive stacking | |
| Music et al. | Classification of leaf type using multilayer perceptron, naive bayes and support vector machine classifiers | |
| Abdel-Azim et al. | Textile defects identification based on neural networks and mutual information |
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 |
Application publication date: 20190607 |
|
| RJ01 | Rejection of invention patent application after publication |