CN107220671B - Artificial olfaction system online correction sample generation method based on self-organizing map - Google Patents

Artificial olfaction system online correction sample generation method based on self-organizing map Download PDF

Info

Publication number
CN107220671B
CN107220671B CN201710393408.3A CN201710393408A CN107220671B CN 107220671 B CN107220671 B CN 107220671B CN 201710393408 A CN201710393408 A CN 201710393408A CN 107220671 B CN107220671 B CN 107220671B
Authority
CN
China
Prior art keywords
online
self
neuron
neural network
organizing map
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.)
Active
Application number
CN201710393408.3A
Other languages
Chinese (zh)
Other versions
CN107220671A (en
Inventor
刘涛
李东琦
陈建军
武萌雅
陈艳兵
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.)
Chongqing University
Original Assignee
Chongqing University
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 Chongqing University filed Critical Chongqing University
Priority to CN201710393408.3A priority Critical patent/CN107220671B/en
Publication of CN107220671A publication Critical patent/CN107220671A/en
Application granted granted Critical
Publication of CN107220671B publication Critical patent/CN107220671B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0001—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00 by organoleptic means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Chemical & Material Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Food Science & Technology (AREA)
  • Medicinal Chemistry (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analyzing Materials By The Use Of Electric Means (AREA)
  • Other Investigation Or Analysis Of Materials By Electrical Means (AREA)

Abstract

The invention belongs to the technical field of odor analysis, and discloses an artificial olfaction system online correction sample generation method based on a self-organizing map, which comprises an initial training stage and an online updating stage; in the initial training stage, a self-organizing map neural network multilayer structure is constructed according to the sample category number, the neural network neuron weight of the self-organizing map is initialized according to the training samples, and the neural network neuron weight of the self-organizing map is used as an initial training sample set; and in the online updating stage, the weight of the neuron in the local region is adjusted by using the test sample according to the classification result of the subsequent classifier. And taking the neural network neuron weight of the self-organizing map at the moment as an online training sample set to perform online correction on the mode identification method. The result shows that the long-term drift resistance of the artificial olfaction system can be improved under the online working condition; the correction sample can be automatically generated in the online working process, and the guarantee is provided for the automatic online correction of the artificial olfaction system.

Description

Artificial olfaction system online correction sample generation method based on self-organizing map
Technical Field
The invention belongs to the technical field of odor analysis, and particularly relates to an artificial olfaction system online correction sample generation method based on a self-organizing map.
Background
The artificial olfaction system is a novel odor analysis means and has the advantages of rapid detection, no wound, simple and convenient operation, low cost and the like. The artificial olfaction system is mainly divided into two parts, namely a gas sensor array and a mode identification method, wherein the gas sensor array adopts a low-cost gas sensor with cross sensitivity to obtain an odor map; the mode identification adopts methods such as artificial intelligence, machine learning and the like to carry out qualitative and quantitative analysis on the odor. The long-term drift of the gas sensor is a problem that an artificial olfactory system cannot avoid, and the odor spectrum of the gas sensor array can slowly and irregularly change along with the prolonging of the use time; the recognition accuracy of the mode recognition method is gradually reduced along with time, and finally the mode recognition method becomes unreliable; data acquisition of specific samples is carried out on an artificial olfactory system regularly, and pattern recognition algorithm correction is carried out according to the acquired data so that the algorithm is matched with the odor map subjected to drift. However, the conventional calibration method needs to stop the normal work of the artificial olfactory system, needs to configure and purchase a special test sample, needs a professional operator to regulate and control, and has high comprehensive cost. It is not feasible to disassemble a system that is mobile and requires on-line monitoring.
In summary, the problems of the prior art are as follows: 1) for equipment needing long-term online work, conventional correction cannot be carried out, so that the system identification accuracy rate can be obviously reduced along with the working time; 2) the conventional calibration method needs to prepare a specific calibration sample and cannot work normally during the calibration of the equipment, thereby wasting a great deal of financial resources, manpower and material resources.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides an artificial olfaction system online correction sample generation method based on a self-organizing map.
The invention is realized in this way, a self-organizing map based artificial olfaction system online correction sample generation method, the self-organizing map based artificial olfaction system online correction sample generation method includes an initial training phase and an online updating phase;
in the initial training stage, each detection is expressed as a p-dimensional vector and comprises K samples, a K-layer self-organizing map neural network is constructed, each layer of the neural network is composed of neurons to form a two-dimensional plane with an N × M structure, the number of the neurons in the neural network is K × N × M, and the weight matrix of the neural network is made to be
Figure BDA0001308032890000021
Wherein
Figure BDA0001308032890000022
Representing a p-dimensional weight vector with (N, M) layer plane coordinates, K ∈ {1, …, K }, N ∈ {1, …, N }, M ∈ {1, …, M }, and when a certain p-dimensional training sample x is trainedtAfter entering the network, initializing;
after the initial training is completed in the online updating stage, testing is carried out, and a sample x is obtainedpThe class is unknown, and the subsequent classifier follows the online training sample pair x in WpAnd (5) classifying, wherein the classification result of the subsequent classifier is the kth class.
Further, the initializing specifically includes:
1) from training samples xtClass (D) of
Figure BDA0001308032890000023
Calculate the first
Figure BDA0001308032890000024
All neuron weights and x of layertThe distance of (c):
Figure BDA0001308032890000025
taking the neuron with the minimum distance as a winning neuron, and saving the positions of the neurons:
Figure BDA0001308032890000026
wherein the content of the first and second substances,
Figure BDA0001308032890000027
and
Figure BDA0001308032890000028
respectively winning neurons in the first
Figure BDA0001308032890000029
Two-dimensional coordinates of the layer;
2) to the first
Figure BDA00013080328900000210
The layer neuron weights are iterated as follows:
Figure BDA00013080328900000211
where α∈ (0,1) is the learning rate, defined as follows:
Figure BDA00013080328900000212
3) repeating 1) and 2) until all training samples enter the neural network for initial training;
4) repeating 1) -3), and enabling the weight distribution of the neural network of the self-organizing map to be close to that of the training sample.
Further, the online update stage specifically includes:
1) calculating all neuron weights and xpThe distance of (c):
Figure BDA0001308032890000031
taking the neuron with the minimum distance as a winning neuron, and saving the position of the neuron:
Figure BDA0001308032890000032
wherein
Figure BDA0001308032890000033
To be the number of the layer where the winning neuron is located,
Figure BDA0001308032890000034
and
Figure BDA0001308032890000035
respectively winning neurons in the first
Figure BDA0001308032890000036
Two-dimensional coordinates of the layer;
2) iterating the k-th layer neuron weight as follows:
Figure BDA0001308032890000037
where α∈ (0,1) is the learning rate, defined as follows:
Figure BDA0001308032890000038
3) and (3) training the classifier by using the weight matrix W of the neural network or carrying out next classification by using an 'online training sample' in the weight matrix W.
The invention also aims to provide an artificial olfactory system applying the artificial olfactory system on-line correction sample generation method based on the self-organizing map.
Another object of the present invention is to provide a gas sensor applying the method for online correction of sample generation by the artificial olfaction system based on self-organizing maps.
The invention has the advantages and positive effects that: the results in table 2 show that the artificial olfactory system can complete the modification of the pattern recognition method without the conventional correction by using the method of the invention, and compared with the case without the conventional correction, the system recognition accuracy rate is increased from 42.56% to 48.51% in the multi-class classification application.
Drawings
Fig. 1 is a flow chart of a sample generation method for online correction of an artificial olfactory system based on a self-organizing map according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The following detailed description of the principles of the invention is provided in connection with the accompanying drawings.
As shown in fig. 1, the method for generating the sample for online correction of the artificial olfactory system based on the self-organizing map provided by the embodiment of the invention comprises the following steps:
s101: each detection is expressed as a vector, a self-organizing map neural network is constructed, and a training sample enters the neural network to be initialized;
s102: after the initial training is finished, testing is carried out, and a subsequent classifier is classified according to the on-line training sample pair.
The method for generating the on-line correction sample of the artificial olfactory system based on the self-organizing map comprises two stages of initial training and on-line updating.
1. Initial training
Each detection is expressed as a p-dimensional vector, and if K types of samples are contained, a K-layer Self-organized map (Self-organized map) neural network is constructed, each layer consists of neurons to form a two-dimensional plane with an N × M structure, so that the neural network has K × N × M total neurons, and the weight matrix of the neural network is set to be
Figure BDA0001308032890000041
Wherein
Figure BDA0001308032890000042
Represents a p-dimensional weight vector with (N, M) plane coordinates of the K-th layer, and K ∈ {1, …, K }, N ∈ {1, …, N }, M ∈ {1, …, M }.
When a certain p-dimension training sample xtAfter entering the network, initializing:
1) from training samples xtClass (D) of
Figure BDA0001308032890000043
Calculate the first
Figure BDA0001308032890000044
All neuron weights and x of layertThe distance of (c):
Figure BDA0001308032890000045
taking the neuron with the minimum distance as a winning neuron, and saving the position of the neuron:
Figure BDA0001308032890000046
wherein the content of the first and second substances,
Figure BDA0001308032890000051
and
Figure BDA0001308032890000052
respectively winning neurons in the first
Figure BDA0001308032890000053
Two-dimensional coordinates of the layer.
2) To the first
Figure BDA0001308032890000054
The layer neuron weights are iterated as follows:
Figure BDA0001308032890000055
where α∈ (0,1) is the learning rate, defined as follows:
Figure BDA0001308032890000056
3) and repeating the step 1) and the step 2) until all the training samples enter the neural network for initial training.
4) And (5) repeating the steps 1) to 3), and stopping after a certain cycle number is met.
At this time, the neural network weight matrix W is an initialized "online training sample" set, and the N × M neuron weights at the kth layer are the "online training sample" set of the kth class sample.
2. Online update
Completing the initial training toThen, enter the test, sample xpThe class is unknown, and the subsequent classifier pairs x according to the 'on-line training sample' in WpClassifying, and setting the classification result of the subsequent classifier as the kth class, wherein the classification method comprises the following steps of:
1) calculating all neuron weights and xpThe distance of (c):
Figure BDA0001308032890000057
taking the neuron with the minimum distance as a winning neuron, and saving the position of the neuron:
Figure BDA0001308032890000058
wherein
Figure BDA0001308032890000059
To be the number of the layer where the winning neuron is located,
Figure BDA00013080328900000510
and
Figure BDA00013080328900000511
respectively winning neurons in the first
Figure BDA00013080328900000512
Two-dimensional coordinates of the layer.
2) Iterating the k-th layer neuron weight as follows:
Figure BDA00013080328900000513
where α∈ (0,1) is the learning rate, defined as follows:
Figure BDA00013080328900000514
3) and (3) training the classifier by using the weight matrix W of the neural network or carrying out next classification by using an 'online training sample' in the weight matrix W.
The online updating keeps the neuron weight (namely an 'online training sample' set) synchronous with the current sample change, can provide support for the real-time updating of a subsequent classifier, and enables the classifier to be matched with the response of the artificial olfactory system after the drift is generated.
The effect of the present invention will be described in detail with reference to the detection.
Six kinds (i.e., K ═ 6) of gases, such as ammonia gas, acetaldehyde, acetone, ethylene, ethanol, and toluene, were detected using an artificial olfactory system including 16 metal oxide gas sensors. 2 steady-state features and 6 dynamic features are extracted from each sensor response curve in each detection, and each detection is characterized as a vector of 128 dimensions (p ═ 128). And (3) inhibiting drift, wherein the detection time span is 36 months, all detection data are arranged into 10 data sets, and the corresponding relation between the data contained in each data set and the detection time is shown in table 1. Taking N as M as 10, taking the data in the data set 1 as an initial training sample, taking the data in the data set 2-10 as a test sample, adopting a nearest neighbor method (KNN) as a classification method, and comparing the two conditions of adopting the method and not adopting the method, wherein the result is shown in Table 2. The results in table 2 show that the invention can improve the long-term drift resistance of the artificial olfactory system under on-line working conditions.
TABLE 1 data set and detection time
Data set numbering Time of detection
Data set 1 Month 1-2
Data set 2 Month 3-10
Data set 3 Month 11 to 13
Data set 4 Month 14-15
Data set 5 Month 16
Data set 6 Month 17-20
Data set 7 Month 21
Data set 8 Months 22-23
Data set 9 Month 24 to 35
Data set 10 Month 36
TABLE 2 test results
Figure BDA0001308032890000061
Figure BDA0001308032890000071
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims (4)

1. An artificial olfaction system online correction sample generation method based on a self-organizing map is characterized by comprising an initial training stage and an online updating stage;
in the initial training stage, each detection is expressed as a p-dimensional vector and comprises K samples, a K-layer self-organizing map neural network is constructed, each layer of the neural network is composed of neurons to form a two-dimensional plane with an N × M structure, the number of the neurons in the neural network is K × N × M, and the weight matrix of the neural network is made to be
Figure FDA0002424156560000011
Wherein
Figure FDA0002424156560000012
Representing a p-dimensional weight vector with (N, M) layer plane coordinates, K ∈ {1, …, K }, N ∈ {1, …, N }, M ∈ {1, …, M }, and when a certain p-dimensional training sample x is trainedtAfter entering the network, initializing;
after the initial training is completed in the online updating stage, testing is carried out, and a sample x is obtainedpThe class is unknown, and the subsequent classifier follows the online training sample pair x in WpClassifying, wherein the classification result of the subsequent classifier is the kth class;
the initialization specifically includes:
1) from training samples xtClass (D) of
Figure FDA0002424156560000013
Calculate the first
Figure FDA0002424156560000014
All neuron weights and x of layertThe distance of (c):
Figure FDA0002424156560000015
taking the neuron with the minimum distance as a winning neuron, and saving the positions of the neurons:
Figure FDA0002424156560000016
wherein the content of the first and second substances,
Figure FDA0002424156560000017
and
Figure FDA0002424156560000018
respectively winning neurons in the first
Figure FDA0002424156560000019
Two-dimensional coordinates of the layer;
2) to the first
Figure FDA00024241565600000110
Iteration is carried out on the layer neuron weight:
Figure FDA00024241565600000111
where α∈ (0,1) is the learning rate, defined as follows:
Figure FDA00024241565600000112
3) repeating 1) and 2) until all training samples enter the neural network for initial training;
4) repeating 1) -3), and stopping after a certain cycle number is met.
2. The self-organizing map-based artificial olfactory system online correction sample generation method of claim 1, wherein the online update phase specifically comprises:
1) calculating all neuron weights and xpThe distance of (c):
Figure FDA0002424156560000021
taking the neuron with the minimum distance as a winning neuron, and saving the position of the neuron:
Figure FDA0002424156560000022
wherein
Figure FDA0002424156560000023
To be the number of the layer where the winning neuron is located,
Figure FDA0002424156560000024
and
Figure FDA0002424156560000025
respectively winning neurons in the first
Figure FDA0002424156560000026
Two-dimensional coordinates of the layer;
2) iterating the k-th layer neuron weight as follows:
Figure FDA0002424156560000027
where α∈ (0,1) is the learning rate, defined as follows:
Figure FDA0002424156560000028
3) and training the classifier by using the weight matrix W of the neural network or carrying out next classification by using an online training sample in W.
3. An artificial olfaction system applying the artificial olfaction system based on the self-organizing map to correct the sample generation method on line according to any one of claims 1-2.
4. A gas sensor applying the method for generating the sample through online correction of the artificial olfaction system based on the self-organizing map as claimed in any one of claims 1-2.
CN201710393408.3A 2017-05-27 2017-05-27 Artificial olfaction system online correction sample generation method based on self-organizing map Active CN107220671B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710393408.3A CN107220671B (en) 2017-05-27 2017-05-27 Artificial olfaction system online correction sample generation method based on self-organizing map

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710393408.3A CN107220671B (en) 2017-05-27 2017-05-27 Artificial olfaction system online correction sample generation method based on self-organizing map

Publications (2)

Publication Number Publication Date
CN107220671A CN107220671A (en) 2017-09-29
CN107220671B true CN107220671B (en) 2020-07-14

Family

ID=59946994

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710393408.3A Active CN107220671B (en) 2017-05-27 2017-05-27 Artificial olfaction system online correction sample generation method based on self-organizing map

Country Status (1)

Country Link
CN (1) CN107220671B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109540978A (en) * 2018-12-13 2019-03-29 清华大学 Odor identification equipment
CN111914082A (en) * 2019-05-08 2020-11-10 天津科技大学 Online knowledge aggregation method based on SOM neural network algorithm

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102507676A (en) * 2011-11-01 2012-06-20 重庆大学 On-line drift compensation method of electronic nose based on multiple self-organizing neural networks
CN103489033A (en) * 2013-09-27 2014-01-01 南京理工大学 Incremental type learning method integrating self-organizing mapping and probability neural network
CN103499664A (en) * 2013-07-30 2014-01-08 中国标准化研究院 Method for extracting and researching atlas information representing differences of honey quality
CN105823801A (en) * 2016-03-03 2016-08-03 重庆大学 Deep belief network characteristic extraction-based electronic nose drift compensation method
CN105913079A (en) * 2016-04-08 2016-08-31 重庆大学 Target domain migration extreme learning-based electronic nose heterogeneous data identification method
CN106226464A (en) * 2016-07-04 2016-12-14 重庆大学 A kind of Electronic Nose gas detecting system with defencive function

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009013754A1 (en) * 2007-07-24 2009-01-29 Technion Research And Development Foundation Ltd. Chemically sensitive field effect transistors and use thereof in electronic nose devices

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102507676A (en) * 2011-11-01 2012-06-20 重庆大学 On-line drift compensation method of electronic nose based on multiple self-organizing neural networks
CN103499664A (en) * 2013-07-30 2014-01-08 中国标准化研究院 Method for extracting and researching atlas information representing differences of honey quality
CN103489033A (en) * 2013-09-27 2014-01-01 南京理工大学 Incremental type learning method integrating self-organizing mapping and probability neural network
CN105823801A (en) * 2016-03-03 2016-08-03 重庆大学 Deep belief network characteristic extraction-based electronic nose drift compensation method
CN105913079A (en) * 2016-04-08 2016-08-31 重庆大学 Target domain migration extreme learning-based electronic nose heterogeneous data identification method
CN106226464A (en) * 2016-07-04 2016-12-14 重庆大学 A kind of Electronic Nose gas detecting system with defencive function

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"A Novel Retraining Method of Multiple Self-Organizing Maps for Gas Sensor Drift Compensation. Sensors and Materials";Tao Liu, Kadri chaibou, Zhiyong Huang.;《Sensors and Materials》;20131231;第25卷(第2期);第109-120页 *
"一种基于多重自组织图的电子鼻漂移抑制方法";刘涛，黄智勇;《仪器仪表学报》;20120630;第33卷(第6期);第1287-1292页 *

Also Published As

Publication number Publication date
CN107220671A (en) 2017-09-29

Similar Documents

Publication Publication Date Title
CN106874581B (en) Building air conditioner energy consumption prediction method based on BP neural network model
Bharati et al. Deep learning techniques—R-CNN to mask R-CNN: a survey
CN102866179B (en) Online recognition and inhibition method based on non-target interference smell in electronic nose of artificial intelligent learning machine
CN105913079B (en) Electronic nose isomeric data recognition methods based on the study of the target domain migration limit
CN105938116A (en) Gas sensor array concentration detection method based on fuzzy division and model integration
CN110770752A (en) Automatic pest counting method combining multi-scale feature fusion network with positioning model
CN108173708A (en) Anomalous traffic detection method, device and storage medium based on incremental learning
CN107563999A (en) A kind of chip defect recognition methods based on convolutional neural networks
CN106022528A (en) Short-term power prediction method of photovoltaic power plant based on density peak hierarchical clustering
CN104317681A (en) Behavioral abnormality automatic detection method and behavioral abnormality automatic detection system aiming at computer system
CN109407649B (en) Fault type matching method based on fault characteristic variable selection
CN113569742B (en) Broadband electromagnetic interference source identification method based on convolutional neural network
CN109143408B (en) Dynamic region combined short-time rainfall forecasting method based on MLP
CN105956612A (en) Hyperspectral image classification method based on active learning and semi-supervised learning
CN108596274A (en) Image classification method based on convolutional neural networks
CN110543906B (en) Automatic skin recognition method based on Mask R-CNN model
CN105224960A (en) Based on the corn seed classification hyperspectral imagery model of cognition update method of clustering algorithm
CN107220671B (en) Artificial olfaction system online correction sample generation method based on self-organizing map
CN107145778B (en) Intrusion detection method and device
CN110880369A (en) Gas marker detection method based on radial basis function neural network and application
CN109242552B (en) Shop positioning method based on big data
CN112135248A (en) WIFI fingerprint positioning method based on K-means optimal estimation
CN116229380A (en) Method for identifying bird species related to bird-related faults of transformer substation
CN111598460A (en) Method, device and equipment for monitoring heavy metal content in soil and storage medium
Bhardwaj et al. Weather forecasting using soft computing techniques

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
GR01 Patent grant
GR01 Patent grant