CN102592171A - Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network - Google Patents

Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network Download PDF

Info

Publication number
CN102592171A
CN102592171A CN2011104521254A CN201110452125A CN102592171A CN 102592171 A CN102592171 A CN 102592171A CN 2011104521254 A CN2011104521254 A CN 2011104521254A CN 201110452125 A CN201110452125 A CN 201110452125A CN 102592171 A CN102592171 A CN 102592171A
Authority
CN
China
Prior art keywords
network
neural network
parameter
training
cognition
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
CN2011104521254A
Other languages
Chinese (zh)
Inventor
孙雁飞
亓晋
李施
朱磊
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.)
Nanjing Post and Telecommunication University
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing Post and Telecommunication 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 Nanjing Post and Telecommunication University filed Critical Nanjing Post and Telecommunication University
Priority to CN2011104521254A priority Critical patent/CN102592171A/en
Publication of CN102592171A publication Critical patent/CN102592171A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Data Exchanges In Wide-Area Networks (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method for predicting the cognitive network performance based on a BP (Back Propagation) neural network. According to the method, the predication of the cognitive network performance is carried out by using the BP neural network and taking a protocol stack parameter and a performance parameter as inputs. The invention also discloses a device for predicting the cognitive network performance based on the BP neural network. The device comprises an information perception module, a data preprocessing module and a prediction module which are sequentially connected according to the signal flow direction. The cognitive network performance at future time can be accurately predicted, the cognitive network can be self-adaptively reacted in a network environment and the controllability, the management and the reliability of the network are realized.

Description

Cognition network performance prediction method and device based on the BP neural network
Technical field
The present invention relates to a kind of cognition network performance prediction method and device, relate in particular to a kind of cognition network performance prediction method and device based on the BP neural network.
Background technology
Along with the increase of miscellaneous service and the raising of customer requirements, network management will be a difficult problem with control.In order better to manage and Control Network, improve network service quality (QoS), introduced the notion of cognition network.
Cognition network is on the basis of cognitive radio, to propose, the notion of cognition connected from wireless single-hop extend to whole network, and be a kind of network with cognitive function.It can the sensing network situation and makes a strategic decision in view of the above, reasoning, learns and take appropriate action.Cognition network is attempted intelligent decision is incorporated in the network; Make network have ability from management, self study, self-optimizing; Can be when being desirably in change of network environment through reasoning, adaptive active is made a response, thereby really realize network may command, can manage, trusted.
Traditional network can only passive the making a response in back occur at environment change or fault (problem); The superiority of cognition network is that it can self-adaptation initiatively make a response in change of network environment; Can, network condition just take measures before degenerating, to prevent the appearance of this situation.Therefore, concerning next constantly the evaluation prediction of network condition for very important the cognition network.Through the network performance parameter that known parameter reasoning draws, be the important evidence of network state assessment, also be the basis that cognition network is made decisions on one's own.
Summary of the invention
Technical matters to be solved by this invention is to overcome the deficiency of prior art, and a kind of cognition network performance prediction method and device based on the BP neural network is provided, and has the more accurate prediction result.
The following technical scheme of the concrete employing of the present invention:
Based on the cognition network performance prediction method of BP neural network, the network parameter current according to cognition network predicted the cognition network performance parameter in the moment in future; This method as input, utilizes the BP neural network to carry out the cognition network performance prediction with the protocol stack parameter of cognition network and performance parameter; Said BP neural network obtains according to following method training:
Step 1, the protocol stack parameter and the performance parameter of gathering cognition network, and it is carried out normalization handle, training sample obtained;
Step 2, three layers of BP neural network of structure, wherein the input layer number is N-1, output layer node number is 1; N is the protocol stack parameter of the cognition network gathered and the species number of performance parameter; The number of hidden nodes is confirmed according to following method: the average with error sum of squares is the accuracy evaluation index, through loop test, chooses the number of hidden nodes that makes that the error sum of squares average is minimum;
Step 3, initial parameter are selected;
Step 4, utilize the training sample that obtains in the step 1 three layers of constructed BP neural network to be trained the BP neural network after obtaining training.
Said normalization processing is meant and all sample datas is transformed in the scope of (0,1) according to following formula:
Figure 2011104521254100002DEST_PATH_IMAGE002
Wherein,
Figure 2011104521254100002DEST_PATH_IMAGE004
and
Figure 2011104521254100002DEST_PATH_IMAGE006
is respectively maximal value and minimum value in the raw sample data;
Figure 2011104521254100002DEST_PATH_IMAGE008
is raw sample data, be the numerical value after the conversion.
Said initial parameter comprises learning rate, initial weight, cycle index, and wherein the learning rate span is (0.01,0.8), and initial weight is chosen equally distributed random number between (1,1), and cycle index is confirmed by the loop test in the step 2.
Preferably, said protocol stack parameter comprises average queue length, end-to-end round-trip delay and tcp window size; Said performance parameter comprises time delay, shake, handling capacity, packet loss, number of dropped packets.
Said step 4 specifically comprises:
Step 401, will pass through pretreated sample and be input to the BP neural network;
Step 402, carry out the parameter initialization setting, comprise maximum frequency of training, training objective, the number of hidden nodes, initial weight, threshold value, initial learn speed;
Step 403, calculate the input and output value of each layer, calculate the error of each layer;
Step 404, according to the weights and the threshold value of adaptive learning rate algorithms correction neural network;
Step 405, as reaching training objective or frequency of training greater than preset maximum frequency of training, then stop training; As not, then return step 404.
Based on the cognition network performance prediction device of BP neural network, this device comprises by signal flow to the information sensing module, data preprocessing module, the prediction module that connect successively;
Said information sensing module is used for gathering in real time the protocol stack parameter and the performance parameter of cognition network, and with data transmission to data preprocessing module, said protocol stack parameter comprises average queue length, end-to-end round-trip delay and tcp window size; Said performance parameter comprises time delay, shake, handling capacity, packet loss, number of dropped packets;
The data that said data preprocessing module is carried out after normalization is handled and will be handled the data that receive are sent to prediction module; Said normalization processing is meant and all sample datas is transformed in the scope of (0,1) according to following formula:
Figure 257763DEST_PATH_IMAGE002
Wherein,
Figure 112587DEST_PATH_IMAGE004
and
Figure 331472DEST_PATH_IMAGE006
is respectively maximal value and minimum value in the raw sample data; is raw sample data,
Figure 416419DEST_PATH_IMAGE010
be the numerical value after the conversion;
The data that said prediction module transmits data preprocessing module are as input, and good BP neural network is predicted to utilize training in advance, obtains the following network parameter of cognition network constantly; Said BP neural network obtains according to following method training:
Step 1, the protocol stack parameter and the performance parameter of gathering cognition network, and it is carried out normalization handle, training sample obtained;
Step 2, three layers of BP neural network of structure, wherein the input layer number is N-1, output layer node number is 1; N is the protocol stack parameter of the cognition network gathered and the species number of performance parameter; The number of hidden nodes is confirmed according to following method: the average with error sum of squares is the accuracy evaluation index, through loop test, chooses the number of hidden nodes that makes that the error sum of squares average is minimum;
Step 3, initial parameter are selected;
Step 4, utilize the training sample that obtains in the step 1 three layers of constructed BP neural network to be trained the BP neural network after obtaining training.
Compare prior art, the present invention has following beneficial effect:
(l) network protocol stack parameter and performance parameter are associated, when network condition changes, the reason that can very fast finding network is changed, thus important basis is provided for dealing with problems.
(2) the BP neural network structure is simple, adjustable parameter is many, training algorithm is many, can be handling good, the present invention is the basis with the self-learning property of BP neural network, through selecting suitable network parameter, can reach very high precision of prediction.
Description of drawings
Fig. 1 is the structural representation that the present invention is based on the cognition network performance prediction device of BP neural network;
Fig. 2 is the algorithm flow chart of BP neural metwork training among the present invention.
Embodiment
Below in conjunction with accompanying drawing technical scheme of the present invention is elaborated:
Cognition network performance prediction device based on the BP neural network of the present invention, its structure is as shown in Figure 1, comprises by signal flow to the information sensing module, data preprocessing module, the prediction module that connect successively.The course of work below in conjunction with this device is elaborated to Forecasting Methodology of the present invention.
The information sensing module is used for gathering in real time the network parameter of cognition network.The network parameter of being gathered among the present invention comprises protocol stack parameter and network performance parameter.Wherein performance parameter comprises time delay, shake, handling capacity, packet loss, number of dropped packets; The protocol stack parameter comprises average queue length, end-to-end round-trip delay and tcp window size.Above-mentioned network parameter can obtain through 100/1000M FE capture card, atm link capture card and SDH link capture card.
For avoiding the excessive network paralysis that causes of raw data, handle raw data.The present invention carries out normalization through pre-processing module to data and handles, and through the normalization formula all data conversions is arrived in the scope of (0,1).The normalization formula is as follows:
Figure 176565DEST_PATH_IMAGE002
Wherein,
Figure 879817DEST_PATH_IMAGE004
and is respectively maximal value and minimum value in the sample data;
Figure 622962DEST_PATH_IMAGE008
is raw sample data,
Figure 554009DEST_PATH_IMAGE010
be the numerical value after the conversion.So not only avoid the input data to fall into the zone of saturation, also kept original characteristic of data.
The data that prediction module transmits data preprocessing module are as input, and good BP neural network is predicted to utilize training in advance, obtains the following network parameter of cognition network constantly; Said BP neural network obtains according to following method training:
Step 1, the protocol stack parameter and the performance parameter of gathering cognition network, and it is carried out normalization handle, training sample obtained.
Step 2, structure BP neural network.
The latent number of plies of the design packet includes network of network topology structure, input and output node layer number.This method adopts three layers of BP neural network, and the promptly latent number of plies is 1.The node number of input layer and output layer is by the parameter determining of gathering.If the parameter kind of gathering is N, then the input layer number is N-1, and output layer node number is 1.
Need confirm the number of hidden nodes then.
Select the reference formula of best the number of hidden nodes
Figure 2011104521254100002DEST_PATH_IMAGE012
as follows:
Figure 2011104521254100002DEST_PATH_IMAGE014
Wherein, m is the output neuron number, and n is the input block number, and a is the constant between [1,10].
The present invention with the average
Figure 2011104521254100002DEST_PATH_IMAGE016
of error sum of squares as the accuracy estimating index.The definition of error sum of squares average is following:
Definition: the expectation derived value of supposing the system is
Figure 2011104521254100002DEST_PATH_IMAGE018
; Corresponding network real output value is ; Reasoning error e is , and then the formula of the average of error sum of squares
Figure 798084DEST_PATH_IMAGE016
is:
Figure DEST_PATH_IMAGE024
Wherein,
Figure DEST_PATH_IMAGE026
is desired output, and
Figure DEST_PATH_IMAGE028
is the actual output of network.
At first calculate the scope of best the number of hidden nodes through the hidden node formula; Be the accuracy evaluation index with the error sum of squares average then; In the scope of having calculated, carry out loop test, choose suitable the number of hidden nodes and make the error sum of squares average minimum.
Step 3, initial parameter are selected.
Initial parameter comprises learning rate, initial weight, cycle index.In order to keep network stabilization and rate of convergence, the learning rate value is between (0.01,0.8).Initial weight is chosen and is obeyed equally distributed random number between (1,1), to improve the training speed of network.Cycle index is confirmed through the loop test in the step 2.
Step 4, utilize the training sample that obtains in the step 1 three layers of constructed BP neural network to be trained the BP neural network after obtaining training.
The present invention adopts the trainlm function that the BP neural network that makes up is trained; The learning algorithm of this function is the Levenberg-Marquadt back propagation algorithm; This algorithm is a kind of based on the theoretical training algorithm of numerical optimization, and learning rate is adaptive, therefore; The speed of convergence of function trainlm is very fast, and the training error of network is also less.
Exist corresponding funtcional relationship between the various network parameters.The present invention adopts the BP neural network to describe the funtcional relationship of this implicit expression between protocol stack parameter and the network performance parameter; Under the situation that network structure has been confirmed; Through seeking the weights of an appropriate, make network approach the funtcional relationship between protocol stack parameter and the performance parameter as much as possible.Training sample is offered the BP neural network, and neuronic activation value is propagated to output layer through each middle layer from input layer, obtains the input response of network at each neuron of output layer.Then,, oppositely get back to input layer, respectively connect weights thereby successively revise through each middle layer from output layer according to the direction that reduces error between target output and the actual output.Concrete training step is following:
Step 401, will pass through pretreated sample and be input to the BP neural network;
Step 402, carry out the parameter initialization setting, comprise maximum frequency of training Max_epoch, error precision
Figure DEST_PATH_IMAGE030
(being also referred to as training objective), the number of hidden nodes, initial weight, threshold value, initial learn speed;
Step 403, calculate the input and output value of each layer, calculate the error of each layer;
Step 404, according to the weights and the threshold value of adaptive learning rate algorithms correction neural network;
Step 405, if error precision greater than
Figure 161195DEST_PATH_IMAGE030
, then return step 404; If error precision less than or frequency of training during greater than its maximum frequency of training Max_epoch, stops training.
The BP neural network that trains has been described the funtcional relationship of implicit expression between protocol stack parameter and the network performance parameter.With network protocol stack parameter and the network performance parameter fan-in network gathered, utilize the stable network structure (comprising training parameter), connection weights and the threshold values that have obtained to the test sample book analysis, dope following network performance parameter constantly.Thereby can initiatively make a policy according to the network condition that is doped, satisfy user's demand with maximization.

Claims (7)

1. based on the cognition network performance prediction method of BP neural network, the network parameter current according to cognition network predicted the cognition network performance parameter in the moment in future; It is characterized in that this method as input, utilizes the BP neural network to carry out the cognition network performance prediction with the protocol stack parameter of cognition network and performance parameter; Said BP neural network obtains according to following method training:
Step 1, the protocol stack parameter and the performance parameter of gathering cognition network, and it is carried out normalization handle, training sample obtained;
Step 2, three layers of BP neural network of structure, wherein the input layer number is N-1, output layer node number is 1; N is the protocol stack parameter of the cognition network gathered and the species number of performance parameter; The number of hidden nodes is confirmed according to following method: the average with error sum of squares is the accuracy evaluation index, through loop test, chooses the number of hidden nodes that makes that the error sum of squares average is minimum;
Step 3, initial parameter are selected;
Step 4, utilize the training sample that obtains in the step 1 three layers of constructed BP neural network to be trained the BP neural network after obtaining training.
2. according to claim 1 based on the cognition network performance prediction method of BP neural network, it is characterized in that said protocol stack parameter comprises average queue length, end-to-end round-trip delay and tcp window size; Said performance parameter comprises time delay, shake, handling capacity, packet loss, number of dropped packets.
3. according to claim 1 based on the cognition network performance prediction method of BP neural network, it is characterized in that said normalization processing is meant and all sample datas is transformed in the scope of (0,1) according to following formula:
Figure 2011104521254100001DEST_PATH_IMAGE002
Wherein,
Figure 2011104521254100001DEST_PATH_IMAGE004
and
Figure 2011104521254100001DEST_PATH_IMAGE006
is respectively maximal value and minimum value in the raw sample data;
Figure 2011104521254100001DEST_PATH_IMAGE008
is raw sample data,
Figure 2011104521254100001DEST_PATH_IMAGE010
be the numerical value after the conversion.
4. according to claim 1 based on the cognition network performance prediction method of BP neural network; It is characterized in that said initial parameter comprises learning rate, initial weight, cycle index, wherein the learning rate span is (0.01; 0.8); Initial weight is chosen equally distributed random number between (1,1), and cycle index is confirmed by the loop test in the step 2.
5. according to claim 1 based on the cognition network performance prediction method of BP neural network, it is characterized in that said step 4 specifically comprises:
Step 401, will pass through pretreated sample and be input to the BP neural network;
Step 402, carry out the parameter initialization setting, comprise maximum frequency of training, training objective, the number of hidden nodes, initial weight, threshold value, initial learn speed;
Step 403, calculate the input and output value of each layer, calculate the error of each layer;
Step 404, according to the weights and the threshold value of adaptive learning rate algorithms correction neural network;
Step 405, as reaching training objective or frequency of training greater than preset maximum frequency of training, then stop training; As not, then return step 404.
6. according to claim 1 based on the cognition network performance prediction method of BP neural network, it is characterized in that, the BP neural network is trained adopt the Levenberg-Marquadt back propagation algorithm.
7. based on the cognition network performance prediction device of BP neural network, it is characterized in that this device comprises by signal flow to the information sensing module, data preprocessing module, the prediction module that connect successively;
Said information sensing module is used for gathering in real time the protocol stack parameter and the performance parameter of cognition network, and with data transmission to data preprocessing module, said protocol stack parameter comprises average queue length, end-to-end round-trip delay and tcp window size; Said performance parameter comprises time delay, shake, handling capacity, packet loss, number of dropped packets;
The data that said data preprocessing module is carried out after normalization is handled and will be handled the data that receive are sent to prediction module; Said normalization processing is meant and all sample datas is transformed in the scope of (0,1) according to following formula:
Figure 715453DEST_PATH_IMAGE002
Wherein,
Figure 894761DEST_PATH_IMAGE004
and
Figure 107568DEST_PATH_IMAGE006
is respectively maximal value and minimum value in the raw sample data;
Figure 190187DEST_PATH_IMAGE008
is raw sample data,
Figure 463036DEST_PATH_IMAGE010
be the numerical value after the conversion;
The data that said prediction module transmits data preprocessing module are as input, and good BP neural network is predicted to utilize training in advance, obtains the following network parameter of cognition network constantly; Said BP neural network obtains according to following method training:
Step 1, the protocol stack parameter and the performance parameter of gathering cognition network, and it is carried out normalization handle, training sample obtained;
Step 2, three layers of BP neural network of structure, wherein the input layer number is N-1, output layer node number is 1; N is the protocol stack parameter of the cognition network gathered and the species number of performance parameter; The number of hidden nodes is confirmed according to following method: the average with error sum of squares is the accuracy evaluation index, through loop test, chooses the number of hidden nodes that makes that the error sum of squares average is minimum;
Step 3, initial parameter are selected;
Step 4, utilize the training sample that obtains in the step 1 three layers of constructed BP neural network to be trained the BP neural network after obtaining training.
CN2011104521254A 2011-12-30 2011-12-30 Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network Pending CN102592171A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2011104521254A CN102592171A (en) 2011-12-30 2011-12-30 Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2011104521254A CN102592171A (en) 2011-12-30 2011-12-30 Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network

Publications (1)

Publication Number Publication Date
CN102592171A true CN102592171A (en) 2012-07-18

Family

ID=46480780

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2011104521254A Pending CN102592171A (en) 2011-12-30 2011-12-30 Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network

Country Status (1)

Country Link
CN (1) CN102592171A (en)

Cited By (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103561094A (en) * 2013-11-04 2014-02-05 成都数之联科技有限公司 Intelligent monitoring method for sleep condition of infant
CN104931538A (en) * 2015-06-10 2015-09-23 中冶南方工程技术有限公司 Learning type hydrochloric acid concentration and Fe ion concentration on-line detection system and method
CN105612492A (en) * 2013-10-02 2016-05-25 高通股份有限公司 Method and apparatus for producing programmable probability distribution function of pseudo-random numbers
CN105989407A (en) * 2015-02-12 2016-10-05 中国人民解放军信息工程大学 Neural network based short wave median field intensity prediction system, method and device
CN106204246A (en) * 2016-08-18 2016-12-07 易联众信息技术股份有限公司 A kind of BP neutral net credit estimation method based on PCA
CN106447103A (en) * 2016-09-26 2017-02-22 河海大学 Deep learning based QoS prediction method of Web service
CN107528649A (en) * 2017-08-30 2017-12-29 华南师范大学 A kind of Forecasting Methodology of the cognitive radio networks frequency spectrum based on Bagging
CN107852771A (en) * 2015-07-28 2018-03-27 华为技术有限公司 Network anomaly detection based on adaptive-filtering
WO2018053835A1 (en) * 2016-09-26 2018-03-29 Intel Corporation Method and apparatus for reducing parameter density of deep neural network (dnn)
CN108055149A (en) * 2017-12-08 2018-05-18 国网辽宁省电力有限公司本溪供电公司 End-to-end Traffic Anomaly feature extracting method in a kind of Time and Frequency Synchronization application
CN108564000A (en) * 2018-03-20 2018-09-21 中北大学 A kind of fragmentation signal automatic-identifying method based on neural network
WO2018176768A1 (en) * 2017-03-27 2018-10-04 烽火通信科技股份有限公司 Network architecture of humanoid network and implementation method
TWI646808B (en) * 2016-01-29 2019-01-01 中華電信股份有限公司 Request traffic prediction method
CN109684182A (en) * 2018-12-06 2019-04-26 中国联合网络通信集团有限公司 A kind of method and apparatus that prediction network congestion influences network quality
CN110378052A (en) * 2019-07-25 2019-10-25 北京航空航天大学 It is looked to the future the equipment method for predicting residual useful life of operating condition based on Recognition with Recurrent Neural Network
CN110580385A (en) * 2019-08-21 2019-12-17 南京博阳科技有限公司 data processing method, device and equipment for underground pipeline and computer storage medium
CN111163530A (en) * 2019-11-18 2020-05-15 浙江万胜智能科技股份有限公司 Wireless local area network performance enhancing method based on neural network algorithm
CN111245678A (en) * 2020-01-09 2020-06-05 同方电子科技有限公司 Adaptive time domain telescopic window short wave radio station microphone PTT signal jitter elimination algorithm
CN111861115A (en) * 2020-06-15 2020-10-30 复旦大学 Neural experiment test calculation method for consumer cognition of internet financial products
CN111884870A (en) * 2020-06-28 2020-11-03 苏州浪潮智能科技有限公司 Method, device, terminal and storage medium for predicting network sub-health state
CN112149912A (en) * 2020-09-30 2020-12-29 中煤能源研究院有限责任公司 Coal mine working face mine pressure data prediction method based on reverse propagation neural network
CN113141318A (en) * 2021-03-29 2021-07-20 重庆邮电大学 Improved TFRC congestion control system and method based on video traffic prediction model
CN113348691A (en) * 2018-11-28 2021-09-03 诺基亚通信公司 Method and apparatus for failure prediction in network management

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101834753A (en) * 2010-03-30 2010-09-15 华中科技大学 Network flow prediction method based on covariant orthogonality and neural network
CN102025620A (en) * 2010-12-07 2011-04-20 南京邮电大学 Cognitive network QoS (quality of service) guarantee method on basis of service differentiation
CN102056183A (en) * 2010-12-10 2011-05-11 北京交通大学 Network flow prediction method and device based on cognitive network
CN102130732A (en) * 2011-04-01 2011-07-20 北京邮电大学 Cooperative spectrum detection method for cognitive radio based on neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101834753A (en) * 2010-03-30 2010-09-15 华中科技大学 Network flow prediction method based on covariant orthogonality and neural network
CN102025620A (en) * 2010-12-07 2011-04-20 南京邮电大学 Cognitive network QoS (quality of service) guarantee method on basis of service differentiation
CN102056183A (en) * 2010-12-10 2011-05-11 北京交通大学 Network flow prediction method and device based on cognitive network
CN102130732A (en) * 2011-04-01 2011-07-20 北京邮电大学 Cooperative spectrum detection method for cognitive radio based on neural network

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
《桂林电子工业学院学报》 20000331 舒朝海等 基于速率的ATM流量反馈控制-一种基于神经网络预测的方法 71-76 1-7 第20卷, 第1期 *
乔爱萍等: "网络行为测量研究", 《计算机应用研究》 *
李丹丹等: "认知网络中基于蚁群算法的网络流量预测模型", 《电子学报》 *
柳小桐: "BP神经网络输入层数据归一化研究", 《机械工程与自动化》 *
舒朝海等: "基于速率的ATM流量反馈控制—一种基于神经网络预测的方法", 《桂林电子工业学院学报》 *

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105612492B (en) * 2013-10-02 2018-08-03 高通股份有限公司 Method, apparatus, equipment and the medium of spike are reduced in Artificial Neural System
CN105612492A (en) * 2013-10-02 2016-05-25 高通股份有限公司 Method and apparatus for producing programmable probability distribution function of pseudo-random numbers
CN103561094A (en) * 2013-11-04 2014-02-05 成都数之联科技有限公司 Intelligent monitoring method for sleep condition of infant
CN105989407A (en) * 2015-02-12 2016-10-05 中国人民解放军信息工程大学 Neural network based short wave median field intensity prediction system, method and device
CN104931538A (en) * 2015-06-10 2015-09-23 中冶南方工程技术有限公司 Learning type hydrochloric acid concentration and Fe ion concentration on-line detection system and method
CN107852771A (en) * 2015-07-28 2018-03-27 华为技术有限公司 Network anomaly detection based on adaptive-filtering
CN107852771B (en) * 2015-07-28 2020-06-02 华为技术有限公司 Network anomaly detection based on adaptive filtering
TWI646808B (en) * 2016-01-29 2019-01-01 中華電信股份有限公司 Request traffic prediction method
CN106204246A (en) * 2016-08-18 2016-12-07 易联众信息技术股份有限公司 A kind of BP neutral net credit estimation method based on PCA
WO2018053835A1 (en) * 2016-09-26 2018-03-29 Intel Corporation Method and apparatus for reducing parameter density of deep neural network (dnn)
CN106447103A (en) * 2016-09-26 2017-02-22 河海大学 Deep learning based QoS prediction method of Web service
WO2018176768A1 (en) * 2017-03-27 2018-10-04 烽火通信科技股份有限公司 Network architecture of humanoid network and implementation method
CN107528649A (en) * 2017-08-30 2017-12-29 华南师范大学 A kind of Forecasting Methodology of the cognitive radio networks frequency spectrum based on Bagging
CN108055149A (en) * 2017-12-08 2018-05-18 国网辽宁省电力有限公司本溪供电公司 End-to-end Traffic Anomaly feature extracting method in a kind of Time and Frequency Synchronization application
CN108564000A (en) * 2018-03-20 2018-09-21 中北大学 A kind of fragmentation signal automatic-identifying method based on neural network
CN108564000B (en) * 2018-03-20 2021-09-28 中北大学 Fragment signal automatic identification method based on neural network
CN113348691A (en) * 2018-11-28 2021-09-03 诺基亚通信公司 Method and apparatus for failure prediction in network management
CN109684182A (en) * 2018-12-06 2019-04-26 中国联合网络通信集团有限公司 A kind of method and apparatus that prediction network congestion influences network quality
CN109684182B (en) * 2018-12-06 2022-07-22 中国联合网络通信集团有限公司 Method and device for predicting influence of network congestion on network quality
CN110378052A (en) * 2019-07-25 2019-10-25 北京航空航天大学 It is looked to the future the equipment method for predicting residual useful life of operating condition based on Recognition with Recurrent Neural Network
CN110580385A (en) * 2019-08-21 2019-12-17 南京博阳科技有限公司 data processing method, device and equipment for underground pipeline and computer storage medium
CN111163530A (en) * 2019-11-18 2020-05-15 浙江万胜智能科技股份有限公司 Wireless local area network performance enhancing method based on neural network algorithm
CN111245678A (en) * 2020-01-09 2020-06-05 同方电子科技有限公司 Adaptive time domain telescopic window short wave radio station microphone PTT signal jitter elimination algorithm
CN111245678B (en) * 2020-01-09 2022-08-16 同方电子科技有限公司 Adaptive time domain telescopic window short-wave radio station microphone PTT signal jitter elimination algorithm
CN111861115A (en) * 2020-06-15 2020-10-30 复旦大学 Neural experiment test calculation method for consumer cognition of internet financial products
CN111861115B (en) * 2020-06-15 2022-03-18 复旦大学 Neural experiment test calculation method for consumer cognition of internet financial products
CN111884870A (en) * 2020-06-28 2020-11-03 苏州浪潮智能科技有限公司 Method, device, terminal and storage medium for predicting network sub-health state
CN111884870B (en) * 2020-06-28 2021-10-22 苏州浪潮智能科技有限公司 Method, device, terminal and storage medium for predicting network sub-health state
CN112149912A (en) * 2020-09-30 2020-12-29 中煤能源研究院有限责任公司 Coal mine working face mine pressure data prediction method based on reverse propagation neural network
CN113141318A (en) * 2021-03-29 2021-07-20 重庆邮电大学 Improved TFRC congestion control system and method based on video traffic prediction model

Similar Documents

Publication Publication Date Title
CN102592171A (en) Method and device for predicting cognitive network performance based on BP (Back Propagation) neural network
Qu et al. Cross-layer congestion control of wireless sensor networks based on fuzzy sliding mode control
Jiang et al. When machine learning meets congestion control: A survey and comparison
CN111526096B (en) Intelligent identification network state prediction and congestion control system
CN107634911B (en) Adaptive congestion control method based on deep learning in information center network
CN113966596B (en) Method and apparatus for data traffic routing
Gomez et al. Intelligent active queue management using explicit congestion notification
Patel et al. A new modified dropping function for congested AQM networks
CN113727306B (en) Decoupling C-V2X network slicing method based on deep reinforcement learning
Xu et al. Resource allocation algorithm based on hybrid particle swarm optimization for multiuser cognitive OFDM network
Su et al. QRED: A Q-learning-based active queue management scheme
Quan et al. Cybertwin-driven DRL-based adaptive transmission scheduling for software defined vehicular networks
CN107070802A (en) Wireless sensor network Research of Congestion Control Techniques based on PID controller
CN105704055A (en) Network congestion preventing method based on natural heuristic algorithm
Huang Quality of service optimization in wireless transmission of industrial Internet of Things for intelligent manufacturing
Falahatraftar et al. An intelligent congestion avoidance mechanism based on generalized regression neural network for heterogeneous vehicular networks
Liu et al. Deep reinforcement learning based adaptive transmission control in vehicular networks
Mehta Genetic algorithm based bi-objective optimization of sigmoidal utility and throughput in ad-hoc wireless networks
CN115037672B (en) Multipath congestion control method and device
Meng et al. Intelligent routing orchestration for ultra-low latency transport networks
CN105425590A (en) PID active queue management method based on improved immune algorithm
Amar et al. Online learning for shortest path and backpressure routing in wireless networks
Doibale et al. Wireless sensor networks congestion and role of artificial intelligence
Bhaskar et al. Deep Neural Network Algorithm to Improve Link Reliability in Wireless Sensor Networks
Prakash et al. A rapidly-exploring random tree-based intelligent congestion control through an alternate routing for WSNs

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20120718