CN107124761A - Merge PSO and SS ELM Wireless Location in Cellular Network method - Google Patents
Merge PSO and SS ELM Wireless Location in Cellular Network method Download PDFInfo
- Publication number
- CN107124761A CN107124761A CN201710140253.2A CN201710140253A CN107124761A CN 107124761 A CN107124761 A CN 107124761A CN 201710140253 A CN201710140253 A CN 201710140253A CN 107124761 A CN107124761 A CN 107124761A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- elm
- mtd
- pso
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
- H04W64/006—Locating users or terminals or network equipment for network management purposes, e.g. mobility management with additional information processing, e.g. for direction or speed determination
-
- 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
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The invention discloses fusion PSO and SS ELM Wireless Location in Cellular Network method.The realization of the present invention is including the use of having label training data and SS ELM output layer weighting parameter β is trained without label training data, PSO is introduced during being trained to SS ELM Automatic Optimal is carried out to SS ELM hyper parameter, PSO fitness value calculation function, which has used label training data during training and screening is optimized to SS ELM without label training data, obtains optimal SS ELM parameters as regression model for tuning on-line service.The implementation defiber and online two parts of the present invention is implemented.The present invention reduces the cellular network location based on RSS finger print datas to there is a dependence of the RSS finger print datas of label, reduces the cost of artificial gathered data, and reduces the workload of the manual adjustment parameter during Algorithm for Training.
Description
Technical field
The invention belongs to a kind of cellular network wireless positioning method in pattern-recognition and computational intelligence field, and in particular to
Particle group optimizing (Particle Swarm Optimization, PSO) and semi-supervised extreme learning machine (Semi-Supervised
Extreme Learning Machine,SS-ELM)。
Background technology
Highly developed beehive network system and cellular network signals covering the whole world cause beehive network system to be the most
Widely used GSM, and the popularization of smart mobile phone cause the location technology based on cellular network system to become one kind
Important outdoor positioning technology.Especially in global position system, in the case of such as global positioning system (GPS) is disabled, intelligence
Mobile phone can only carry out outdoor positioning by beehive network system.At the same time with the development of technology of Internet of things, honeycomb is accessed
The smart machine of network will be more and more, obtain the positional information of smart machine and will turn into the prerequisite bar of various application scenarios
Part.
With TOA, TDOA, the traditional position location techniques based on geometric distance such as AOA are compared, received signal strength (RSS) and machine
Device learning algorithm is more suitable for positioning mobile device in the environment of radio signal non-line-of-sight propagation (NLOS).But
Due to because there is the RSS finger print datas required for the machine learning algorithm training of supervision, i.e., with positional information and reception signal
The collection of " the having label training data " of intensity (RSS) is, it is necessary to by moveable signal collecting device in the target area of positioning
Received signal strength and corresponding positional information (two-dimentional longitude and latitude or one-dimensional band of position sign) are obtained in domain, especially
It is to be collected into enough RSS finger print datas in wide outdoor environment then to need more time and human cost.Conversely obtain
Take the cost paid without label received signal strength (RSS) data without positional information many less, we can be from
MR (Measurement Report) extracting data that mobile phone in destination service area is uploaded goes out a large amount of received signal strengths
(RSS) data, but these data do not include positional information, and we term it " no label training data ".In area of pattern recognition
Semi-supervised learning, no label data is the supplement for having label training data, in the case where there is label training data less without mark
Label data can improve the accuracy rate of algorithm prediction and identification.
Semi-supervised extreme learning machine (Semi-supervised extreme learning machine) hereinafter referred to as SS-
ELM is a kind of neutral net of single hidden layer, with the fast generalization ability of training speed it is good the characteristics of, while the algorithm can be combined
There is label and be trained without label training data.But SS-ELM pairsNorm and the constraint of manifold canonical are very sensitive, in reality
Not having systematic theoretical direction in the application on border, we remove the hyper parameter of optimization constraint, and SS-ELM training process is only right in itself
Output layer weights β is trained optimization, and also the theory in the absence of system is instructed SS-ELM hyperparameter optimizations, so SS-
The optimization of ELM hyper parameters often according to specific business scenario can only carry out experiment repeatedly by experienced staff.
The content of the invention
Embodiment of the present invention is divided into two big steps:Purpose is to solve the cellular network ring based on RSS finger print datas
Positioning under border needs collection is more to have label training data, using SS-ELM in the case where realizing equal positioning precision, SS-ELM
What is needed has label training data less, so as to reduce the cost for artificially collecting label training data.The present invention is simultaneously
SS-ELM and PSO combination, carries out Automatic Optimal to SS-ELM hyper parameter using PSO, reduces in algorithm parameter tuning
Manual intervention, improves the efficiency of production application.The present invention is achieved through the following technical solutions.
PSO and SS-ELM Wireless Location in Cellular Network method is merged, it is including the use of having label training data and without label
Training data is trained to SS-ELM output layer weighting parameter β, using PSO to SS-ELM in SS-ELM training process
Hyper parameter carry out Automatic Optimal, the calculating of PSO fitness value calculation function covered label training data and without label
Training data, the optimal SS-ELM parameters obtained after PSO optimal screenings are used for tuning on-line service as regression model.
Further, using equipment received signal strength (RSS) as input, SS-ELM is trained and optimization is obtained
SS-ELM parameters provide the tuning on-line service under cellular network environment as regression model.
Further, label training number has been used simultaneously in the training process to SS-ELM output layer weighting parameters β
According to without label training data.
Further, the calculating of the fitness value calculation function of PSO algorithms has covered label training and without label training
Data, particle fitness calculates being specifically defined as function:
Wherein i, j, k, N represent the numbering of particle, have label training data example number, the dimension of label vector to compile respectively
Number, the total dimension of label vector, l is that training data is concentrated and has the number of label training data, and u is that training data is concentrated without label instruction
Practice the number of data,Represent SS-ELM on training dataset to the Tag Estimation value for thering is label training data to be predicted, Y
The label actual value of label training data is indicated,Produced by being given a forecast for SS-ELM in each repetitive exercise on training set
Extremum ratio.
Further, key parameters of the PSO to SS-ELM is used during SS-ELM is trained and is optimized:Norm surpasses
Parameter cβ, manifold canonical hyper parameter λ is adjusted and optimizes;PSO is to SS-ELM hyper parameters cβ, λ optimization calculation it is specific such as
Under:PSO is randomly generated particle in defined search space, and each particle is moved to global search most with certain speed
The position of excellent solution;Each particle can be optimal with itself according to the momentum of oneself in the iteration each time of particle cluster algorithm
Position PbWith global optimal PgThe influence factor of position adjusts the speed of oneself, while calculating oneself in current iteration
Position;The dimension of the search space of population is 2, and total population number is n, some particle ithThe position of each iteration can be with table
It is shown as vectorial Xi=(xi1,xi2),xi1∈[-10,0],xi2∈[-10,0];The relation of particle position and hyper parameter is expressed as Personal best particle of the particle since search so far is expressed as Pib=(pi1,pi2), particle
Movement velocity be expressed as vectorial Vi=(vi1,vi2), global optimum position Pg=(pg1,pg2), in iteration particle meeting each time
The current speed of renewal and position, and Proper treatment f (X are utilized according to particle current locationi) particle fitness is calculated, update individual
Body optimal location and global optimum position;Renewal of each variables of PSO in iteration can be represented with following formula:
vid(t+1)=vid(t)+c1*rand()*[pid(t)-xid(t)]+c2*and()*[pgd(t)-Xid(t)] (4)
xid(t+1)=xid(t)+vid(t+1)1≤i≤n,1≤d≤2 (5)
Wherein positive number c1,c2It is acceleration factor, rand () is the random number between 0 and 1;vmaxAnd vminIt is particle respectively
The upper bound of speed and lower bound, t are the algebraically of algorithm iteration;When the particle fitness corresponding to global optimum position in iterative process
Stop iterative calculation when convergence no longer changes with iteration, pass through global optimum's particle position and the corresponding relation of hyper parameter
Optimal hyper parameter can be obtained.
Compared with prior art, the beneficial effects of the present invention are:
1) compared in the case where training data is less and use supervised learning method, use semi-supervised learning algorithm SS-ELM
More preferable positioning precision can be obtained.
2) selections of the SS-ELM to hyper parameter is more sensitive, generally requires skilled engineer according to different business
Scene carries out experiment regulation parameter repeatedly, and Automatic Optimal and regulation are carried out to SS-ELM hyper parameter present invention uses PSO,
Reduce manual intervention.
3) from traditional PSO only be used only label training data it is different the present invention PSO be based on simultaneously be based on have label instruction
Practice data and be iterated training without label training data, and employ formula (1) as fitness value calculation function so that passing through
The more healthy and strong stabilizations of SS-ELM that PSO optimized are crossed, the positioning to user is more accurate.
Brief description of the drawings
Fig. 1 for the inventive method specific implementation process in off-line training part flow chart.
Fig. 2 for the inventive method specific implementation process in on-line prediction part flow chart.
Fig. 3 is SS-ELM neutral nets topological diagram used in the present invention.
Fig. 4 is variation tendency of the PSO fitness values in repetitive exercise in the specific embodiment of the invention.
Embodiment
Specific implementation below in conjunction with accompanying drawing and example to the present invention is described further, but the implementation and protection of the present invention
Not limited to this.The present invention is divided into two big steps by following embodiment:
Step 1: having label training data and without label training data by merging PSO to SS-ELM algorithms using part
Model carry out off-line training, it is different from traditional artificial parameter distribution method, merge PSO and SS-ELM purpose be to search using PSO
Rope and the optimal super ginseng in different business scene SS-ELM is filtered out, draw optimal model.
Step 2: carrying out tuning on-line to the equipment of user using optimal SS-ELM models.
The crucial computational methods of SS-ELM and PSO training are described below in above-mentioned steps one:
Assuming that giving RSS finger print data the sample sets { (r for having l tape labelk,yk) | k=1,2,3 ..., l }lWith u
Not the RSS data sample set of tape label, { rk| k=1,2,3 ..., u }uWherein r components are received signal strength (RSS), y points
Measure as positional information, choose sigmoid functions as SS-ELM hidden layer activation primitive, then SS-ELM training needs to meet
Following object function:
S.t.F=H β
Wherein
A is the weight that input layer connects hidden layer, and b is the biasing of hidden layer, and β is connection hidden layer and output in SS-ELM
The weight of layer, cβCorresponded to respectively with λThe hyper parameter of norm and manifold regular terms.CeIt is (l+u) × (l+u) weight square formation,
[Ce]jj=1, j=1,2 ... l, it is to have label training data and without label training data respectively that remaining element, which is equal to zero, l and u,
Number.The mark of Tr () representing matrix, L is Laplacian Matrix, and D is diagonal matrix, wherein.wijIt is xiAnd xjBetween category
Property similarity:
L=D-W (12)
Because extreme learning machine only needs to train and corrects the weight beta between output layer and hidden layer, so to SS-ELM
Training be equal to the weight beta that solution meets formula (7).So carrying out derivation to formula (7), its derivative is made to obtain following for 0
Formula:
If the line number of matrix H is more than columnsThen β approximate solution is in formula (13):
If H columns is more than line number
β*=HT(cβI(l+u)×(1+u)+CeHHT+λLHHT)-1CeT (15)
Wherein I isTie up unit matrix.
PSO is randomly generated particle in defined search space, and each particle is moved to the overall situation with certain speed
The position of the optimal solution of search.Each particle can be optimal with itself according to the momentum of oneself in PSO iteration each time
Position (Pb) and the optimal (P of the overall situationg) influence factor of position adjusts the speed of oneself, while calculating oneself in current iteration
In position.Assuming that the dimension that our search space is is D, total population number is n, some particle ithThe position of each iteration
Vectorial X can be expressed as by puttingi=(xi1,xi2,…,xiD), personal best particle of the particle since search so far is expressed as Pib
=(pi1,pi2,…,piD), the movement velocity of particle is expressed as vectorial Vi=(vi1,vi2,…,viD), global optimum position Pg=
(pg1,pg2,…,pgD), current speed and position can be updated in iteration particle each time, and utilize according to particle current location
Proper treatment f (Xi) calculate particle Moderate renewal personal best particle and global optimum position.Each variables of PSO are in iteration
Renewal can be represented with following formula:
vid(t+1)=vid(t)+c1*rand()*[pid(t)-xid(t)]+c2*rand()*[pgd(t)-Xid(t)] (18)
xid(t+1)=xid(t)+vid(t+1)1≤i≤n,1≤d≤D(19)
Wherein positive number c1,c2It is acceleration factor, rand () is the random number between 0 and 1;vmaxAnd vminIt is particle respectively
The upper bound of speed and lower bound, t are the algebraically of algorithm iteration.
Embodiment of the present invention is described further by taking two-dimensional space as an example below.
Step 1:The a collection of quantity of random initializtion is n particles, the initial position X of each particlei=(xi1,xi2), wherein xi1
∈[-10,0]、xi2∈ [- 10,0], i represents the numbering of particle.Accordingly a SS-ELM example is built for each particle:Wherein, ai、bi、βi、cβi、λiInput layer weight, hidden layer biasing, output layer power are represented respectively
Weight,The hyper parameter of norm hyper parameter, manifold regular terms.Of SS-ELM input layers is set according to input parameter dimension
NumberHidden layer neuron number is setTo SS-ELM input layer weight ai, the biasing b of hidden layeriRandom assignment is carried out to take
It is [0,1] to be worth scope, with season SS-ELMIt is bottom that the hyper parameter of norm and manifold regular terms, which is expressed as with 10,
Exponential representation:
Initialize weight square formation CeDiagonal element [Ce]jjThe subscript of=1, j=1,2 ... l, j expression matrix element, l tables
It is shown with the number of label training data, square formation CeRemaining element is initialized as 0.
Step 2:Table 1 is used to position preceding 35 RSS informations for being classified as equipment in the RSS finger print data examples of target area, table 1
Alphabetical r represents that next two columns represent that the two-dimensional position coordinate of corresponding device represents then there is label training number in the example with letter y
{ (r can be expressed as according to collectionk,yk) | k=1,2,3 ..., l }l, wherein l is the number for having label training data.Respectively to preceding 35
The RSS and next two columns location coordinate information of row calculate hiding for SS-ELM after being normalized according to formula (8)~(12)
Layer output H and Laplacian Matrix L.
Table 1.RSS data instances
Step 3:Example calculates the corresponding output layer weight of each SS-ELM examples according to formula (14) or formula (15)
And be predicted by formula below using current iteration training output weight on training dataset
Secondly, statistical forecast collectionMiddle extremum sample proportion
The definition of extremum is:For arbitraryIf be unsatisfactory forThen the sample belongs to pole
End value.Assuming that extremum number of samples is κ, it can obtain
The last fitness value that each particle is calculated according to formula (22)
According to formula (16)~(20) to global optimum position of the position of each particle and speed and individual optimum position etc.
Variable is updated, and recalculates according to the particle after renewal SS-ELM hyper parameter cβiAnd λi。
Step 4:The calculating of repeat step 3 no longer changes such as Fig. 3 until the fitness value convergence of global optimum's particle
It is shown.
Step 5:Choose global optimal particle P in above-mentioned stepsg, and corresponding SS-ELM parameters It is used for tuning on-line as final regression model.
Step 6:Current device received signal strength information is inputted for online Location Request, is used according in step 2
Method for normalizing normalized after received signal strength dataCalculated by equation below
It is rightDo the final position that equipment is obtained after renormalization.
Above-mentioned middle step 1 is off-line training part to step 5, and step 6 is tuning on-line part.Above-mentioned flow is the present invention
Preferably embodiment, but protection scope of the present invention is not limited thereto, any technology people for being familiar with the art
Member the invention discloses technical scope in, the change or replacement that can be readily occurred in should all be covered in protection scope of the present invention
Within.Therefore, protection scope of the present invention should be defined by scope of the claims.
Claims (5)
1. merge PSO and SS-ELM Wireless Location in Cellular Network method, it is characterised in that including the use of have label training data and
SS-ELM output layer weighting parameter β is trained without label training data, PSO pairs is used in SS-ELM training process
SS-ELM hyper parameter carries out Automatic Optimal, the calculating of PSO fitness value calculation function covered label training data and
Without label training data, the optimal SS-ELM parameters obtained after PSO optimal screenings are used for tuning on-line as regression model
Service.
2. fusion PSO and SS-ELM according to claim 1 Wireless Location in Cellular Network method, it is characterised in that:To set
Standby received signal strength is trained to SS-ELM as input and optimization obtains SS-ELM parameters as regression model and provides honeybee
Tuning on-line service under nest network environment.
3. a kind of fusion PSO and SS-ELM according to claim 1 Wireless Location in Cellular Network method, it is characterised in that:
Label training data has been used in the training process to SS-ELM output layer weighting parameters β and without label training number simultaneously
According to.
4. fusion PSO and SS-ELM according to claim 1 Wireless Location in Cellular Network method, it is characterised in that:PSO is calculated
The calculating of the fitness value calculation function of method has covered label training and without label training data, and particle fitness calculates function
Be specifically defined as:
Wherein i, j, k, N represent respectively the numbering of particle, have label training data example number, the dimension of label vector numbering,
The total dimension of label vector, l is that training data concentrates the number for having label training data, and u is that training data is concentrated without label training
The number of data,Represent SS-ELM on training dataset to the Tag Estimation value for thering is label training data to be predicted, Y tables
The label actual value of label training data is shown with,Produced by being given a forecast for SS-ELM in each repetitive exercise on training set
The ratio of extremum.
5. fusion PSO and SS-ELM according to claim 1 Wireless Location in Cellular Network method, it is characterised in that:In SS-
Key parameters of the PSO to SS-ELM is used during ELM training and optimization:Norm hyper parameter cβ, manifold canonical hyper parameter λ enters
Row regulation and optimization;PSO is to SS-ELM hyper parameters cβ, λ optimization calculation it is specific as follows:PSO is in defined search space
Particle is inside randomly generated, each particle is moved to the position of the optimal solution of global search with certain speed;In population
Each particle can be according to the momentum of oneself and itself optimal position P in the iteration each time of algorithmbWith global optimal PgPosition
The influence factor put adjusts the speed of oneself, while calculating oneself position in current iteration;The search space of population
Dimension be 2, total population number be n, some particle ithThe position of each iteration can be expressed as vectorial Xi=(xi1,xi2),
xi1∈[-10,0],xi2∈[-10,0];The relation of particle position and hyper parameter is expressed asThe grain
Personal best particle of the son since search so far is expressed as Pib=(pi1,pi2), the movement velocity of particle is expressed as vectorial Vi=
(vi1,vi2), global optimum position Pg=(pg1,pg2), it can update current speed and position, and root in iteration particle each time
Proper treatment f (X are utilized according to particle current locationi) particle fitness is calculated, update personal best particle and global optimum position;
Renewal of each variables of PSO in iteration can be represented with following formula:
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>i</mi>
<mi>b</mi>
</mrow>
</msub>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo><</mo>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>i</mi>
<mi>b</mi>
</mrow>
</msub>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>e</mi>
<mi>l</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>P</mi>
<mi>g</mi>
</msub>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo><</mo>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>P</mi>
<mi>g</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>P</mi>
<mi>g</mi>
</msub>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>e</mi>
<mi>l</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
vid(t+1)=vid(t)+c1*rand()*[pid(t)-xid(t)]
+c2*rand()
xid(t+1)=xid(t)+vid(t+1)1≤i≤n,1≤d≤2 (5)
<mrow>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>v</mi>
<mrow>
<mi>i</mi>
<mi>d</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mi>v</mi>
<mi>max</mi>
</msub>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<msub>
<mi>v</mi>
<mrow>
<mi>i</mi>
<mi>d</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>></mo>
<msub>
<mi>v</mi>
<mi>max</mi>
</msub>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>v</mi>
<mrow>
<mi>i</mi>
<mi>d</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mi>v</mi>
<mi>min</mi>
</msub>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<msub>
<mi>v</mi>
<mrow>
<mi>i</mi>
<mi>d</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo><</mo>
<msub>
<mi>v</mi>
<mi>min</mi>
</msub>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<mn>1</mn>
<mo>&le;</mo>
<mi>i</mi>
<mo>&le;</mo>
<mi>n</mi>
<mo>,</mo>
<mn>1</mn>
<mo>&le;</mo>
<mi>d</mi>
<mo>&le;</mo>
<mn>2</mn>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein positive number c1,c2It is acceleration factor, rand () is the random number between 0 and 1;vmaxAnd vminIt is particle rapidity respectively
The upper bound and lower bound, t be algorithm iteration algebraically;When the particle fitness corresponding to global optimum position is restrained in iterative process
Stop iterative calculation when no longer being changed with iteration, can be obtained by the corresponding relation of global optimum's particle position and hyper parameter
To optimal hyper parameter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710140253.2A CN107124761B (en) | 2017-03-10 | 2017-03-10 | Cellular network wireless positioning method fusing PSO and SS-ELM |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710140253.2A CN107124761B (en) | 2017-03-10 | 2017-03-10 | Cellular network wireless positioning method fusing PSO and SS-ELM |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107124761A true CN107124761A (en) | 2017-09-01 |
CN107124761B CN107124761B (en) | 2021-02-19 |
Family
ID=59718289
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710140253.2A Active CN107124761B (en) | 2017-03-10 | 2017-03-10 | Cellular network wireless positioning method fusing PSO and SS-ELM |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107124761B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107770719A (en) * | 2017-09-27 | 2018-03-06 | 无锡神探电子科技有限公司 | A kind of object localization method based on drive test data and machine learning |
CN109100682A (en) * | 2018-08-01 | 2018-12-28 | 南京邮电大学 | Localization method based on multicore extreme learning machine and received signal strength |
CN109871850A (en) * | 2019-01-21 | 2019-06-11 | 北京大学 | A kind of classification method of the mobile lidar data based on neural network model |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101820676A (en) * | 2010-04-12 | 2010-09-01 | 中国科学技术大学 | Sensor node positioning method |
CN103002575A (en) * | 2012-10-16 | 2013-03-27 | 桂林电子科技大学 | Method for locating underwater wireless sensor network nodes on the basis of particle swarm optimization |
CN104992184A (en) * | 2015-07-02 | 2015-10-21 | 东南大学 | Multiclass image classification method based on semi-supervised extreme learning machine |
WO2016187746A1 (en) * | 2015-05-22 | 2016-12-01 | 华为技术有限公司 | Method and device for improving positioning performance of artificial neural network |
-
2017
- 2017-03-10 CN CN201710140253.2A patent/CN107124761B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101820676A (en) * | 2010-04-12 | 2010-09-01 | 中国科学技术大学 | Sensor node positioning method |
CN103002575A (en) * | 2012-10-16 | 2013-03-27 | 桂林电子科技大学 | Method for locating underwater wireless sensor network nodes on the basis of particle swarm optimization |
WO2016187746A1 (en) * | 2015-05-22 | 2016-12-01 | 华为技术有限公司 | Method and device for improving positioning performance of artificial neural network |
CN104992184A (en) * | 2015-07-02 | 2015-10-21 | 东南大学 | Multiclass image classification method based on semi-supervised extreme learning machine |
Non-Patent Citations (1)
Title |
---|
王润元: "基于深度图像和彩色图像融合的头姿估计算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107770719A (en) * | 2017-09-27 | 2018-03-06 | 无锡神探电子科技有限公司 | A kind of object localization method based on drive test data and machine learning |
CN109100682A (en) * | 2018-08-01 | 2018-12-28 | 南京邮电大学 | Localization method based on multicore extreme learning machine and received signal strength |
CN109871850A (en) * | 2019-01-21 | 2019-06-11 | 北京大学 | A kind of classification method of the mobile lidar data based on neural network model |
Also Published As
Publication number | Publication date |
---|---|
CN107124761B (en) | 2021-02-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wu et al. | Evolving RBF neural networks for rainfall prediction using hybrid particle swarm optimization and genetic algorithm | |
Ferreira et al. | Multi-step ahead forecasting of daily reference evapotranspiration using deep learning | |
US11874429B2 (en) | High-temperature disaster forecast method based on directed graph neural network | |
Li et al. | Urban traffic flow forecasting using Gauss–SVR with cat mapping, cloud model and PSO hybrid algorithm | |
Li et al. | Evaluation of urban green space landscape planning scheme based on PSO-BP neural network model | |
Jeon et al. | Artificial intelligence for traffic signal control based solely on video images | |
CN108629288A (en) | A kind of gesture identification model training method, gesture identification method and system | |
CN107705556A (en) | A kind of traffic flow forecasting method combined based on SVMs and BP neural network | |
CN114998712B (en) | Image recognition method, storage medium, and electronic device | |
CN110347971A (en) | Particle filter method, device and storage medium based on TSK fuzzy model | |
CN110781595B (en) | Method, device, terminal and medium for predicting energy use efficiency (PUE) | |
CN107124761A (en) | Merge PSO and SS ELM Wireless Location in Cellular Network method | |
CN102646164A (en) | Land use change modeling method and system implemented in combination with spatial filtering | |
CN107945534A (en) | A kind of special bus method for predicting based on GMDH neutral nets | |
Salami et al. | A hybrid short-term load forecasting model developed by factor and feature selection algorithms using improved grasshopper optimization algorithm and principal component analysis | |
CN108984995A (en) | A kind of ecology garden landscape design method of evaluation simulation | |
Lu et al. | Image classification and identification for rice leaf diseases based on improved WOACW_SimpleNet | |
Jiang et al. | Hybrid PSO and GA for neural network evolutionary in monthly rainfall forecasting | |
Akar et al. | Evaluation of tree-based machine learning and deep learning techniques in temperature-based potential evapotranspiration prediction | |
CN117116048A (en) | Knowledge-driven traffic prediction method based on knowledge representation model and graph neural network | |
Alif et al. | Crop prediction based on geographical and climatic data using machine learning and deep learning | |
Hou et al. | An occluded cherry tomato recognition model based on improved YOLOv7 | |
Harders et al. | UAV-based real-time weed detection in horticulture using edge processing | |
Li et al. | Spatial-temporal attention network with multi-similarity loss for fine-grained skeleton-based action recognition | |
Li et al. | Object detection based on knowledge graph network |
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 |