CN112261721B - Combined beam distribution method based on Bayes parameter-adjusting support vector machine - Google Patents
Combined beam distribution method based on Bayes parameter-adjusting support vector machine Download PDFInfo
- Publication number
- CN112261721B CN112261721B CN202011115056.3A CN202011115056A CN112261721B CN 112261721 B CN112261721 B CN 112261721B CN 202011115056 A CN202011115056 A CN 202011115056A CN 112261721 B CN112261721 B CN 112261721B
- Authority
- CN
- China
- Prior art keywords
- scheme
- distribution
- support vector
- vector machine
- bayesian
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 30
- 238000005457 optimization Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000002790 cross-validation Methods 0.000 claims abstract description 7
- 230000001413 cellular effect Effects 0.000 claims abstract description 5
- 238000011156 evaluation Methods 0.000 claims abstract description 5
- 238000010295 mobile communication Methods 0.000 claims abstract description 5
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 4
- 230000008602 contraction Effects 0.000 claims 1
- 230000009977 dual effect Effects 0.000 claims 1
- 238000010801 machine learning Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/04—Wireless resource allocation
- H04W72/044—Wireless resource allocation based on the type of the allocated resource
- H04W72/046—Wireless resource allocation based on the type of the allocated resource the resource being in the space domain, e.g. beams
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention provides a combined beam distribution method based on a Bayesian parameter-adjusting support vector machine, which comprises the steps of firstly generating a beam distribution scheme space to be selected, and then randomly or according to a selection scheme in a similar historical distribution record to generate an initial beam distribution scheme set; the method is applied to a cellular mobile communication system, the evaluation of the system on a scheme is obtained, and a training sample is generated by labeling; the Bayesian hyper-parameter optimization method is combined with cross validation to perform parameter optimization, and then a training sample in a (scheme, label) format is learned by using a support vector machine learning algorithm to obtain corresponding constraint conditions; and finally, using a beam scheme optimization algorithm and combining the obtained constraint conditions to screen the beam allocation scheme in the beam scheme space to generate a next beam allocation scheme set. The invention solves the technical problems of low matching degree of beam distribution and low distribution efficiency in the prior art, and has higher real-time performance while ensuring the accuracy.
Description
Technical Field
The invention relates to the technical field of wireless communication, and particularly provides a method for sensing interference among multiple cells and allocating combined beams.
Background
In recent years, with the intensive research on machine learning, machine learning algorithms are widely applied to problem solution in various fields, such as speech recognition, image recognition, and the like, which indicates that machine learning algorithms have strong universality. In the field of wireless communication research, a machine learning algorithm can also be used for solving the traditional communication problem, and both beam forming and signal detection have many successful research applications.
Compared with other machine learning algorithms, a Support Vector Machine (SVM) can generally achieve better effect on a small sample training set and has excellent generalization capability, because the optimization goal of the SVM is to minimize the risk of structuring rather than experience risk, and therefore, the structured description of data distribution is obtained through the concept of margin, so that the requirements on data scale and data distribution are reduced, but the traditional optimization method has low beam allocation efficiency and high computation complexity, so that the information rate is influenced.
Disclosure of Invention
In order to solve the problem that the traditional optimization method is low in beam distribution efficiency and high in computation complexity, the invention provides a combined beam distribution method based on a Bayesian parameter-adjusting support vector machine.
In order to solve the technical problems, the invention provides the following technical scheme:
a combined beam distribution method based on a Bayesian parameter-adjusting support vector machine comprises the following steps:
s1, generating a beam distribution scheme space to be selected according to the distribution of a current cell and the link connection condition;
s2, selecting l schemes randomly or according to similar historical distribution records in a generated candidate scheme space to generate an initial beam distribution scheme set;
s3, successively applying the schemes in the beam allocation scheme set to the cellular mobile communication system, obtaining the evaluation of the system to the scheme, and labeling the corresponding schemes according to the evaluation to generate training samples; preferably, the distribution scheme samples with performance higher than the average performance of the distribution scheme set are labeled with 1, and the distribution scheme samples are labeled with-1 otherwise;
s4, carrying out parameter optimization by combining a Bayesian hyper-parameter optimization method with cross validation, and then learning the training sample in a scheme and label format by using a support vector machine learning algorithm to obtain corresponding constraint conditions;
preferably, the beam allocation scheme matrix is zero-padded and vectorized
Where I is the number of samples which then together with the corresponding label constitute a training sample->
Inputting the SVM model for parameter tuning and cross validation.
S5, using a beam scheme optimization algorithm and combining the obtained constraint conditions to screen a beam distribution scheme in a beam scheme space to generate a next beam distribution scheme set; outputting boundary constraint conditions after SVM model learning is finished
The beam scheme selection algorithm bases on £ er>
A new i samples are chosen in the beam allocation scheme space with the maximized minimum distance to all the marked samples. Also, when the number of learns reaches REC, the previous boundary constraint is consulted, and +>
Together, the beam allocation scheme space is constrained, and the adjustment function of exploration and utilization is played.
And S6, repeating S3-S5 until a preset maximum iteration number or a beam distribution scheme optimization algorithm is reached, and failing to pick up a new distribution scheme.
Further, the specific method for generating the candidate beam allocation scheme space according to the distribution of the current cell and the link connection condition in S1 is as follows:
s11 identifying and recording traffic conditions as a matrix S t Including distance and angle information of the vehicle and the base station;
s12 according to S t Generating a beam allocation scheme space Ω t Including all possible beam allocation scheme matrices B t 。
Further, S2, inquiring from the database and sending the past S t′ And S t Performing cosine similarity judgment;
wherein vec (S) t ) S representing zero completion and vectorization t So as to obtain a homography matrix and compare the similarity of vectorization, and identify similar traffic modes. If S is present t′ Satisfy the sum of S t Similarly, the traffic pattern S is read t′ As S, the optimal beam allocation scheme t The initial beam allocation scheme of (1). If not, at omega randomly t Selecting.
Has the advantages that:
aiming at the problem of beam distribution, the invention adopts a machine learning algorithm solution based on a support vector machine, adopts Bayesian super-parameter optimization to enhance the beam distribution performance and simultaneously reduces the calculation complexity, and realizes the generation of an optimal beam distribution scheme suitable for the traffic condition of the current cell according to the distribution and link connection condition of the mobile communication cell so as to maximize the average link information rate of all cells.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
fig. 1 is a flowchart of a joint beam allocation method based on a bayesian parameter-adjusting support vector machine according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a traffic condition of a cell according to an embodiment;
FIG. 3 illustrates the performance of the embodiment in providing a joint beam allocation method based on a Bayesian parameter-adjusting support vector machine with different REC parameters;
FIG. 4 is a comparison between the performance of the combined beam allocation method based on the Bayesian parameter adjusting support vector machine and other methods in the fixed traffic mode;
fig. 5 is a comparison between the performance of the joint beam allocation method based on the bayesian parameter-adjusting support vector machine and other methods in the dynamic traffic mode according to the embodiment.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it should be understood that they are presented herein only to illustrate and explain the present invention and not to limit the present invention.
Examples
A combined beam distribution method based on Bayesian parameter adjustment comprises the following steps:
cell distribution and link state information input, state analysis and beam distribution scheme generator, beam distribution scheme database, distribution scheme optimizer and Bayesian parameter-adjusting support vector machine learner. And finally obtaining a near-optimal beam distribution scheme through multiple iterations. FIG. 1 shows a flow chart of an embodiment of the method of the present invention.
The low complexity of the embodiment means that compared with a traditional optimization problem-based method, the method requires a machine learning mode with less calculation amount, and can search an implicit relation, so that the method can quickly converge to a near-optimal beam allocation scheme, and the allocation efficiency is improved.
The distribution accuracy of the model is supported by cross validation, and the 5-fold cross validation is adopted to powerfully ensure the precision of beam distribution. Meanwhile, the common MATLAB software self-contained fitcsvm is adopted to implement parameter tuning and model training of the SVM model, and a parallel mode is adopted to accelerate the learning speed during training. In this embodiment, the maximum learning time ITER for the fixed traffic pattern is set to 14 times, the history constraint boundary reference period REC is set to 5, and the number of elements l of the newly generated allocation plan set is 5.
Firstly, at time t, according to the distribution and link connection condition of the current cellular system cell, a traffic pattern matrix S is generated by a traffic pattern analyzer t And thereby generating candidate beam allocation scheme space omega t The beam allocation space should contain all or most of the allocation scheme matrix B t . Scene schematic diagram of the embodiment as shown in fig. 2, the present embodiment also takes permanent and temporary shadow effects into account.
Next, it is checked whether a record of highly similar traffic patterns exists in the database. If so, reloading the last learning data, if the same traffic mode does not exist in the database, initializing learning, and then entering a exploration or utilization stage according to the learning record, such asFruit existence S t′ Satisfy the sum of S t Similarly, the traffic pattern S is read t′ As S t The initial beam allocation scheme of (1). If not, it is randomly at omega t Selecting. At either of these two stages, the algorithm will generate l samples to form an initial solution set D 0t . Second, the performance of the solutions is collected and labeled to enable the support vector machine model. For example, in the second iteration, after the feedback collection phase, a tagging function is invoked, feeding back the initial solution set D according to the performance of the cellular system 0t The samples in (1) are labeled. Then, after adjusting the hyper-parameters, cross-validating and model training, the support vector machine model will learn the intrinsic relations from these samples and labels, and then based on the data set
Generating a decision boundary h 1t (x)＝0。
Third, a set of solutions for the next iteration is generated and the performance is collected, e.g., assuming that in the first iteration m in/samples is marked as "1", which means that m solutions are excluded. Then, in a second iteration, the assignment scheme optimization algorithm picks out m schemes in the scheme space constrained by the decision boundary set. The algorithm selects each solution to maximize its minimum distance from all other labeled solutions. To speed up the learning process, the decision boundaries are recorded after REC iterations, where REC is an adjustable parameter to balance utilization (small REC) and exploration (large REC), see fig. 3.
Fourth, the solutions originally labeled "1" and newly generated solutions are labeled, e.g., in a second iteration, after the feedback collection phase, instead of labeling all 2l-m solutions, only the solutions already labeled "1" and those of the newly generated sample set are labeled.
Fifth, the updated set of solutions is trained and learned, for example, in a third iteration by learning these new solutions labeled "1" and the solutions labeled "-1" from the second iteration "L-m schemes of (1), will be in the current data set
Generates new decision boundaries and then uses them to reduce the solution space.
The algorithm repeats these steps until the termination criteria are met, all over again and again. For the traffic pattern at time t, the termination criterion of the algorithm is that the beam allocation scheme selector cannot pick out a new sample or reach the maximum number of iterations ITER. FIG. 4 provides a comparison of the performance of the method of the present invention with other methods when directed to a fixed mode of transportation; FIG. 5 provides a comparison of the performance of the method of the present invention in dynamic traffic mode with other methods; as can be seen from fig. 4 and 5, the present invention can generate an optimal beam allocation scheme suitable for the traffic condition of the current cell according to the distribution and link connection conditions of the mobile communication cells so as to maximize the average link information rate of all cells.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A combined beam distribution method based on a Bayesian parameter-adjusting support vector machine is characterized in that a support vector machine model and a Gaussian kernel function are adopted, and a Bayesian super-parameter optimization method is utilized to optimize super-parameters, and the method specifically comprises the following steps:
s1, generating a beam distribution scheme space to be selected according to the distribution of a current cell and the link connection condition;
s2, selecting one scheme from the generated scheme space to be selected randomly or according to similar historical distribution records to generate an initial beam distribution scheme set;
s3, successively applying the schemes in the beam allocation scheme set to the cellular mobile communication system, obtaining the evaluation of the system to the scheme, and labeling the corresponding schemes according to the evaluation to generate training samples;
s4, carrying out parameter optimization by combining a Bayesian hyper-parameter optimization method with cross validation, and then learning the training sample in a scheme and label format by using a support vector machine learning algorithm to obtain corresponding constraint conditions;
s5, using a beam scheme optimization algorithm and combining the obtained constraint conditions to screen a beam distribution scheme in a beam scheme space to generate a next beam distribution scheme set;
and S6, repeating S3-S5 until a preset maximum iteration number or a beam distribution scheme optimization algorithm is reached, and a new distribution scheme cannot be picked up.
2. The joint beam allocation method based on the bayesian parametrization support vector machine according to claim 1, wherein the specific method for generating the candidate beam allocation scheme space according to the distribution of the current cell and the link connection condition in S1 is as follows:
s11 identifying and recording traffic conditions as a matrix S t Including distance and angle information of the vehicle and the base station;
s12 according to S t Generating a beam allocation scheme space omega t Including all possible beam allocation scheme matrices B t 。
3. The combined beam distribution method based on the bayesian parametrization support vector machine according to claim 1, wherein S2 performs the similarity judgment of the traffic patterns and the identification of the same traffic patterns by the cosine similarity of the traffic pattern matrix after the zero completion and the vectorization, and the formula is as follows:
wherein S t Is a traffic pattern matrix at time t, S t′ Is a traffic condition matrix at the time t'; vec (S) t ) S representing zero completion and vectorization t So as to obtain a homography matrix and compare the similarity of vectorization, and identify similar traffic modes.
4. The bayesian-parametrization support vector machine-based joint beam allocation method of claim 1, wherein the samples of the allocation scheme in S3 having a performance higher than the average performance of the set of allocation schemes are labeled as 1 and vice versa as-1.
5. The joint beam allocation method based on the bayesian parametrization support vector machine as recited in claim 1, wherein the beam allocation scheme matrix is zero-padded and vectorized in S4
Where I is the number of samples which then together with the corresponding label constitute a training sample->
Inputting an SVM model for parameter tuning and cross validation;
B it is the ith beam allocation scheme matrix at time t; then vectorizing it with
Samples characterizing a beam allocation scheme;
Is a label for the scheme, characterizing performance.
6. The joint beam allocation method based on the Bayesian parameter-adjusting support vector machine according to claim 1 or 5, wherein S4 is used for sample training by adopting an SVM classifier model of a dual form of a Gaussian kernel function.
7. The Bayesian parameter-adjusting support vector machine-based joint beam distribution method according to claim 1, wherein S5 employs a beam scheme optimization algorithm and a boundary constraint condition reference period to improve beam space contraction; outputting boundary constraint conditions after SVM model learning is finished
Beam scheme selection algorithm based on &>
A new i samples are chosen in the beam allocation scheme space with the maximized minimum distance to all the marked samples. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011115056.3A CN112261721B (en) | 2020-10-19 | 2020-10-19 | Combined beam distribution method based on Bayes parameter-adjusting support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011115056.3A CN112261721B (en) | 2020-10-19 | 2020-10-19 | Combined beam distribution method based on Bayes parameter-adjusting support vector machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112261721A CN112261721A (en) | 2021-01-22 |
| CN112261721B true CN112261721B (en) | 2023-03-31 |
Family
ID=74243815
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011115056.3A Active CN112261721B (en) | 2020-10-19 | 2020-10-19 | Combined beam distribution method based on Bayes parameter-adjusting support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112261721B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111144581A (en) * | 2019-12-31 | 2020-05-12 | 杭州雅拓信息技术有限公司 | Machine learning hyper-parameter adjusting method and system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10405208B2 (en) * | 2017-11-03 | 2019-09-03 | Salesforce.Com, Inc. | Automatic performance monitoring and health check of learning based wireless optimization framework |
| CN109376869A (en) * | 2018-12-25 | 2019-02-22 | 中国科学院软件研究所 | A kind of super ginseng optimization system of machine learning based on asynchronous Bayes optimization and method |
| CN109617584B (en) * | 2019-01-08 | 2021-12-21 | 南京邮电大学 | MIMO system beam forming matrix design method based on deep learning |
-
2020
- 2020-10-19 CN CN202011115056.3A patent/CN112261721B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111144581A (en) * | 2019-12-31 | 2020-05-12 | 杭州雅拓信息技术有限公司 | Machine learning hyper-parameter adjusting method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112261721A (en) | 2021-01-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114386694B (en) | Drug molecular property prediction method, device and equipment based on contrast learning | |
| US10070264B2 (en) | Automated WLAN radio map construction method and system | |
| CN111191732A (en) | Target detection method based on full-automatic learning | |
| CN111274958B (en) | Pedestrian re-identification method and system with network parameter self-correction function | |
| CN110942091A (en) | Semi-supervised few-sample image classification method for searching reliable abnormal data center | |
| CN113095251B (en) | Human body posture estimation method and system | |
| Dewan et al. | Learning a local feature descriptor for 3d lidar scans | |
| CN114329031B (en) | Fine-granularity bird image retrieval method based on graph neural network and deep hash | |
| CN111985547A (en) | Cross-resolution vehicle re-identification method based on attention-guided generation learning | |
| Zeng et al. | Robust multivehicle tracking with wasserstein association metric in surveillance videos | |
| CN117933748B (en) | Urban land use planning method and system based on machine learning | |
| CN109121081B (en) | Indoor positioning method based on position candidate set and EM algorithm | |
| CN115063664A (en) | Model learning method, training method and system for industrial vision detection | |
| CN112261721B (en) | Combined beam distribution method based on Bayes parameter-adjusting support vector machine | |
| CN113837046A (en) | Small sample remote sensing image scene classification method based on iterative feature distribution learning | |
| CN117854104A (en) | Feature alignment-based unsupervised pedestrian re-identification method | |
| US20230368038A1 (en) | Improved fine-tuning strategy for few shot learning | |
| CN114116669B (en) | Huo Fuding tree-based multi-label stream data classification method | |
| CN114970682A (en) | Pseudo label propagation optimization time sequence clustering method based on multiple regularization constraints | |
| CN115329821A (en) | Ship noise identification method based on pairing coding network and comparison learning | |
| CN114020948A (en) | Sketch image retrieval method and system based on sorting clustering sequence identification selection | |
| CN111737386A (en) | Two-stage neighbor graph searching method applied to large-scale high-dimensional data | |
| CN114298168B (en) | Three-dimensional point cloud processing method and system | |
| CN113572732B (en) | Multi-step attack modeling and prediction method based on VAE and aggregated HMM | |
| CN114463602B (en) | Target identification data processing method based on big data |
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 |