CN111130557A - Data reconstruction method based on distributed quasi-Newton projection tracking - Google Patents
Data reconstruction method based on distributed quasi-Newton projection tracking Download PDFInfo
- Publication number
- CN111130557A CN111130557A CN201911409497.1A CN201911409497A CN111130557A CN 111130557 A CN111130557 A CN 111130557A CN 201911409497 A CN201911409497 A CN 201911409497A CN 111130557 A CN111130557 A CN 111130557A
- Authority
- CN
- China
- Prior art keywords
- data
- support set
- reconstructed
- computing node
- representing
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 239000011159 matrix material Substances 0.000 claims abstract description 28
- 238000005259 measurement Methods 0.000 claims abstract description 23
- 238000004364 calculation method Methods 0.000 claims description 12
- 239000000126 substance Substances 0.000 claims description 4
- 238000011084 recovery Methods 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
- H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a data reconstruction method based on distributed quasi-Newton projection tracking, which comprises the following steps: dividing data to be reconstructed into a common part and an individual part; each computing node sends a support set of data to be reconstructed to a neighbor computing node; each computing node obtains a public part support set according to the support set computed by the computing node and the support set obtained from the neighbor computing node; and each computing node iteratively reconstructs the compressed data according to the common part support set, the collected compressed data, the measurement matrix and the sparsity of the common part and the individual part. In the data reconstruction process, each computing node can obtain data in the whole network to obtain global information, and the method can be applied to data recovery in a distributed network with public components and meet the requirements of more scenes; and the data reconstruction speed is higher, and the precision is higher.
Description
Technical Field
The invention relates to a data reconstruction method based on distributed quasi-Newton projection tracking, and belongs to the technical field of data reconstruction.
Background
With the continuous development of information technology, the scale of data acquisition by people is larger and larger. How to be able to quickly and efficiently reconstruct data plays an important role in many distributed applications. Data to be processed in the distributed system is dispersed in a plurality of nodes, and the data needs to be transmitted to a server for joint reconstruction of the data. Therefore, a large amount of information needs to be transferred in the distributed network at the time of data reconstruction, thereby causing consumption of bandwidth and delay of data reconstruction.
At present, a centralized processing method is mainly adopted for data reconstruction in a distributed system, all node data are transmitted to a server for centralized processing, a large amount of bandwidth is consumed, a long time delay is generated, and the requirement for rapid and accurate reconstruction of compressed data in distributed systems with larger and larger scales cannot be met. The existing distributed Bayesian algorithm carries out iterative reconstruction by decomposing data into a public part and an individual part and utilizing variational Bayesian inference, but because data information of the public part is only interacted among all computing nodes, all the computing nodes cannot acquire global information and cannot meet certain application scenarios (for example, in intelligent transportation, all the computing nodes need to know the global information for adjustment), and the speed and the precision of data reconstruction still need to be improved.
Disclosure of Invention
The invention aims to provide a data reconstruction method based on distributed quasi-Newton projection tracking, which can effectively solve the problems in the prior art, realize faster and more accurate data reconstruction, and simultaneously can acquire global information to meet the requirements of more scenes.
In order to solve the technical problems, the invention adopts the following technical scheme: the data reconstruction method based on the distributed quasi-Newton projection tracking comprises the following steps: dividing data to be reconstructed into a common part and an individual part; each computing node sends a support set of data to be reconstructed to a neighbor computing node; each computing node obtains a public part support set according to the support set computed by the computing node and the support set obtained from the neighbor computing node; and each computing node iteratively reconstructs the compressed data according to the common part support set, the collected compressed data, the measurement matrix and the sparsity of the common part and the individual part.
Preferably, the method specifically comprises the following steps:
s1, dividing the data to be reconstructed into a public part and an individual part; initializing a support set of data to be reconstructed of each computing node and a data residual error;
s2, each computing node sends the latest support set of the data to be reconstructed to the neighbor computing nodes;
s3, each computing node obtains a public part support set according to the latest support set computed by the computing node and the latest support set obtained from the neighbor computing node;
s4, obtaining an updated support set of data to be reconstructed, updated reconstructed sparse data and updated data residual errors by each computing node according to the common part support set, the collected compressed data, the collected measurement matrix and the sparsity of the common part and the individual part by using an MODQNPP function;
s5, determine if the square of the two-norm of the updated data residual is less than the square of the two-norm of the last obtained data residual? If yes, go to S2; otherwise, outputting the reconstructed sparse data obtained last time as final reconstructed data.
The data reconstruction is carried out by the method, particularly, the data reconstruction is carried out by combining with an MODQNPP function, so that the data iteration times in the reconstruction process are less, the reconstruction speed is higher, the reconstruction precision is higher, and each computing node can acquire the global information of the network.
More preferably, in step S3, the compute node selects the top K with the highest frequency of occurrence according to the latest support set computed by itself and the latest support set obtained from the neighboring compute node(c)As a common part supporting set, wherein K(c)Is the sparsity of the common part. The method obtains the common part support set, thereby ensuring that the calculation is simpler under the condition of being as accurate as possible and being beneficial to improving the speed of data reconstruction.
In the aforementioned data reconstruction method based on distributed quasi-newton projection tracking, step S4 specifically includes the following steps: and S41, initializing data:if it isThen r is0=resid(y,Ax0) Otherwisek=0,x0=0,Wherein the content of the first and second substances,sparsity K representing common part(c)And sparsity of individual partsSumming; t is0A set of supports is represented that is,a supporting set representing a common part in the entire network,representing the measurement matrix A on the basis of the support set T0Y represents the collected compressed data, x0Sparse signal representing initialization reconstruction, and resid representing the calculation y anda difference of (d);
s42, let k equal to k +1, calculateWherein d iskDenotes the Newton direction, xk-1Representing the reconstructed sparse signal obtained at the k-1 st iteration, I representing the identity matrix, ΛkRepresenting a vector middle frontIndex value of maximum value, and taking corresponding calculation result based on ΛkProjection of (2);
s43, calculatingWherein, mukDenotes the step size, k denotes the number of iterations, Tk-1Representing the support set calculated in the last iteration,representing the measurement matrix A on the basis of Tk-1Projection of (2);
s44, calculating and obtaining an updated support set T of the data to be reconstructedk:Wherein, TkRepresenting the support set, x, obtained at the k-th iterationk-1Denotes the reconstructed sparse signal obtained at the k-1 st iteration, and max _ indices denotes xk-1+μkdkFront ofAn index value of the maximum value;
s45, calculating and obtaining the updated data x to be reconstructedk:Wherein T iskRepresenting the support set calculated for the kth iteration,representing the measurement matrix A on the basis of TkProjection of (2);
S47, judgmentIs there any? If yes, go to S42, otherwise, directly output T obtained by the iterationk、xkAnd rk。
T is obtained by the above methodk、xkAnd rkSimple calculation, high reconstruction precision and high reconstruction speed, and particularly the method for reconstructing the dkThe updating method of the invention ensures that the data reconstruction precision is higher and the reconstruction speed is higher.
In the data reconstruction method based on distributed quasi-newton projection tracking, in step S1, the support set of the data to be reconstructed and the data residual of each computing node are initialized by the following method: initializing sparsity of a public part and an individual part, and obtaining a support set of data to be reconstructed and a data residual error according to the collected compressed data, a measurement matrix and sparsity of the public part and the individual part by using an MODQNPP function; wherein the common part support set is set empty.
In the above data reconstruction method based on distributed quasi-newton projection tracking, in step S2, the compressed data is iteratively reconstructed, that is, an objective function is constructed so that it converges:
wherein y is the collected compressed data, a is the measurement matrix, and x is the sparse signal to be reconstructed.
Compared with the prior art, the support set calculated by each calculation node in each iteration is exchanged with other calculation nodes, so that the support set of the public sparse part can be obtained and is used as data of the next iteration for iterative calculation until the data is converged. In the data reconstruction process, each computing node can obtain data in the whole network to obtain global information, and the method can be applied to data recovery in a distributed network with public components and meet the requirements of more scenes; and the data reconstruction speed is higher, and the precision is higher.
In addition, the inventor randomly selects K cables not more than n in simulation data (namely, the measurement matrix A is a randomly generated m multiplied by n matrix, and K cables not more than n are randomly selected through the analysis of the existing data and a large amount of experimental comparisonIndexing values and randomly generating data at their corresponding index values as a sparse signal x to be reconstructediBy means of a randomly generated signal xiCalculating yi=Aixi+eiWherein e isiGaussian noise). At a calculation node number of 5, sparsity120 m and 300 n, support setSupporting set of common parts in the whole networkk=0,x00, data residual r0=yiUnder the condition, the reconstruction error of the data is 5.5965e-4 by using the method, which is improved by 16.8% and the reconstruction speed is improved by 21.3% compared with the reconstruction error of 6.7254e-4 of the existing distributed compressed data reconstruction algorithm (namely the 'distributed Bayesian algorithm' mentioned in the background technology); under the condition that a data set of real data (2012 year black river basin midstream ecological hydrological wireless sensor network soil temperature observation data set http:// westdc. westgis. ac. cn/heihe/view/uid/0 a2e1ce6-f322-4d0f-82ee-70446123dba1) is adopted, when the number of nodes is calculated to be 35, m is 60, and n is 128, the data reconstruction error is 0.0083 by using the method provided by the invention, compared with the existing distributed compressed data reconstruction algorithm (namely the distributed Bayesian algorithm mentioned in the background technology), the reconstruction error of the distributed compressed data reconstruction algorithm designed by using the method provided by the invention is 0.0095 and is improved by 12.6%, and the reconstruction speed is improved by 15.4%.
Drawings
FIG. 1 is a method flow diagram of one embodiment of the present invention;
FIG. 2 is a flow chart of a method of obtaining an updated support set of data to be reconstructed, updated reconstructed sparse data, and updated data residuals.
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
Detailed Description
The embodiment of the invention comprises the following steps: the data reconstruction method based on the distributed quasi-Newton projection tracking, as shown in FIG. 1, includes the following steps: dividing data to be reconstructed into a common part and an individual part; each computing node sends a support set of data to be reconstructed to a neighbor computing node; each computing node obtains a public part support set according to the support set computed by the computing node and the support set obtained from the neighbor computing node; and each computing node iteratively reconstructs the compressed data according to the common part support set, the collected compressed data, the measurement matrix and the sparsity of the common part and the individual part.
In order to further improve the accuracy and speed of data reconstruction, optionally, the method specifically includes the following steps:
s1, dividing the data to be reconstructed into a public part and an individual part; initializing a support set of data to be reconstructed of each computing node and a data residual error;
s2, each computing node sends the latest support set of the data to be reconstructed to the neighbor computing nodes;
s3, each computing node obtains a public part support set according to the latest support set computed by the computing node and the latest support set obtained from the neighbor computing node;
s4, obtaining an updated support set of data to be reconstructed, updated reconstructed sparse data and updated data residual errors by each computing node according to the common part support set, the collected compressed data, the collected measurement matrix and the sparsity of the common part and the individual part by using an MODQNPP function;
s5, determine if the square of the two-norm of the updated data residual is less than the square of the two-norm of the last obtained data residual? If yes, go to S2; otherwise, outputting the reconstructed sparse data obtained last time as final reconstructed data.
In step S4, the subspace tracking (SP) and the compressed sample matching pursuit (CoSaMP) algorithm may be further used to obtain the updated support set of the data to be reconstructed, the updated reconstructed sparse data, and the updated data residual.
Optionally, in the present invention, in step S3, the computing node selects the top K with the highest frequency of occurrence according to the latest support set computed by itself and the latest support set obtained from the neighboring computing node(c)As a common part supporting the sets (i.e.Wherein the content of the first and second substances,a supporting set representing a common part in the entire network,represents a collection of other computing nodes that are communicable with the current computing node,a set of supporting sets representing all data computed by the current compute node and other compute nodes in communication with the current compute node), where K(c)Is the sparsity of the common part.
Optionally, as shown in fig. 2, step S4 specifically includes the following steps:
and S41, initializing data:if it isThen r is0=resid(y,Ax0) Otherwisek=0,x0=0,Wherein the content of the first and second substances,sparsity K representing common part(c)And sparsity of individual partsSumming; t is0A set of supports is represented that is,a supporting set representing a common part in the entire network,representing the measurement matrix A on the basis of the support set T0Y represents the collected compressed data, x0Sparse signal representing initialization reconstruction, and resid representing the calculation y anda difference of (d);
s42, let k equal to k +1, calculateWherein d iskDenotes the Newton direction, xk-1Representing the reconstructed sparse signal obtained at the k-1 st iteration, I representing the identity matrix, ΛkRepresenting a vector middle frontIndex value of maximum value, and taking corresponding calculation result based on ΛkProjection of (2);
s43, calculatingWherein, mukDenotes the step size, k denotes the number of iterations, Tk-1Representing the support set calculated in the last iteration,representing the measurement matrix A on the basis of Tk-1Projection of (2);
s44, calculating and obtaining an updated support set T of the data to be reconstructedk:Wherein, TkRepresenting the support set, x, obtained at the k-th iterationk-1Denotes the reconstructed sparse signal obtained at the k-1 st iteration, and max _ indices denotes xk-1+μkdkFront ofAn index value of the maximum value;
s45, calculating and obtaining the updated data x to be reconstructedk:Wherein T iskRepresenting the support set calculated for the kth iteration,representing the measurement matrix A on the basis of TkProjection of (2);
S47, judgmentIs there any? If yes, go to S42, otherwise, directly output T obtained by the iterationk、xkAnd rk。
Optionally, in step S1, the support set of the data to be reconstructed and the data residuals of each computing node are initialized by the following method: initializing sparsity of a public part and an individual part, and obtaining a support set of data to be reconstructed and a data residual error according to the collected compressed data, a measurement matrix and sparsity of the public part and the individual part by using an MODQNPP function; wherein the common part support set is set empty.
Optionally, in step S2, the compressed data is iteratively reconstructed, that is, an objective function is constructed so that it converges:
wherein y is the collected compressed data, a is a measurement matrix which is an m × n matrix, and x is the sparse signal to be reconstructed.
The distributed system in the invention is composed of N computing nodes, wherein each computing node manages p member nodes, each computing node comprises data in each member node, and q is the sum of all data contained in the node.
Claims (6)
1. The data reconstruction method based on the distributed quasi-Newton projection tracking is characterized by comprising the following steps of: dividing data to be reconstructed into a common part and an individual part; each computing node sends a support set of data to be reconstructed to a neighbor computing node; each computing node obtains a public part support set according to the support set computed by the computing node and the support set obtained from the neighbor computing node; and each computing node iteratively reconstructs the compressed data according to the common part support set, the collected compressed data, the measurement matrix and the sparsity of the common part and the individual part.
2. The data reconstruction method based on the distributed quasi-Newtonian projection tracking according to claim 1, comprising the following steps:
s1, dividing the data to be reconstructed into a public part and an individual part; initializing a support set of data to be reconstructed of each computing node and a data residual error;
s2, each computing node sends the latest support set of the data to be reconstructed to the neighbor computing nodes;
s3, each computing node obtains a public part support set according to the latest support set computed by the computing node and the latest support set obtained from the neighbor computing node;
s4, obtaining an updated support set of data to be reconstructed, updated reconstructed sparse data and updated data residual errors by each computing node according to the common part support set, the collected compressed data, the collected measurement matrix and the sparsity of the common part and the individual part by using an MODQNPP function;
s5, determine if the square of the two-norm of the updated data residual is less than the square of the two-norm of the last obtained data residual? If yes, go to S2; otherwise, outputting the reconstructed sparse data obtained last time as final reconstructed data.
3. The method for reconstructing data based on distributed quasi-Newtonian projection pursuit as claimed in claim 2, wherein in step S3, the top K with the highest frequency of occurrence is selected by the computing node according to the latest support set computed by itself and the latest support set obtained from the neighboring computing nodes(c)As a common part supporting set, wherein K(c)Is the sparsity of the common part.
4. The method for reconstructing data based on distributed quasi-Newtonian projection tracking according to claim 2, wherein the step S4 comprises the following steps:
and S41, initializing data:if it isThen r is0=resid(y,Ax0) Otherwisek=0,x0=0,Wherein the content of the first and second substances,sparsity K representing common part(c)And sparsity of individual partsSumming; t is0A set of supports is represented that is,a supporting set representing a common part in the entire network,representing the measurement matrix A on the basis of the support set T0Y represents the collected compressed data, x0Sparse signal representing initialization reconstruction, and resid representing the calculation y anda difference of (d);
s42, let k equal to k +1, calculateWherein d iskDenotes the Newton direction, xk-1Representing the reconstructed sparse signal obtained at the k-1 st iteration, I representing the identity matrix, ΛkRepresenting a vector middle frontIndex value of maximum value, and taking corresponding calculation result based on ΛkProjection of (2);
s43, calculatingWherein, mukDenotes the step size, k denotes the number of iterations, Tk-1Representing the support set calculated in the last iteration,representing the measurement matrix A on the basis of Tk-1Projection of (2);
s44, calculating and obtaining an updated support set T of the data to be reconstructedk:Wherein, TkRepresenting the support set, x, obtained at the k-th iterationk-1Denotes the reconstructed sparse signal obtained at the k-1 st iteration, and max _ indices denotes xk-1+μkdkFront ofAn index value of the maximum value;
s45, calculating and obtaining the updated data x to be reconstructedk:Wherein T iskRepresenting the support set calculated for the kth iteration,representing the measurement matrix A on the basis of TkProjection of (2);
5. The method for reconstructing data based on distributed quasi-Newtonian projection pursuit according to claim 2 or 4, wherein in step S1, the support set of data to be reconstructed and the data residuals of each computing node are initialized by the following method: initializing sparsity of a public part and an individual part, and obtaining a support set of data to be reconstructed and a data residual error according to the collected compressed data, a measurement matrix and sparsity of the public part and the individual part by using an MODQNPP function; wherein the common part support set is set empty.
6. The method for reconstructing data based on distributed quasi-newton projection pursuit according to claim 1, wherein in step S2, the compressed data is iteratively reconstructed, i.e. an objective function is constructed so that it converges:
wherein y is the collected compressed data, a is the measurement matrix, and x is the sparse signal to be reconstructed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911409497.1A CN111130557B (en) | 2019-12-31 | 2019-12-31 | Data reconstruction method based on distributed quasi-Newton projection tracking |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911409497.1A CN111130557B (en) | 2019-12-31 | 2019-12-31 | Data reconstruction method based on distributed quasi-Newton projection tracking |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111130557A true CN111130557A (en) | 2020-05-08 |
CN111130557B CN111130557B (en) | 2023-11-17 |
Family
ID=70506801
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911409497.1A Active CN111130557B (en) | 2019-12-31 | 2019-12-31 | Data reconstruction method based on distributed quasi-Newton projection tracking |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111130557B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101895297A (en) * | 2010-07-30 | 2010-11-24 | 哈尔滨工业大学 | Compressed sensing-oriented block-sparse signal reconfiguring method |
CN102611455A (en) * | 2012-03-05 | 2012-07-25 | 哈尔滨工业大学 | Compressed sensing-oriented sparse multiband signal reconstruction method |
CN103300859A (en) * | 2013-05-31 | 2013-09-18 | 王勇 | High-quality rapid CS-MRI (Compressed Sensing-Magnetic Resonance Imaging) method of mixed norm |
CN106291452A (en) * | 2016-09-12 | 2017-01-04 | 电子科技大学 | A kind of direction of arrival estimation method based on enhanced greedy algorithm |
CN106596724A (en) * | 2016-12-08 | 2017-04-26 | 清华大学 | Method for sparse compression optimization reconstruction of narrowband Lamb wave detection data |
US20170300021A1 (en) * | 2016-04-18 | 2017-10-19 | Tsinghua University | Frequency control method for micro-grid and control device |
CN108282788A (en) * | 2018-01-19 | 2018-07-13 | 南京航空航天大学 | A kind of resource allocation methods of the Energy Efficient based on quasi- newton interior point method |
-
2019
- 2019-12-31 CN CN201911409497.1A patent/CN111130557B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101895297A (en) * | 2010-07-30 | 2010-11-24 | 哈尔滨工业大学 | Compressed sensing-oriented block-sparse signal reconfiguring method |
CN102611455A (en) * | 2012-03-05 | 2012-07-25 | 哈尔滨工业大学 | Compressed sensing-oriented sparse multiband signal reconstruction method |
CN103300859A (en) * | 2013-05-31 | 2013-09-18 | 王勇 | High-quality rapid CS-MRI (Compressed Sensing-Magnetic Resonance Imaging) method of mixed norm |
US20170300021A1 (en) * | 2016-04-18 | 2017-10-19 | Tsinghua University | Frequency control method for micro-grid and control device |
CN106291452A (en) * | 2016-09-12 | 2017-01-04 | 电子科技大学 | A kind of direction of arrival estimation method based on enhanced greedy algorithm |
CN106596724A (en) * | 2016-12-08 | 2017-04-26 | 清华大学 | Method for sparse compression optimization reconstruction of narrowband Lamb wave detection data |
CN108282788A (en) * | 2018-01-19 | 2018-07-13 | 南京航空航天大学 | A kind of resource allocation methods of the Energy Efficient based on quasi- newton interior point method |
Also Published As
Publication number | Publication date |
---|---|
CN111130557B (en) | 2023-11-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yao et al. | Dr2-net: Deep residual reconstruction network for image compressive sensing | |
Condat et al. | Cadzow denoising upgraded: A new projection method for the recovery of Dirac pulses from noisy linear measurements | |
CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
CN110427654B (en) | Landslide prediction model construction method and system based on sensitive state | |
CN111383741B (en) | Method, device and equipment for establishing medical imaging model and storage medium | |
Li et al. | Compressed sensing via dictionary learning and approximate message passing for multimedia Internet of Things | |
US20210004645A1 (en) | Unified shape representation | |
Jin et al. | Deep learning for seasonal precipitation prediction over China | |
CN111383742A (en) | Method, device, equipment and storage medium for establishing medical imaging model | |
Okuno et al. | Combining multiple forecasts for multivariate time series via state-dependent weighting | |
CN115859116A (en) | Marine environment field reconstruction method based on radial basis function regression interpolation method | |
CN112163106A (en) | Second-order similarity perception image Hash code extraction model establishing method and application thereof | |
CN117454495A (en) | CAD vector model generation method and device based on building sketch outline sequence | |
Quesada-Chacón et al. | Repeatable high-resolution statistical downscaling through deep learning | |
Kartheeswaran et al. | A data-parallelism approach for PSO-ANN based medical image reconstruction on a multi-core system | |
Wei et al. | Probabilistic optimal interpolation for data assimilation between machine learning model predictions and real time observations | |
CN116562650A (en) | Short-term wind power prediction method and device and computer readable storage medium | |
Li et al. | Towards communication-efficient digital twin via AI-powered transmission and reconstruction | |
CN111130557A (en) | Data reconstruction method based on distributed quasi-Newton projection tracking | |
Hernández-Sánchez et al. | Higher order Hamiltonian Monte Carlo sampling for cosmological large-scale structure analysis | |
Zhang et al. | Distributed compressive sensing via LSTM-Aided sparse Bayesian learning | |
Wang et al. | Application of hybrid Kalman filter for improving water level forecast | |
CN114155410A (en) | Graph pooling, classification model training and reconstruction model training method and device | |
Fan et al. | A spatiotemporal-aware weighting scheme for improving climate model ensemble predictions | |
Zhao et al. | Spatio-temporal model combining vmd and am for wind speed prediction |
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 |