CN104570757A - Method and apparatus for operating integrated control assembly - Google Patents
Method and apparatus for operating integrated control assembly Download PDFInfo
- Publication number
- CN104570757A CN104570757A CN201410527571.0A CN201410527571A CN104570757A CN 104570757 A CN104570757 A CN 104570757A CN 201410527571 A CN201410527571 A CN 201410527571A CN 104570757 A CN104570757 A CN 104570757A
- Authority
- CN
- China
- Prior art keywords
- data
- function model
- gridden
- vector
- model
- 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
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
Abstract
The invention relates to a method and apparatus for operating an integrated control assembly. The invention relates to a method calculating function models based on data. The function model is definated via parameter groups and net point data. The net point data comprsies at least partialy same net point data points; the same net point data points of the function models based on data are stored in a storage area for one time; and teh storage area is visited by a plurality of times for calculation of the function models based on the data.
Description
Technical field
The present invention relates to the Control Component for implementing controlling functions especially in a motor vehicle.The present invention relates to the integrating control assembly with the hardware based model computing unit for calculating the function model based on data in addition, and described function model is defined by hyper parameter and gridden data (St ü tzstellendaten).
Background technology
In order to measure for fulfill model in opertaing device, device for controlling engine especially at internal combustion engine, be set using the function model based on data.Represent non-parametric model based on the flexible program of the function model of data and can create from the training data set of training data point (also i.e.) without specifically prespecified.
Example based on the function model of data is so-called Gaussian process model, and it returns based on Gaussian process.It is for carrying out many-sided method of modeling to complicated physical system based on data that Gaussian process returns.Regretional analysis is usually based on large training data set, thus meaningfully, use the solution of approaching, this solution can be asked for (ausgewertet) efficiently.
Function model based on data is defined by gridden data and hyper parameter usually.They are stored in the storage area of oneself of internal storage unit for each function model based on data.Especially, gridden data has significant storage demand, defines more than 100 gridden datas because usually utilize based on the function model of data.Because gridden data has the dimension D of multiple input parameter and (pretreated if desired) output parameter respectively, therefore arrange (dominieren) for the matrix of the storage space needed for Gaussian process model by gridden data.Because usually a large amount of function model must be calculated in opertaing device, be therefore huge for the storage demand calculated as these function models of the function model based on data.
By the known opertaing device with integrating control assembly of prior art, described Control Component has main computation unit and the independent model computing unit for calculating the function model based on data.Thus publication DE 10 2,010 028 266A1 such as illustrates a kind of with the Control Component of additional logical circuit as model computing unit, this logical circuit is configured to purely based on hardware ground gauge index function and addition and multiplying.This makes it possible to realize, and supports the calculating of Bayes (Bayes) homing method in hardware cell, and this Bayesian regression method is required in particular for calculating Gaussian process model.
Model computing unit is designed to perform the mathematical procedure for calculating based on parameter and net point or training data based on the function model of data generally.Especially, be used for the function of the model computing unit of gauge index and summing function efficiently purely with hardware implementing, make it possible to realize: compared with this can carry out in by the main computation unit of software control with higher computing velocity to calculate Gaussian process model.
Preparing to supply a model the calculating of computing unit and before then carried out the calculating based on configuration data of function model by the hardware of model computing unit, usually comprise configuration data, for calculating the parameter (hyper parameter) of the function model based on data and gridden data or the address pointer to address area, wherein gridden data is stored in this address area.
Summary of the invention
Arrange according to method for calculating multiple function model based on data according to claim 1 and according to model computing unit and the integrating control assembly for especially calculating multiple function model based on data in integrated memory assembly described in claim arranged side by side according to the present invention.
Favourable expansion scheme in addition illustrates in the dependent claims.
Arrange the method for calculating multiple function model based on data according to first aspect, described function model is defined by parameter group and gridden data.Gridden data comprises Grid dimension strong point identical at least in part, wherein the identical Grid dimension strong point for multiple function model based on data once is stored in storage area, wherein in order to the function model calculated based on data repeatedly accesses described storage area.
The thought of said method is, by multiple function model based on data based on gridden data only arrange once in the storage area of storage unit.By correspondingly reference, therefore model computing unit can access (common) storage area distributed to and distribute to accordingly multiple function model based on data based on the hyper parameter of the function model of data and access, so that the function model based on data involved by calculating.
Commonly in the prior art strategically for each function model based on data, gridden data and super data are stored in corresponding storage area individually to be different from, in the method for advising above for multiple function model based on data by a part for stored data, also namely especially gridden data only stores once, can save obviously a large amount of storage spaces thus.
Especially for the function model based on data asked for based on identical test device, test macro or checking system or experiment process, also namely in parameter/input space at multiple output parameters of a series of some place measuring system that can be given in advance, said method be suitable for.Therefore the output valve for different measurement parameters or output parameter is mutually different, and normally identical for the coordinate of the measurement point of the training data of all output parameters.
In addition, can produce from the common testing process checking system for the parameter group of multiple function model based on data and gridden data.
According to a kind of embodiment, parameter group can have vector respectively, this DUAL PROBLEMS OF VECTOR MAPPING is accordingly based on the functional value of the function model of data, wherein this vector has N dimension, wherein N is corresponding to the quantity at the Grid dimension strong point of gridden data, wherein when the element distributing to involved Grid dimension strong point of this vector is set to 0, Grid dimension strong point maintenance in one of multiple function model based on data of gridden data is not considered.
Can specify, one of the function model based on data is revised by additional Grid dimension strong point, its mode is, Grid dimension strong point is added and the respective element of the vector of (function model of unmodified) parameter group is set to 0 to described gridden data, a wherein said vector based on the function model of data utilizes the gridden data re-training being expanded Grid dimension strong point, to ask for a described respective element based on the vector of the parameter group of the function model of data.
In addition, described multiple function model based on data can corresponding to Gaussian process model, RBF net or support vector machine (Support Vector Machines).
According to another aspect, arrange the model computing unit for calculating multiple function model based on data in integrating control assembly, described function model is defined by parameter group and gridden data, comprising:
-for the gridden data for multiple function model based on data being stored in the internal storage in common storage area,
Wherein model computing unit is configured to repeatedly territory, access storage areas to calculate the function model based on data.
According to a further aspect, the integrating control assembly of main computation unit and above-mentioned model computing unit is provided with.
Accompanying drawing explanation
Set forth further preferred embodiment by accompanying drawing below.Wherein:
Fig. 1 illustrates the schematic diagram of the integrating control assembly with hardware based model computing unit; And
Fig. 2 illustrates the process flow diagram for showing the method for running the integrating control assembly for calculating multiple function model based on data.
Embodiment
Fig. 1 illustrates the schematic diagram of the hardware structure of the integrating control assembly 1 of such as microcontroller form, arranges main computation unit 2 in an integrated manner and calculate the model computing unit 3 of the function model based on data for pure based on hardware in described integrating control assembly.Main computation unit 2 and model computing unit 3 are connected 4 via intercommunication, such as system bus communicates to connect mutually.
In principle, model computing unit 3 is hard wired substantially and is correspondingly configured to implement software code unlike main computation unit 2.Alternatively, following solution is possible, wherein provides restricted for the model computing unit 3 calculated based on the function model of data, the instruction group of height specialization.Not placement processor in model computing unit 3.This realizes such model computing unit 3 or the area-optimized structure of integrated morphology mode with making it possible to resource optimization.
Model computing unit 3 has calculating core 31, and it implements the calculating of algorithm given in advance purely in hardware.Calculate core 31 to be connected with interrupt location (Abbrucheinheit) 32, wherein when there is interrupt condition, this interrupt location signals the interruption that algorithm calculates.Model computing unit 3 can comprise the local SRAM 33 for store configuration data in addition.Model computing unit 3 can comprise local DMA unit 34(DMA=Direct Memory Access equally, direct memory access).Can the integrated resource of access control components 1 by DMA unit 34, especially access internal storage 5.
Control Component 1 can comprise internal storage 5 and another DMA unit 6(DMA=Direct Memory Access, direct memory access).Internal storage 5 and another DMA unit 6 described in an appropriate manner, be such as connected 4 via intercommunication and be interconnected.Internal storage 5 can comprise (for main computation unit 2, model computing unit 3 and unit other if desired) common SRAM memory and the flash memory for configuration data (parameter and gridden data).
Non-parametric, based on the use of the function model of data based on Bayesian regression method.The basis of Bayesian regression such as exists
deng people "
" be described in (MIT Press 2006).Bayesian regression is the method based on data, and the method is based on model.In order to create this model, the affiliated output data of the measurement point needing training data and the output parameter treating modeling.The establishment of this model is by making for realizing gridden data, and these gridden datas produce corresponding to training data or from described training data completely or partially.In addition, determine abstract hyper parameter, described hyper parameter carries out parametrization to the space of pattern function and is effectively weighted the impact of each measurement point on model prediction afterwards of training data.
Abstract hyper parameter is determined by optimization method.Possibility for this optimization method is edge likelihood (Marginal Likelihood)
optimization.Edge likelihood
describe the rationality (Plausibilitaet) of the measured y value of training data, it is represented as vectorial Y, provides the x value of model parameter H and training data.In model training, right
maximize, its mode finds suitable hyper parameter, and described hyper parameter results through the change curve of the pattern function that hyper parameter and training data are determined and maps training data as far as possible exactly.In order to simplify calculating, maximize
logarithm because logarithm does not change the continuity of rational function.
The calculating of Gaussian process model is carried out corresponding to the step schematically shown in fig. 2.For test point u(input parameter vector) input value
first be standardized, and be standardized according to formula below:
。
At this, m
xcorresponding to the mean function of the mean value of the input value about gridden data, s
ycorresponding to the variance of the input value of gridden data and d corresponding to the index of the dimension D for test point u.
As the result of the establishment of the non-parametric function model based on data, obtain:
。
The model value v asked for like this carrys out standardization by outputting standard, and carrys out standardization according to following formula:
。
At this, v is corresponding to the input parameter vector at standardized test point u(dimension D) the standardized model value (output valve) at place,
corresponding in (nonstandardized technique) test point
(nonstandardized technique) model value (output valve) at (the input parameter vector of dimension D) place,
corresponding to the net point of gridden data, N is corresponding to the quantity of the net point of gridden data, and D is corresponding to the dimension in input data space/training data space/gridden data space, and I
dwith
corresponding to the hyper parameter from model training.Vector
it is the parameter calculated by hyper parameter and training data.In addition, m
ycorresponding to the mean function of the mean value of the output valve about gridden data and S
ycorresponding to the variance of the output valve of gridden data.Be alternatively to above-mentioned strategy, also can carry out the calculating of the function model based on data in nonstandardized technique space, thus cancel input and output standardization.
Because the calculating of Gaussian process model typically occurs in standardized space, therefore perform input and output standardization.
In order to start-up simulation, computing unit 2 especially can indicate DMA unit 34 or other DMA unit 6, and to be transferred to by the configuration data relating to function model to be calculated in model computing unit 3 and start-up simulation, described calculating performs by configuration data.These configuration datas comprise hyper parameter and the gridden data of Gaussian process model, and it preferably illustrates by the address pointer tasking the address area of model computing unit 3 that divides to internal storage 5.Especially, also can use the SRAM memory 33 for model computing unit 3, this SRAM memory especially can be arranged in model computing unit 3 or be arranged in model computing unit 3 place for this reason.Internal storage 5 and SRAM memory 33 also can be used in combination.
Calculating in model computing unit 3 model computing unit 3 by the hardware structure of Implementation of pseudocode below in carry out, it is corresponding to above-mentioned calculation criterion.As can be seen from false code, carry out calculating in Inner eycle and outer circulation and its partial results is accumulated.When starting model and calculating, the representative value for counter Initial Parameters Nstart is 0.
/ * the stage 1: input standardization */
/ * the stage 2: the calculating * of outer circulation/
/ * stage 2a: the calculating * of Inner eycle/
/ * stage 2b: gauge index function */
/ * stage 2c:*/
/ * the stage 3: outputting standard */
Therefore comprise hyper parameter and gridden data for calculating based on the model data needed for the function model of data, it is stored distributing in the storage area of the involved function model based on data in the memory unit.According to above-mentioned false code, the parameter based on the function model of data comprises normalizing parameter
, vectorial Q
y, the quantity N at Grid dimension strong point, the quantity D of dimension of input parameter, outer circulation initial value nStart, when the calculating restarting Inner eycle loop index
the length dimension l of (usually=0) and each dimension for input parameter.
Can arrange the function model based on data, it is respectively based on the parameter group of oneself of above-mentioned parameter and the common matrix of gridden data.In the current situation, the matrix of gridden data comprises the following measurement point of test status measurement or test macro: create multiple function model based on data based on described measurement point.Also namely specify: only provide the matrix of a gridden data for multiple function model based on data and leave in the storage area for this reason arranged of storage unit.
The weighting at each Grid dimension strong point of gridden data is by hyper parameter Q
yrealize, its dimension is corresponding to the quantity of N, also i.e. gridden data.But hyper parameter Q
ybe set up individually for each function model based on data, because be indirectly stored in wherein for the measured value of the output parameter of institute's modeling.
Below, the process flow diagram by Fig. 2 describes how to carry out computing function value for two function models based on data.
First is provided in step sl respectively based on the first parameter group P of the function model of data in the first parameter storage area
1, for second based on the second parameter group P of the function model of data
2, and for first and second based on the net point storage area of the gridden data of the function model of data.
In step s 2, from the first parameter storage area, the first parameter group P is called based on the function model of data in order to calculating first
1and call gridden data from net point storage area, so that computing function value.
Similarly, from the second parameter storage area, the second parameter group P is called based on the function model of data in order to calculating second in step s3
1and call gridden data from net point storage area, so that computing function value.
For the function models based on data many arbitrarily, can repeat described strategy, wherein these function models are at least in part based on the gridden data stored in net point storage area.
In order to create gridden data for the function model based on data and parameter group, now based on have test plan given in advance measurement series sight in measure two output parameter A and B based on identical training data point and in addition for each the training Gaussian process model in these output parameters A, B, the matrix making gridden data for two models is identical.
If Grid dimension strong point is deleted during the model training about one of output parameter, then this point is removed usually from the matrix of gridden data.Therefore this matrix can have dimension (N-1) xD and affiliated hyper parameter
(k=1,2 ...: and the ordinal number of parameter group) there is dimension (N-1) x1.
But because the matrix of gridden data is also used to second based on the function model of data, the Grid dimension strong point of therefore only deleting based on the function model of data for first can not be removed simply from gridden data set.But must guarantee, the Grid dimension strong point being arranged in gridden data set does not have impact to first based on the calculating of the function model of data.
Because vectorial
used in model computing unit 3, its mode is multiplied with the functional value at i-th Grid dimension strong point by every i-th item, therefore can get rid of Grid dimension strong point from this calculating, and its mode will (in this case the second function model) vector
corresponding i-th be set to zero.I-th training data point no longer summation is had an impact in this kind of situation and thus no longer to model prediction, to be also (first) have an impact based on the functional value of the function model of data.Therefore possible that, by vector that is affiliated, also kth model when there is deviation between two gridden datas based on the function model of data
in item be set to zero, distribute to this net point to get rid of from the calculating of the function model based on data.
If determined, relate to the Grid dimension strong point determined based on neither one needs in the function model of data of common gridden data set, so can delete involved Grid dimension strong point from the set of gridden data.Then, also can from the vector of the function model based on data
middle deletion be accordingly 0 item and correspondingly adapted mesh to count the quantity at strong point.
The deletion at Grid dimension strong point can realize as described above, and adds other Grid dimension strong point also can to the set of gridden data.In order to realize the other use of the matrix to common grid Grid data, first by capable for the matrix-expand N+1 of gridden data, it comprises new training data point.In addition, accordingly based on the vector of the function model of data
be expanded N+1 item, first it be set to 0.Then, ask for during the training stage utilizing the matrix that is made up of gridden data and affiliated impact point for first based on the vector of the function model of data
.
Especially can specify, the Grid dimension strong point only used by a part for the function model based on data is attached to the end of the matrix of the gridden data jointly used by multiple function model.In this case, calculating based on the function model of data can be set to, make to interrupt described calculating after the Grid dimension strong point reaching quantification, make to be arranged for the other available grids point data describing the other function model based on data and keep not being considered.Alternatively, vector can be used
the flexible program presented hereinbefore of zero setting of each additional weight, such as other Grid dimension strong points can be added to one based on the model of data after a while, and do not improve vector
dimension.
Based on the similarity asking for formula and RBF net and support vector machine of Gaussian process, also set forth methodology can be used for such model based on data.
Claims (10)
1. for calculating the method for multiple function model based on data, described function model is defined by parameter group and gridden data, wherein gridden data comprises Grid dimension strong point identical at least in part, wherein the identical Grid dimension strong point for multiple function model based on data once being stored in storage area, wherein repeatedly accessing described storage area to calculate described multiple function model based on data.
2. method according to claim 1, wherein produces in checking system for the parameter group of described multiple function model based on data and gridden data from common testing process.
3. method according to claim 1 and 2, wherein, described parameter group has vector respectively, this DUAL PROBLEMS OF VECTOR MAPPING is accordingly based on the functional value of the function model of data, wherein this vector has N dimension, wherein N is corresponding to the quantity of the net point of gridden data, and wherein when the element distributing to involved Grid dimension strong point of this vector is set to 0, Grid dimension strong point maintenance in one of described multiple function model based on data of gridden data is not considered.
4. according to the method one of claims 1 to 3 Suo Shu, wherein revise one of the function model based on data by additional Grid dimension strong point, its mode is, Grid dimension strong point is added and the respective element of the vector of parameter group is set to 0 to described gridden data, a wherein said vector based on the function model of data utilizes the gridden data re-training being expanded Grid dimension strong point, to ask for a described respective element based on the vector of the parameter group of the function model of data.
5., according to the method one of Claims 1-4 Suo Shu, wherein said multiple function model based on data is corresponding to Gaussian process model, RBF net or support vector machine.
6. the model computing unit (3) for calculating multiple function model based on data in integrating control assembly (1), described function model is defined by parameter group and gridden data, comprising:
-for the gridden data for multiple function model based on data being stored in the internal storage (5) in common storage area,
Wherein model computing unit (3) is configured to repeatedly territory, access storage areas to calculate the function model based on data.
7. with the integrating control assembly of main computation unit (2) and model computing unit according to claim 6 (3).
8. computer program, it is set up for implementing according to the method one of claim 1 to 5 Suo Shu.
9. electronic storage medium, computer program according to claim 8 is stored on described electronic storage medium.
10. electronic control unit, it has electronic storage medium according to claim 9.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102013220425.6 | 2013-10-10 | ||
| DE201310220425 DE102013220425A1 (en) | 2013-10-10 | 2013-10-10 | Method and device for operating an integrated control module for calculating data-based function models |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104570757A true CN104570757A (en) | 2015-04-29 |
| CN104570757B CN104570757B (en) | 2020-03-17 |
Family
ID=52737865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410527571.0A Active CN104570757B (en) | 2013-10-10 | 2014-10-09 | Method and device for operating an integrated control assembly |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN104570757B (en) |
| DE (1) | DE102013220425A1 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102018206660A1 (en) * | 2018-04-30 | 2019-10-31 | Sms Group Gmbh | Method for operating a cooling section and installation for producing rolled products |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0559949A2 (en) * | 1992-03-07 | 1993-09-15 | Bundesrepublik Deutschland, vertr. durch Vorstand der D. Bundespost TELEKOM, vertr. durch Präs. des Fernmeldetechn. Z.amtes | Method for determining the optical location and power levels of emitters in picocellular radio networks |
| US6195659B1 (en) * | 1998-07-14 | 2001-02-27 | Trw Inc. | Method and apparatus for morphological clustering having multiple dilation and erosion of switchable grid data cells |
| CN1894930A (en) * | 2003-12-15 | 2007-01-10 | 国际商业机器公司 | Apparatus, system, and method for grid based data storage |
| CN101021954A (en) * | 2007-03-23 | 2007-08-22 | 东南大学 | Three-dimensional scanning point cloud compressing method |
| CN101241602A (en) * | 2008-03-05 | 2008-08-13 | 罗笑南 | Computer three-dimensional grid pattern abridging method |
| CN101625660A (en) * | 2009-08-04 | 2010-01-13 | 威盛电子股份有限公司 | Data storage device and data storage method |
| CN102096106A (en) * | 2010-12-10 | 2011-06-15 | 中国石油集团川庆钻探工程有限公司 | Method for drawing reverse fault-contained isopleths based on space gridding algorithm |
| CN102411616A (en) * | 2011-10-28 | 2012-04-11 | 北京地拓科技发展有限公司 | Method and system for storing data and data management method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102010028266A1 (en) | 2010-04-27 | 2011-10-27 | Robert Bosch Gmbh | Control device and method for calculating an output for a controller |
-
2013
- 2013-10-10 DE DE201310220425 patent/DE102013220425A1/en active Pending
-
2014
- 2014-10-09 CN CN201410527571.0A patent/CN104570757B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0559949A2 (en) * | 1992-03-07 | 1993-09-15 | Bundesrepublik Deutschland, vertr. durch Vorstand der D. Bundespost TELEKOM, vertr. durch Präs. des Fernmeldetechn. Z.amtes | Method for determining the optical location and power levels of emitters in picocellular radio networks |
| US6195659B1 (en) * | 1998-07-14 | 2001-02-27 | Trw Inc. | Method and apparatus for morphological clustering having multiple dilation and erosion of switchable grid data cells |
| CN1894930A (en) * | 2003-12-15 | 2007-01-10 | 国际商业机器公司 | Apparatus, system, and method for grid based data storage |
| CN101021954A (en) * | 2007-03-23 | 2007-08-22 | 东南大学 | Three-dimensional scanning point cloud compressing method |
| CN101241602A (en) * | 2008-03-05 | 2008-08-13 | 罗笑南 | Computer three-dimensional grid pattern abridging method |
| CN101625660A (en) * | 2009-08-04 | 2010-01-13 | 威盛电子股份有限公司 | Data storage device and data storage method |
| CN102096106A (en) * | 2010-12-10 | 2011-06-15 | 中国石油集团川庆钻探工程有限公司 | Method for drawing reverse fault-contained isopleths based on space gridding algorithm |
| CN102411616A (en) * | 2011-10-28 | 2012-04-11 | 北京地拓科技发展有限公司 | Method and system for storing data and data management method |
Non-Patent Citations (2)
| Title |
|---|
| SANDRO FIORE 等: "Data Virtualization in Grid Environments through the GRelC Data Access and Integration Service", 《2009 BY THE INSTITUTE OF ELECTRICAL AND ELECTRONICS ENGINEERS》 * |
| 尤晓刚 等: "网格环境下的数据存储管理", 《信息技术》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104570757B (en) | 2020-03-17 |
| DE102013220425A1 (en) | 2015-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7413580B2 (en) | Generating integrated circuit floorplans using neural networks | |
| CN102859158B (en) | For calculating control apparatus and the method for the output parameter controlled | |
| JP6784780B2 (en) | How to build a probabilistic model for large-scale renewable energy data | |
| CN108984211A (en) | Neural network processor and the method for executing Givens rotation instruction using processor | |
| CN104573822B (en) | For integrated control component for calculating the model computing unit of LOLIMOT | |
| CN102314522B (en) | Optimizing method of analogue integrated circuit design | |
| JP2007249974A (en) | Method for characterizing processor design, device, logical program and system | |
| CN111126668A (en) | Spark operation time prediction method and device based on graph convolution network | |
| CN104102137A (en) | Model calculation unit, control unit and method for calibrating a data-based function model | |
| JP2014529795A (en) | How to create control unit functions | |
| US20160239278A1 (en) | Generating a schedule of instructions based on a processor memory tree | |
| CN111814376A (en) | Method for extracting rigidity result of vehicle body attachment point and electronic equipment | |
| US20150186332A1 (en) | Method and device for providing a sparse gaussian process model for calculation in an engine control unit | |
| CN110018882B (en) | Virtual machine performance prediction method based on width learning | |
| CN104102129A (en) | Method and device for creating a data-based function model | |
| CN104111871A (en) | Method and device used for executing dynamic load balancing in circuit simulation | |
| US9977417B2 (en) | Method for optimizing utilization of programmable logic elements in control units for vehicles | |
| US20200117841A1 (en) | Method and apparatus for analyzing fluid around tire | |
| CN104570757A (en) | Method and apparatus for operating integrated control assembly | |
| CN112182819B (en) | Structure topology optimization method and system based on weighted graph and readable storage medium | |
| CN116757146A (en) | Distributed random walk parasitic capacitance extraction method, device, equipment and medium | |
| CN116626504A (en) | Power battery performance determining method, apparatus, computer device and storage medium | |
| CN117272896A (en) | Machine learning techniques for circuit design verification | |
| CN107609234B (en) | Robustness analysis method and system based on uncertainty | |
| KR102257530B1 (en) | Method and device for determining a gradient of a data-based function model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |