CN105389615A - Nested dynamic environment change detection method - Google Patents
Nested dynamic environment change detection method Download PDFInfo
- Publication number
- CN105389615A CN105389615A CN201510915881.4A CN201510915881A CN105389615A CN 105389615 A CN105389615 A CN 105389615A CN 201510915881 A CN201510915881 A CN 201510915881A CN 105389615 A CN105389615 A CN 105389615A
- Authority
- CN
- China
- Prior art keywords
- environment
- change
- environmental
- fitness
- forward step
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5017—Task decomposition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Genetics & Genomics (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Complex Calculations (AREA)
Abstract
The present invention discloses a nested dynamic environment change detection method. The method comprises: extracting characteristic parameters related to a current division task set; setting a sampling number as M; determining whether the current environment division scale is equal to the previous environment division scale or not; evaluating the fitness of M sampling individuals in the current environment, determining whether the fitness is the same with that under the previous environment evaluation or not, so as to determine whether a fitness function of the current environment is equal to that in the previous environment or not; successively determining whether each component in a performance constraint set of the current environment is equal to each corresponding component in a performance constraint set of the previous environment or not; successively determining whether values at M corresponding positions of an adjacency matrix in the current environment are equal to those at M corresponding positions of an adjacency matrix in the previous environment; successively determining whether values at M corresponding positions in node attributes of the current environment are equal to those at M corresponding positions in node attributes of the previous environment; and ending the detection. The nested dynamic environment change detection method disclosed by the present invention enables an algorithm to be able to predict changes of dynamic problems more accurately.
Description
Technical field
The present invention relates to a kind of dynamic environment change detecting method.Particularly relate to a kind of dynamic environment change detecting method of nested type.
Background technology
Dynamic hardware-software partition problem refers to when solving hardware-software partition problem, system faced by task-set to be divided be uncertain, optimized algorithm must the structure of Detection task collection could be determined to adopt what strategy to carry out dividing to adapt to up-to-date environment.In dynamic environment, focus on very much the real-time of partitioning algorithm, require that partitioning algorithm has the ability processing dynamic hardware-software partition.Therefore, partitioning algorithm needs for the characteristic of dynamic environment, design adaptive algorithm, to meet the requirement of real system to performance and reliability.
The mathematical model of dynamic hardware-software partition: the model of hardware-software partition problem uses a Flow chart task (TaskGraph) to describe usually, whole Flow chart task is a directed acyclic graph (DAG figure), can with two element group representations a: G=(V, E).Wherein V={V
0, V
1..., V
nthe set of TU task unit in whole system, V
irepresent i-th task node, E={ (V
i, V
j) | V
i, V
j∈ V} represents data dependence relation between two task nodes or Row control, e
i=(V
i, V
j) represent data dependence relation between i-th node and a jth node or Row control.V
ibe called V
jpredecessor node, V
jbe called V
idescendant node, both create unidirectional dependence.Fig. 1 is the hardware-software partition system model figure that a DAG figure represents.
The attributive character that each node is detailed describes:
N
i={T
sw,T
hw,A
sw,A
hwC
sw,C
hw,P
sw,P
hw,S
sw,S
hw,…,x(i)}
Wherein, T
swand T
hwrepresent the task execution time that task node realizes with software and hardware respectively, A
swand A
hwhardware area required when representing that task node software and hardware realizes respectively, C
swand C
hwrepresent cost when task node software and hardware realizes respectively, P
swand P
hwrepresent power consumption when task node software and hardware realizes respectively, S
swand S
hwrepresent storage overhead when task node software and hardware realizes respectively.X (i) represent this node the mapped mode selected, the software and hardware implementation selected by expression task.Limit collection E is the category that task scheduling needs to consider, describing and strengthening the specific aim to partition problem, not introducing the gain to reality total execution time that task scheduling is brought in partition problem to simplify system.Therefore to the value of scheduling parameter E be thought of as ideally 0.According to the above-mentioned definition to system task attribute, the fitness function of hardware-software partition problem and constraint function are defined as:
In dynamic hardware-software partition problem, following mathematical model can be described as.
Process the optimization problem of dynamic hardware-software partition, be first a very crucial step to the accurate detection of environmental change, this is evolution algorithm makes correlated response in time prerequisite to environmental change.The quality of detection method directly can have influence on time efficiency and the system reliability of partitioning algorithm entirety.But, in existing most of scientific research, be all suppose that algorithm is own through knowing the situation of change of environment in advance, the situation of the correct time point of such as environmental change and task-set change substantially, but actual conditions are really not so.Such as, in the problem of dynamic dispatching, scheduler task all likely changes at any time.So, problem has just been come.How to detect efficiently and accurately external environment condition is the top priority that dynamic evolution calculates research.
Evolution algorithm, as the blind optimized algorithm of one, does not namely directly use the variable of objective function and computing formula to guide searching process, but objective function is considered as black box, by finding optimum solution to the iterative test of objective function input and output.Because the computation process of objective function is sightless for blind optimized algorithm, therefore this optimization is called blind optimization.Under normal circumstances, evolution algorithm is can not the change of sensing external environment, and need by some external householder methods, help the change of evolution algorithm testing environment, this detection is very important.If detection algorithm complexity is high, implement difficulty, valuable computational resource can be taken, be unfavorable for the division of task; If detection algorithm accuracy rate is low, usually occur the erroneous judgement to environmental change, this can badly influence the reliability of system.If the reaction velocity of detection algorithm is very slow, after actual environment change for a long time after just can be detected, this can waste a part of computational resource on idle work, and system reliability also can be made to reduce.Therefore, good detection method can detect the change of environment fast, vacates the more time and prepares to carry out the division of task, improve the change that the efficiency divided can predict environment exactly for partitioning algorithm, the reliability of raising system, avoids expending valuable computational resource.Finally, the complexity that good detection algorithm implements is low, and the computational resource expended is few.
Conventional dynamic environment detection method has:
1, sampling Evaluation Method.
In the search volume of feasible zone, the several point of stochastic sampling, records these sampled points and carries out Fitness analysis, if as long as wherein there is the fitness value of body one by one to there occurs change, so just think that environment there occurs change.The performance of this method depends on the quantity of sampled point.If sampled point is too much, sampling and assessment all may take valuable computational resource.If sampled point is very few, the solution space at sampled point place does not change, then also cannot detect the change of environment, affects the accuracy rate of the judgement of algorithm.
2, optimum evaluation method.
Before the next generation evolves generation, individual optimal solution best is at present reappraised, if fitness value there occurs change at the moment, so just thinks that environment there occurs change.This method can detect the dynamic environment that optimum solution position changes and obtain good Detection results, but is invalid to the environmental change that optimum solution position does not change.In most of the cases, the change of environment can bring the change of optimum solution position.But in some cases, single optimum solution sampled point cannot judge the change of environment exactly.
3, statistical estimation method.
The degeneration of population performance (deterioration), or the indicator whether reduction of best individual average behavior changes as environment, usual average fitness and fitness variance are as the criterion of population performance degradation.This method is that the change of assumptions' environment causes original performance of separating to reduce as prerequisite.In general, the change of environment can cause the degeneration of population performance, so it is adapted to majority of case.But in some cases, the performance of population may improve, and this method just can not be suitable for.
In a word, if the sampled point detected is more, then certainly will take computational resource limited, valuable in a large number, the cost of flower in the detection of environmental change is higher, is unfavorable for the tracking of algorithm to the optimum solution of change.Consider that in real world, the change of environment is that certain interval of time just occurs, some algorithms have employed the method that interval some generations carry out environment measuring again.
But these detection methods all exist one-sidedness, for special circumstances to making correct response, can not can only be suitable for majority of case.For the system that reliability requirement is very high, this will cause catastrophic effect.
Summary of the invention
Technical matters to be solved by this invention is, provides a kind of dynamic environment change detecting method of nested type.
The technical solution adopted in the present invention is: a kind of dynamic environment change detecting method of nested type, comprises the steps:
1) extract the characteristic parameter relevant with current division task-set, environmental change zone bit flag=0 is set, and regulation is when environmental change zone bit flag is 1, represent that environment changes; When environmental change zone bit flag is 0, represent that environment does not change;
2) arranging number of samples is M;
3) judge that current environment divides scale n
curenvironment before divides scale n
prewhether equal;
4) assess the individual fitness of M sampling under the present circumstances, judge that whether described fitness is identical with the fitness under environmental assessment before, judge current environment fitness function f with this
curwith last environmental adaptation degree function f
prewhether equal;
5) current environment performance constraints group L is judged successively
curin each component and environmental performance set of constraints L before
prein each corresponding component whether equal;
6) current environment adjacency matrix E is judged successively
curenvironment adjacent matrix E before
prein M correspondence position on value whether equal;
7) current environment nodal community Node is judged successively
curenvironment nodes attribute Node before
prein M correspondence position on value whether equal;
8) detection of end.
Step 1) described in characteristic parameter comprise: node number n, fitness function f, adjacency matrix E
n × n, performance constraints vector L
n × 1, nodal community matrix N ode
n × 2.
Step 3) if in unequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise, forward step 4 to).
Step 4) if in fitness under the individual fitness of M sampling and environmental assessment before entirely inequal, environmental change zone bit flag is put 1, and expression environment changes, and forwards step 8 to); Otherwise, forward step 5 to).
Step 5) if in all inequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise, forward step 6 to).
Step 6) if in all inequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise, forward step 7 to).
Step 7) if in all inequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise directly forward step 8 to).
The dynamic environment change detecting method of a kind of nested type of the present invention, when not affecting environment measuring accuracy, is optimized the structure of detection method.For the most cases of environmental change, further optimization is carried out to algorithm in the basis increasing complexity not significantly.By the sample detecting of the key parameter to task-set to be divided, determine the change of dynamic environment, compare and detect by fitness the accuracy rate that improve detection method significantly, make algorithm can predict the change of dynamic problem more exactly, for hardware/software partitioning algorithms afterwards provides division information exactly.
Accompanying drawing explanation
Fig. 1 is the hardware-software partition system model figure that a DAG figure represents;
Fig. 2 is the process flow diagram of the dynamic environment change detecting method of nested type of the present invention.
Embodiment
Below in conjunction with embodiment and accompanying drawing, the dynamic environment change detecting method to a kind of nested type of the present invention is described in detail.
As shown in Figure 2, the dynamic environment change detecting method of a kind of nested type of the present invention, is characterized in that, comprise the steps:
1) extract the characteristic parameter relevant with current division task-set, environmental change zone bit flag=0 is set, and regulation is when environmental change zone bit flag is 1, represent that environment changes; When environmental change zone bit flag is 0, represent that environment does not change;
Described characteristic parameter comprises: node number n, fitness function f, adjacency matrix E
n × n, performance constraints vector L
n × 1, nodal community matrix N ode
n × 2.
2) arranging number of samples is M;
3) judge that current environment divides scale n
curenvironment before divides scale n
prewhether equal;
If unequal, environmental change zone bit flag is put 1, represent that environment changes, forward step 8 to); Otherwise, forward step 4 to).
4) assess the individual fitness of M sampling under the present circumstances, judge that whether described fitness is identical with the fitness under environmental assessment before, judge current environment fitness function f with this
curwith last environmental adaptation degree function f
prewhether equal;
If the fitness under M sample individual fitness and environmental assessment is before entirely inequal, environmental change zone bit flag is put 1, represent that environment changes, forward step 8 to); Otherwise, forward step 5 to).
5) current environment performance constraints group L is judged successively
curin each component and environmental performance set of constraints L before
prein each corresponding component whether equal;
If all inequal, environmental change zone bit flag is put 1, represent that environment changes, forward step 8 to); Otherwise, forward step 6 to).
6) current environment adjacency matrix E is judged successively
curenvironment adjacent matrix E before
prein M correspondence position on value whether equal;
If all inequal, environmental change zone bit flag is put 1, represent that environment changes, forward step 8 to); Otherwise, forward step 7 to).
7) current environment nodal community Node is judged successively
curenvironment nodes attribute Node before
prein M correspondence position on value whether equal;
If all inequal, environmental change zone bit flag is put 1, represent that environment changes, forward step 8 to); Otherwise directly forward step 8 to).
8) detection of end.
(1) produce 8 static hardware-software partition environment with TGFF instrument, computer data is packaged into enviroline, and is numbered from 1 to 8.The parameter configuration of TGFF instrument is as shown in table 1.
The optimum configurations of table 1TGFF instrument
Nodal point number | Software is consuming time | Hardware is consuming time | Hardware area | Communicate consuming time | Area-constrained | |
1 | 15 | 423±34 | 107±22 | 55±18 | 41±9 | 450 |
2 | 30 | 273±30 | 83±16 | 76±27 | 38±8 | 1150 |
3 | 45 | 119±26 | 57±11 | 98±38 | 33±7 | 2200 |
4 | 60 | 443±62 | 153±29 | 71±26 | 60±15 | 1900 |
5 | 75 | 774±103 | 252±47 | 43±12 | 89±23 | 1500 |
6 | 90 | 513±66 | 179±32 | 58±16 | 62±16 | 3000 |
7 | 105 | 258±33 | 110±17 | 73±20 | 35±10 | 4500 |
8 | 115 | 88±11 | 64±8 | 87±24 | 15±5 | 5500 |
(2) employing number M=10 is set, environmental change zone bit flag=0.
(3) successively the DAG figure being numbered 1 ~ 8 in table 1 is called in current division environment.
(4) the dynamic environment change detecting method of nested type of the present invention is utilized to judge the characteristic parameter of environment respectively: node number n, fitness function f, adjacency matrix E
n × n, performance constraints vector L
n × 1, nodal community matrix N ode
n × 2whether change.If detect that environment changes and environment there occurs change really.
Claims (7)
1. a dynamic environment change detecting method for nested type, is characterized in that, comprise the steps:
1) extract the characteristic parameter relevant with current division task-set, environmental change zone bit flag=0 is set, and regulation is when environmental change zone bit flag is 1, represent that environment changes; When environmental change zone bit flag is 0, represent that environment does not change;
2) arranging number of samples is M;
3) judge that current environment divides scale n
curenvironment before divides scale n
prewhether equal;
4) assess the individual fitness of M sampling under the present circumstances, judge that whether described fitness is identical with the fitness under environmental assessment before, judge current environment fitness function f with this
curwith last environmental adaptation degree function f
prewhether equal;
5) current environment performance constraints group L is judged successively
curin each component and environmental performance set of constraints L before
prein each corresponding component whether equal;
6) current environment adjacency matrix E is judged successively
curenvironment adjacent matrix E before
prein M correspondence position on value whether equal;
7) current environment nodal community Node is judged successively
curenvironment nodes attribute Node before
prein M correspondence position on value whether equal;
8) detection of end.
2. the dynamic environment change detecting method of a kind of nested type according to claim 1, is characterized in that, step 1) described in characteristic parameter comprise: node number n, fitness function f, adjacency matrix E
n × n, performance constraints vector L
n × 1, nodal community matrix N ode
n × 2.
3. the dynamic environment change detecting method of a kind of nested type according to claim 1, is characterized in that, step 3) if in unequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise, forward step 4 to).
4. the dynamic environment change detecting method of a kind of nested type according to claim 1, it is characterized in that, step 4) if in fitness under the individual fitness of M sampling and environmental assessment before entirely inequal, environmental change zone bit flag is put 1, represent that environment changes, forward step 8 to); Otherwise, forward step 5 to).
5. the dynamic environment change detecting method of a kind of nested type according to claim 1, is characterized in that, step 5) if in all inequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise, forward step 6 to).
6. the dynamic environment change detecting method of a kind of nested type according to claim 1, is characterized in that, step 6) if in all inequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise, forward step 7 to).
7. the dynamic environment change detecting method of a kind of nested type according to claim 1, is characterized in that, step 7) if in all inequal, environmental change zone bit flag is put 1, represent environment change, forward step 8 to); Otherwise directly forward step 8 to).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510915881.4A CN105389615B (en) | 2015-12-09 | 2015-12-09 | A kind of dynamic hardware-software partition environmental change detection method of nested type |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510915881.4A CN105389615B (en) | 2015-12-09 | 2015-12-09 | A kind of dynamic hardware-software partition environmental change detection method of nested type |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105389615A true CN105389615A (en) | 2016-03-09 |
CN105389615B CN105389615B (en) | 2018-01-09 |
Family
ID=55421882
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510915881.4A Active CN105389615B (en) | 2015-12-09 | 2015-12-09 | A kind of dynamic hardware-software partition environmental change detection method of nested type |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105389615B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070162475A1 (en) * | 2005-12-30 | 2007-07-12 | Intel Corporation | Method and apparatus for hardware-based dynamic escape detection in managed run-time environments |
CN101515933A (en) * | 2009-03-16 | 2009-08-26 | 中兴通讯股份有限公司 | Method and system for detecting the completeness of network equipment software and hardware |
CN101763288A (en) * | 2010-01-19 | 2010-06-30 | 湖南大学 | Method for dynamic hardware and software partitioning by considering hardware pre-configuration factors |
CN102508721A (en) * | 2011-11-30 | 2012-06-20 | 湖南大学 | Hardware-software partitioning method based on greedy simulated annealing algorithm |
CN104572268A (en) * | 2015-01-14 | 2015-04-29 | 天津大学 | Efficient dynamic division method of software and hardware |
-
2015
- 2015-12-09 CN CN201510915881.4A patent/CN105389615B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070162475A1 (en) * | 2005-12-30 | 2007-07-12 | Intel Corporation | Method and apparatus for hardware-based dynamic escape detection in managed run-time environments |
CN101515933A (en) * | 2009-03-16 | 2009-08-26 | 中兴通讯股份有限公司 | Method and system for detecting the completeness of network equipment software and hardware |
CN101763288A (en) * | 2010-01-19 | 2010-06-30 | 湖南大学 | Method for dynamic hardware and software partitioning by considering hardware pre-configuration factors |
CN102508721A (en) * | 2011-11-30 | 2012-06-20 | 湖南大学 | Hardware-software partitioning method based on greedy simulated annealing algorithm |
CN104572268A (en) * | 2015-01-14 | 2015-04-29 | 天津大学 | Efficient dynamic division method of software and hardware |
Non-Patent Citations (2)
Title |
---|
王光辉等: "多种群协同粒子群算法求解动态环境优化问题", 《第27届中国控制会议》 * |
陈莉: "动态优化问题的粒子群算法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Also Published As
Publication number | Publication date |
---|---|
CN105389615B (en) | 2018-01-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6969637B2 (en) | Causality analysis methods and electronic devices | |
Charras-Garrido et al. | Extreme value analysis: an introduction | |
Nasa et al. | Evaluation of different classification techniques for web data | |
CN111428733B (en) | Zero sample target detection method and system based on semantic feature space conversion | |
CN106030589A (en) | Disease prediction system using open source data | |
CN113218537B (en) | Training method, training device, training equipment and training storage medium for temperature anomaly detection model | |
Zhou et al. | Imbalanced data processing model for software defect prediction | |
CN113011191A (en) | Knowledge joint extraction model training method | |
US10248462B2 (en) | Management server which constructs a request load model for an object system, load estimation method thereof and storage medium for storing program | |
CN113783715B (en) | Opportunistic network topology prediction method adopting causal convolutional neural network | |
US20220327394A1 (en) | Learning support apparatus, learning support methods, and computer-readable recording medium | |
CN107463486B (en) | System performance analysis method and device and server | |
CN112163132B (en) | Data labeling method and device, storage medium and electronic equipment | |
CN117688846A (en) | Reinforced learning prediction method and system for building energy consumption and storage medium | |
CN113657510A (en) | Method and device for determining data sample with marked value | |
CN110532458B (en) | Method and device for determining search mode, server and storage medium | |
CN117036781A (en) | Image classification method based on tree comprehensive diversity depth forests | |
CN116721736A (en) | Exercise effect monitoring method, system, equipment and readable storage medium | |
CN108133234B (en) | Sparse subset selection algorithm-based community detection method, device and equipment | |
US11328225B1 (en) | Automatic spatial regression system | |
US20230186150A1 (en) | Hyperparameter selection using budget-aware bayesian optimization | |
CN105389615A (en) | Nested dynamic environment change detection method | |
US20230022253A1 (en) | Fast and accurate prediction methods and systems based on analytical models | |
CN117999560A (en) | Hardware-aware progressive training of machine learning models | |
CN114239743B (en) | Weather event occurrence time prediction method based on sparse time sequence data |
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 |