CN110443348A - A kind of modeling of the nerve network system based on MSVL and verification method - Google Patents
A kind of modeling of the nerve network system based on MSVL and verification method Download PDFInfo
- Publication number
- CN110443348A CN110443348A CN201910563393.XA CN201910563393A CN110443348A CN 110443348 A CN110443348 A CN 110443348A CN 201910563393 A CN201910563393 A CN 201910563393A CN 110443348 A CN110443348 A CN 110443348A
- Authority
- CN
- China
- Prior art keywords
- network system
- nerve network
- msvl
- modeling
- nerve
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3604—Software analysis for verifying properties of programs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Quality & Reliability (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Stored Programmes (AREA)
Abstract
The invention belongs to system form modeling and verification technique field, modeling and the verification method of a kind of nerve network system based on MSVL are disclosed.Neural network (including DNN, the CNN, RNN etc.) system verified is modeled using MSVL, the information of involved node and side in system is indicated with Multidimensional numerical, in modeling, with the basic operation in function representation neural network system;It determines the shared nature and characteristic of nerve network system for needing to verify, mainly includes correctness and robustness, and describe these properties of nerve network system using PPTL formula;The PPTL formula of the MSVL program of modeling and the shared property of description is unified in UMC4MSVL platform and is verified, judges whether property can be met according to verification result.The programming procedure of formalization is applied to the modeling and verifying of nerve network system by the present invention, each state that program executes can be verified reliably, has effectively ensured the safety of system itself.
Description
Technical field
The invention belongs to system form modelings and verification technique field more particularly to a kind of neural network based on MSVL
The modeling of system and verification method.
Background technique
Currently, nerve network system still remains problems: first, since depth model is all non-convex function, also
The theoretical research of neural network in this respect is allowed to become extremely difficult.For any one nonlinear function, one can be found
Shallow-layer network and depth network indicate that depth model has better expressive ability to nonlinear function than shallow Model, but
The representability of depth network does not represent learnability and interpretation.Second, nerve network system whether be securely and reliably
Restrict the powerful resistance of industry development.There may be serious defect and loopholes for nerve network system itself, it is also possible to by not
Congener malicious attack.Especially in the scene that safety concerns, superfine small mistake all may cause serious consequence.By
This has become hot spot to the attack of neural network and the research of mean of defense.
In academia, numerous achievements are worth further investigation.The prior art one is by that can not examine data sample using certain
The small disturbance felt may cause to the result that neural network classification system generates mistake.The prior art two passes through research nerve
The linear characteristic of network model proposes a kind of simple and quick method FGSM generated to resisting sample, and method is extensive
It to different neural network structures, then applies this method in specific physics scene again, classifying quality receives pair
Resisting sample significantly interferes with.For the prior art three on the basis of FGSM algorithm, the thought for introducing norm has further quantified nerve
The robustness evaluation index of network class model, so as to reach attack on the basis of modifying original sample as small as possible
The purpose of neural network classification model.The prior art four is by Jacobian matrix and significant nomography, by will be in given image
Several pixels be saturated to maximum or minimum value, the iteration above process and finally to be classified into mistake to resisting sample
Target category.
It is extremely in actual production environment that the successful application of above-mentioned four kinds of attack technologies, which shows nerve network system,
Fragile, the safety of neural network system not can guarantee, or specially correctness and robustness, concern safely certain
Fatal problem will likely be caused under scene.The correctness and robustness for examining nerve network system as a result, go deep into neural network
The Formal Theory research of model seems very necessary.For nerve network system, the inspection of safety mainly includes to survey
Two kinds of means of examination and verifying.Test is to meet the example being centainly distributed input by running, and checking system whether there is error row
For.And verifying is then a kind of method of formalization, whether meets specificity to system using technologies such as model inspection, theorem provings
Matter carries out stringent reasoning.
Problem of the existing technology is:
Most nerve network systems carry out safety examination using the method for test.But test can not be demonstrate,proved
Mistake is not present in bright system, and it is obviously less reliable only by the system of test under some scenes to concern safely.
Solve the difficulty of above-mentioned technical problem:
Nerve network system usually may be face identification system, automated driving system, recommender system, translation system etc.,
But code, complicated realization logic and the potential security features of large-scale data, the big scale of construction will allow and model and test
Card process becomes difficult.
Solve the meaning of above-mentioned technical problem:
For the automatic Verification of nerve network system, often the cycle of operation is long for frequent large-scale test,
It is more to consume resource, test scene is limited, and can not obtain error feedback information comprehensively in time, and by the way of Formal Verification
It can more efficiently guarantee the safety of system on the basis of preferably solving the above problems.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of modeling of nerve network system based on MSVL and
Verification method.
The invention is realized in this way modeling and the verification method of a kind of nerve network system based on MSVL, the base
In the nerve network system of MSVL modeling and verification method the following steps are included:
Step 1 models the nerve network system verified with MSVL, indicates system with Multidimensional numerical
In involved node and side information;In modeling, with the basic operation in function representation neural network system;
Step 2 determines the correctness and robustness feature for needing the nerve network system verified, and uses PPTL formula
These properties are described;
The PPTL formula of the MSVL program of modeling and the shared property of description is unified in UMC4MSVL platform and tests by step 3
Card, judges whether property can be met according to verification result.
Further, nerve network system is modeled with MSVL, wherein described with Multidimensional numerical system interior joint and
The information on side, and the basic operation of neural network is then indicated with function.
Further, PPTL formula is determined according to the shared property of the nerve network system specifically to be verified, first basis
The property definition proposition to be verified constructs reasonable and strict PPTL formula then according to the logical relation of proposition.
Further, the modeling method of the nerve network system based on MSVL, specifically includes:
(1) nerve network system structure is defined with Multidimensional numerical.
(2) data sample function reading read_data is defined.
(3) definition structure parameter function reading read_parameter.
(4) propagated forward analog function bpnn_sim is defined.
(5) attack process analog function bpnn_sim_single is defined.
Further, it is described as in the shared property of the nerve network system about correctness, if test is concentrated with
Sample more than 90% (experiment setting value) can correctly be classified by neural network model, and predict that error can be controlled certain
Range, then model is correct effective.PPTL formula is described as follows:
P:bpnn_sim_percent > 0.9 define;
Define q:check_model=1;
Formula fin (p and q) meets, then proves that meeting correctness based on the nerve network system model that MSVL is constructed wants
It asks.
Further, it is described as in the shared property of the nerve network system about robustness, if meeting correctness
It is required that neural network model can withstand 1000 (experiment setting values) iteration attacks of FSGM, and the prediction error of sample
Consistently greater than mean error, then it is believed that model is sufficiently robust.PPTL formula is described as follows:
P:bpnn_sim_single≤1000 define;
Define q:bpnn_sim_single_Ek_ > aveEk;
Define r:bpnn_sim_single_Ek_out > aveEk;
Formula alw (p and q and r) meets, then it is strong to prove that the nerve network system model constructed based on MSVL is met
Strong property requirement.
Another object of the present invention is to provide a kind of modeling of the nerve network system described in application based on MSVL and test
The nerve network system of card method.
In conclusion advantages of the present invention and good effect are as follows: modeled using MSVL to nerve network system, utilize PPTL
Its shared property is described, is finally verified in UMC4MSVL platform, determines that the correctness and robustness in shared property are
It is no to can satisfy.Application field of the invention is the Formal Verification of nerve network system.
Nerve network system of the invention uses MSVL Procedure modeling, and the correctness and robustness shared in property uses
The description of PPTL formula, and PPTL is the proposition subset of PTL, MSVL is the executable subset of PTL, thus MSVL and PPTL can unite
One executes in UMC4MSVL platform, and compared to other methods, the present invention is not needed using other formal language, without right
System is largely tested, and mistake that may be present in system operation is able to carry out and is timely and effectively fed back.It is whole
A verification process is automatically performed by UMC4MSVL, is not necessarily to manual intervention.Due to the invention belongs to formalize field, and Formal
Method itself be again based on strict mathematical reasoning, so, by the programming procedure of formalization be applied to nerve network system
Modeling and verifying, whether software systems or hardware system, each state that program executes can obtain reliably
Verifying, has effectively ensured the safety of system itself.
Detailed description of the invention
Fig. 1 is modeling and the verification method flow chart of the nerve network system provided in an embodiment of the present invention based on MSVL.
Fig. 2 is the flow diagram of embodiment 1 provided in an embodiment of the present invention.
Fig. 3 is the execution flow chart of MSVL program provided in an embodiment of the present invention.
Fig. 4 is the verification result figure of correctness in shared property provided in an embodiment of the present invention.
Fig. 5 is the verification result figure of robustness in shared property provided in an embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
In view of the problems of the existing technology, the present invention provides a kind of modeling of nerve network system based on MSVL and
Verification method is with reference to the accompanying drawing explained in detail the present invention.
As shown in Figure 1, modeling and the verification method packet of the nerve network system provided in an embodiment of the present invention based on MSVL
Include following steps:
S101: modeling neural network (including DNN, the CNN, RNN etc.) system verified with MSVL, uses
The information of involved node and side in Multidimensional numerical expression system;In addition, in modeling, in function representation neural network system
Basic operation;
S102: the correctness and robustness feature of the nerve network system for needing to verify are determined, and public using PPTL
Formula describes these properties;
S103: the PPTL formula of the MSVL program of modeling and the shared property of description is unified in UMC4MSVL platform and is tested
Card, judges whether property can be met according to verification result.
Technical scheme of the present invention will be further described combined with specific embodiments below.
The embodiment of the present invention is by the PPTL formula of the MSVL program of modeling and the shared property of description, same
It is verified in UMC4MSVL platform, if be proved to be successful, the nerve network system of needs verifying meets correctness and robustness
It is required that otherwise the nerve network system of needs verifying just violates property.
The principle of the present invention is to be modeled with MSVL program p for nerve network system, and described using PPTL equation φ
Required property needs to prove the validity of formula p → φ to judge whether nerve network system meets this property,
If p → φ is effective, system meets property, and otherwise system just violates property.
The main sentence that MSVL program is related to:
(1) null statement: empty;
(2) basic assignment statement: x <==e,
(3) next sentence: Ox;
(4) always sentence: x;
(5) projected sentences: (s1,...,sm)prjs;
(6) sequential statement:
(7) parallel subqueries:
(8) conditional statement:
(9) while statement:
(10) state frame sentence:
(11) section frame sentence:
(12) Await sentence:Wherein,
x1,...,xhOccur from the variable in b.
The Data Structures of MSVL have:
(1) shaping: int;
(2) floating type: float;
(3) character type: char;
(4) character string type: string;
(5) basic data type pointer: int*/char*/float*/string*;
(6) structural body: struct;
(7) structural body pointer: struct*.
As shown in Fig. 2, modeling and the verification method of the nerve network system provided in an embodiment of the present invention based on MSVL, with
For BP neural network system, include the following steps:
The first step models nerve network system using MSVL, wherein describing system interior joint with Multidimensional numerical
With the information on side, specifically:
v[IN_N,HIDDEN_N] | Indicate the connection weight of input layer and hidden layer in neural network |
w[HIDDEN_N,1] | Indicate the connection weight of hidden layer and output layer in neural network |
r[HIDDEN_N] | Indicate the connection biasing of input layer and hidden layer in neural network |
o[OUT_N] | Indicate the connection biasing of hidden layer and output layer in neural network |
g[OUT_N] | Indicate the derivative value that node layer is exported in back-propagation process |
e[HIDDEN_N] | Indicate the derivative value of hidden layer node in back-propagation process |
f[IN_N] | Indicate the derivative value of input layer in back-propagation process |
x[IN_N] | Indicate the value of input layer in neural network |
b[HIDDEN_N] | Indicate the value that node layer is hidden in neural network |
y[OUT_N] | Indicate the theoretical value that node layer is exported in neural network |
yc[OUT_N] | Indicate the actual value that node layer is exported in neural network |
And the basic operation of neural network is then indicated with function, specifically:
functionread_data | Indicate reading and the assignment operation of data sample |
functionread_parameter | Indicate reading and the assignment operation of Connecting quantity |
functionbpnn_sim | Indicate propagated forward process |
functionbpnn_sim_single | Indicate that nerve network system receives the process of attack |
Second step is determined the shared property in the nerve network system for needing to verify, and is described using PPTL formula
These properties.PPTL formula is determined according to the shared property specifically to be verified, the correctness verified as needed first
And robustness, proposition is defined, then according to the logical relation of proposition, obtains corresponding reasonable PPTL formula.
It should be noted that PPTL can determine that, and PPTL can express all regular expressions, use it as
The Property specification language of MSVL program can describe more properties, so that it is guaranteed that program meets property of system description and satisfaction property
The demand of judgement, this provides theoretical basis to study the model inspection of sequential logic program.
The PPTL formula of the MSVL program of modeling and the shared property of description is unified in UMC4MSVL platform and tests by third step
Card, judges whether property can be met according to verification result.
It should be noted that being used to describe the MSVL program of nerve network system model and for descriptive model property
PPTL formula may be converted into PTL formula, it is possible to execute, and obtain a result in the same UMC4MSVL platform.
The authentication module of UMC4MSVL platform can complete the Property Verification based on PPTL formula.
Now by the program of one BP neural network system model of building, to the correctness in its potential shared property
It is verified with robustness.
As shown in figure 3, the structural parameters of data set and nerve network system are saved in the text in this program, number
Multidimensional numerical is transferred to handle according to after reading in text.Wherein, model program is stored in M file, and property formula is stored in P text
In part, EXE is the executable file that program compiling generates.The shared property description of nerve network system relate generally to correctness and
Robustness.
The main thought of modeling is as follows:
Firstly, defining nerve network system structure with Multidimensional numerical, wherein IN_N indicates the number of input layer,
HIDDEN_N indicates the number of hidden layer node, and OUT_N indicates the number of output node layer.Defined function read_data is read
Data set, defined function read_parameter read nerve network system structural parameters, defined function bpnn_sim simulation mind
Through network propagated forward process, defined function bpnn_sim_single simulation attack nerve network system.
Further, it defines data sample function reading read_data: reading institute from data set in.txt and out.txt
The sample data needed, constantly modification x [IN_N] and y [OUT_N], with the propagated forward process for nerve network system.
Definition structure parameter function reading read_parameter: trained nerve net is read from bpnn_param.txt
The structural parameters of network system are simultaneously successively assigned to array v [IN_N, HIDDEN_N], w [HIDDEN_N, 1], r [HIDDEN_N] and o
[OUT_N], with the propagated forward process for nerve network system.
It defines propagated forward analog function bpnn_sim: being used for simulative neural network system forward communication process.Function by
Row reads the sample data in data set, then acts on the nerve network system that building is completed, and generation, record, processing are corresponding
Output as a result, and being finally completed the evaluation to nerve network system.
It defines attack process analog function bpnn_sim_single: receiving the mistake of attack for simulative neural network system
Journey.Function uses the FGSM algorithm based on change of gradient, by sign function to original after each back-propagation process
Sample is purposefully disturbed, and the intermediate interference sample of generation is re-entered into nerve network system, recycles above-mentioned steps,
Until meeting termination condition.
The main code of modeling is as follows:
The Multidimensional numerical (part array information) of definition node and side information:
float v[3,5]and skip;
float w[5,1]and skip;
float r[5]and skip;
float o[1]and skip;
……
The part core code (defining neural network attack process) of function bpnn_sim_single:
It in this example, mainly include that correctness in shared property to nerve network system model and robustness are tested
Card.
Correctness is described as, if the sample that test is concentrated with 90% (experiment setting value) or more can be by neural network mould
Type is correctly classified, and predicts that error can be controlled in a certain range, then model is correct effective.
The verifying of PPTL formula:
P:bpnn_sim_percent > 0.9 define;
Define q:check_model=1;
Whether true need to verify fin (p and q).Proposition p indicates that in test set 90% data sample can pass through
Neural network model, proposition q indicate the prediction error of sample within the acceptable range, and fin (p and q) representation program executes
Final state proposition p and proposition q set up simultaneously.
Property Verification result are as follows:
The verification result about correctness in shared property is valid, and verification result is as shown in figure 4, show the mind
Meet correctness requirement through network system model.
Robustness is described as, if the neural network model for meeting correctness requirement can withstand 1000 times of FSGM
(experiment setting value) iteration attack, and the prediction error of sample is consistently greater than mean error, then it is believed that model is sufficiently robust
's.
The verifying of PPTL formula:
P:bpnn_sim_single≤1000 define;
Define q:bpnn_sim_single_Ek_ > aveEk;
Define r:bpnn_sim_single_Ek_out > aveEk;
Whether true need to verify alw (p and q and r).The number that iteration is attacked is limited in by proposition p expression
1000 times, prediction error is greater than mean error when proposition q indicates initial, and prediction error begins in proposition r representation program operational process
It is greater than mean error eventually, each state proposition p, q, r of alw (p and q and r) representation program operation are set up simultaneously.
Property Verification result are as follows:
The verification result about robustness in shared property is valid, and verification result is as shown in figure 5, show the mind
Meet robustness requirement through network system model.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (7)
1. modeling and the verification method of a kind of nerve network system based on MSVL, which is characterized in that the mind based on MSVL
Modeling and verification method through network system the following steps are included:
Step 1, models the nerve network system verified with MSVL, indicates institute in system with Multidimensional numerical
Relate to the information on node and side;In modeling, with the basic operation in function representation neural network system;
Step 2 is determined the correctness and robustness feature for needing the nerve network system verified, and is described using PPTL formula
The correctness and robustness feature of nerve network system;
The PPTL formula of the MSVL program of modeling and the shared property of description is unified in UMC4MSVL platform and verifies by step 3,
Judge whether property can be met according to verification result.
2. the modeling of the nerve network system based on MSVL and verification method as described in claim 1, which is characterized in that use
MSVL models nerve network system, wherein the information on system interior joint and side is described with Multidimensional numerical, and nerve net
The basic operation of network is then indicated with function.
3. the modeling of the nerve network system based on MSVL and verification method as described in claim 1, which is characterized in that PPTL
Formula is determining according to the shared property of the nerve network system specifically to be verified, the property definition proposition to be verified of basis first,
Then according to the logical relation of proposition, reasonable and strict PPTL formula is constructed.
4. the modeling of the nerve network system based on MSVL and verification method as described in claim 1, which is characterized in that described
The modeling method of nerve network system based on MSVL, specifically includes:
(1) nerve network system structure is defined with Multidimensional numerical;
(2) data sample function reading is defined;
(3) definition structure parameter function reading;
(4) propagated forward analog function is defined;
(5) attack process analog function is defined.
5. the modeling of the nerve network system based on MSVL and verification method as described in claim 1, which is characterized in that institute
It states in the shared property of nerve network system and is verified about the PPTL formula of correctness:
P:bpnn_sim_percent > 0.9 define;
Define q:check_model=1;
Formula fin (p and q) meets, then proves to meet correctness requirement based on the nerve network system model that MSVL is constructed.
6. the modeling of the nerve network system based on MSVL and verification method as described in claim 1, which is characterized in that institute
It states in the shared property of nerve network system and is verified about the PPTL formula of robustness:
P:bpnn_sim_single≤1000 define;
Define q:bpnn_sim_single_Ek_ > aveEk;
Define r:bpnn_sim_single_Ek_out > aveEk;
Formula alw (p and q and r) meets, then proves to meet robustness based on the nerve network system model that MSVL is constructed
It is required that.
7. a kind of modeling and verification method using the above-mentioned nerve network system based on MSVL of claim 1~6 any one
Nerve network system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910563393.XA CN110443348A (en) | 2019-06-26 | 2019-06-26 | A kind of modeling of the nerve network system based on MSVL and verification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910563393.XA CN110443348A (en) | 2019-06-26 | 2019-06-26 | A kind of modeling of the nerve network system based on MSVL and verification method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110443348A true CN110443348A (en) | 2019-11-12 |
Family
ID=68429006
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910563393.XA Pending CN110443348A (en) | 2019-06-26 | 2019-06-26 | A kind of modeling of the nerve network system based on MSVL and verification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110443348A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111143539A (en) * | 2019-12-31 | 2020-05-12 | 重庆和贯科技有限公司 | Knowledge graph-based question-answering method in teaching field |
CN111475321A (en) * | 2020-05-08 | 2020-07-31 | 中国人民解放军国防科技大学 | Neural network security property verification method based on iterative abstract analysis |
CN113240075A (en) * | 2021-04-23 | 2021-08-10 | 西安电子科技大学 | BP neural network construction and training method and system based on MSVL |
CN113434132A (en) * | 2021-05-08 | 2021-09-24 | 西安电子科技大学 | Intelligent course arrangement modeling verification method and system |
-
2019
- 2019-06-26 CN CN201910563393.XA patent/CN110443348A/en active Pending
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111143539A (en) * | 2019-12-31 | 2020-05-12 | 重庆和贯科技有限公司 | Knowledge graph-based question-answering method in teaching field |
CN111143539B (en) * | 2019-12-31 | 2023-06-23 | 重庆和贯科技有限公司 | Knowledge graph-based teaching field question-answering method |
CN111475321A (en) * | 2020-05-08 | 2020-07-31 | 中国人民解放军国防科技大学 | Neural network security property verification method based on iterative abstract analysis |
CN111475321B (en) * | 2020-05-08 | 2024-04-26 | 中国人民解放军国防科技大学 | Neural network security property verification method based on iterative abstract analysis |
CN113240075A (en) * | 2021-04-23 | 2021-08-10 | 西安电子科技大学 | BP neural network construction and training method and system based on MSVL |
CN113240075B (en) * | 2021-04-23 | 2023-08-22 | 西安电子科技大学 | Construction and training method and system of BP neural network based on MSVL (modeling, simulation and simulation verification) |
CN113434132A (en) * | 2021-05-08 | 2021-09-24 | 西安电子科技大学 | Intelligent course arrangement modeling verification method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110443348A (en) | A kind of modeling of the nerve network system based on MSVL and verification method | |
Wang et al. | A cooperative game for automated learning of elasto-plasticity knowledge graphs and models with AI-guided experimentation | |
US10521215B2 (en) | Technologies for customized crowd-sourced features, automated safety and quality assurance with a technical computing environment | |
Volodina et al. | The importance of uncertainty quantification in model reproducibility | |
Kessentini et al. | Example-based model-transformation testing | |
Kraus et al. | Artificial intelligence for structural glass engineering applications—overview, case studies and future potentials | |
WO2021115186A1 (en) | Ann-based program test method and test system, and application | |
CN107783758A (en) | A kind of intelligent contract engineering method | |
Sahin et al. | Model transformation testing: a bi‐level search‐based software engineering approach | |
Goknil et al. | Tool support for generation and validation of traces between requirements and architecture | |
Al-Zwainy et al. | Structural equation modeling of critical success factors in the programs of development regional | |
Nguyen et al. | Assertion-based monitoring in practice–checking correctness of an automotive sensor interface | |
US7310790B2 (en) | Automatic symbolic indexing methods for formal verification on a symbolic lattice domain | |
Raunak et al. | Metamorphic testing on the continuum of verification and validation of simulation models | |
Dick et al. | Computational Intelligence in Software Quality Assurance | |
Sohier et al. | Improving simulation specification with MBSE for better simulation validation and reuse | |
Pravin et al. | An efficient programming rule extraction and detection of violations in software source code using neural networks | |
Gaaloul et al. | Combining genetic programming and model checking to generate environment assumptions | |
Weiß et al. | Towards establishing formal verification and inductive code synthesis in the PLC domain | |
Royce | Status report: computer-aided prototyping | |
Hu et al. | Verifying and validating a simulation model | |
US20040107174A1 (en) | Parametric representation methods for formal verification on a symbolic lattice domain | |
Khastgir et al. | Incorporating ISO 26262 Concepts in an Automated Testing Toolchain Using Simulink Design Verifier™ | |
Schaffeld et al. | Towards robust models of Cyber-Physical Systems | |
Khatami Firouzabadi | A Note on Models' Verification, Validation and Calibration |
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 |