CN111753892A - Deep learning-based interpretation method of global visual field network system - Google Patents

Deep learning-based interpretation method of global visual field network system Download PDF

Info

Publication number
CN111753892A
CN111753892A CN202010532906.3A CN202010532906A CN111753892A CN 111753892 A CN111753892 A CN 111753892A CN 202010532906 A CN202010532906 A CN 202010532906A CN 111753892 A CN111753892 A CN 111753892A
Authority
CN
China
Prior art keywords
global
matrix
network system
hypergraph
visual field
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
Application number
CN202010532906.3A
Other languages
Chinese (zh)
Other versions
CN111753892B (en
Inventor
徐明伟
孟子立
王敏虎
白家松
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN202010532906.3A priority Critical patent/CN111753892B/en
Publication of CN111753892A publication Critical patent/CN111753892A/en
Application granted granted Critical
Publication of CN111753892B publication Critical patent/CN111753892B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention relates to the technical field of internet information, in particular to an interpretation method of a global visual field network system based on deep learning. The method carries out causal explanation and conversion on the decision of the computer network system based on deep learning under the condition of global visual field. Firstly, training an original network system by adopting a deep reinforcement learning method, modeling a generated global configuration result in a hypergraph mode after the training of the original system based on the deep learning is completed, analyzing key point-hyper-edge connections in the hypergraph, and scoring the influence of each point-hyper-edge connection on a final global configuration result so that a network administrator understands key components in decision making. The method greatly reduces the understanding difficulty of the original deep learning-based global visual field network system, and facilitates the network administrator to understand the decision process. When the interpretation method is deployed on an actual system, a network administrator is facilitated to understand and correct the decision process of the original global view network system.

Description

Deep learning-based interpretation method of global visual field network system
Technical Field
The invention relates to the technical field of internet information, in particular to an interpretation method of a global visual field network system based on deep learning.
Background
Computer network systems can be generally divided into systems with a local view and systems with a global view. The local view refers to that the network system is deployed on a server side, a client side or middleware, a switch (such as a congestion control system), and the system with the local view can only observe information of one point in the system and make a decision. A system with a global view, such as a network management controller, a traffic engineering scheduling system, a software defined network controller, etc., can observe and make decisions for multiple devices in the network. In decision logic, part of the existing global visual field network systems adopt a deep learning technology as a decision algorithm thereof, such as a software defined network route optimization algorithm based on a deep neural network.
In a decision strategy of a conventional network system, decision is often made through a concise strategy such as 'multiplication and subtraction'. The decision mode of the existing deep learning-based systems cannot be understood by a network administrator: neural networks often contain thousands of neurons, and the final conclusion is reached through a series of nonlinear calculations. Thus, while performing well in training, network administrators are not able to understand the logic of their decisions and therefore tend to have difficulty gaining trust.
Disclosure of Invention
The invention aims to provide an interpretation method of a deep learning-based global view network system, which is used for causally interpreting and converting decisions of the deep learning-based computer network system under the condition of a global view. Firstly, training an original network system by adopting a deep reinforcement learning method, after the training of the original system based on the deep learning is completed, modeling a global configuration result generated by the system in a hypergraph (hypergraph) mode, analyzing key point-edge (hyperedge) connections in the hypergraph, and scoring the influence of each point-edge connection on the final global configuration result so as to enable a network manager to understand key components in decision making.
The invention provides an interpretation method of a global visual field network system based on deep learning, which comprises the following steps:
(1) inputting resources and requests into a global view network system S to be interpreted, outputting to obtain a global configuration result set, and recording the global configuration result set as O;
(2) constructing a global-view computer network system hypergraph H comprising resources and requests, the global-view computer network system hypergraph H having the following four forms:
a. when the global visual field network system S is a software defined network routing optimization system, physical links in the hypergraph H are constructed into points, and routing paths are constructed into hyperedges;
b. when the global visual field network system S is used for placing an optimization system for a virtual network function, a physical server in the hypergraph H is constructed into a point, and the virtual network function is a hyperedge;
c. when the global visual field network system S is a super-dense cellular network optimization system, mobile users in the hypergraph H are constructed into points, and a base station is constructed into a super edge;
d. when the global visual field network system S is a cluster task scheduling optimization system, the request tasks in the hypergraph H are constructed into points, and the dependency relationship among the tasks is constructed into a hypergraph;
representing a hypergraph H of the global visual field computer network system by using an incidence matrix I, wherein the incidence matrix I is a V multiplied by E dimensional 0-1 matrix, when the position (V, E) in the incidence matrix I is 1, the incidence matrix I represents that a point V and a hyperedge E have a connection relation, and when the position (V, E) is 0, the incidence matrix I represents that the point V and the hyperedge E have no connection relation;
(3) and (3) performing feature calculation on the hypergraph H of the global-view computer network system in the step (2), wherein the feature calculation comprises the following steps:
(3-1) constructing an evaluation matrix W for representing the importance of each non-zero element in the incidence matrix I;
(3-2) calculating the performance loss of the evaluation matrix W:
judging the global configuration result set O in the step (1), and if the output result of the global configuration result in the step (1) is a discrete variable, calculating KL divergence D (W, I) of the incidence matrix I and the evaluation matrix W by using the following formula; if the output result of the global configuration result in the step (1) is a continuous variable, calculating the mean square error D (W, I) of the global configuration result correlation matrix I and the evaluation matrix W by using the following formula:
Figure BDA0002536027210000021
in the above formula, f (W) is the evaluation result of the evaluation matrix W by the evaluation function of the global view network system S to be interpreted, and f (I) is the evaluation result of the correlation matrix I of the global view network system S to be interpreted;
(3-3) calculating the conciseness of the evaluation matrix W in the step (3-1) by using the following formula, and defining the conciseness of the evaluation matrix W as:
Figure BDA0002536027210000022
in the above formula, WevTo evaluate the element value of the matrix W at the (e, v) position of the incidence matrix I;
(3-4) representing the certainty of the evaluation matrix W by adopting the entropy H (W) of the evaluation matrix W:
Figure BDA0002536027210000031
(3-5) establishing an optimization model for solving the evaluation matrix W, wherein an objective function of the optimization model is as follows:
minD(W,I)+λ1||W||+λ2H(W)
the constraint conditions of the optimization model are as follows:
Figure BDA0002536027210000032
wherein D (W, I) is the mean square error in step (3-2), | | W | | is the simplicity in step (3-3), H (W) is the entropy of the evaluation matrix W in step (3-4), and λ1And λ2Respectively calculating parameters of simplicity and entropy;
(4) and (3) solving the optimization model in the step (3-5) by adopting a gradient descent method to obtain an evaluation matrix W, and obtaining the representation of the importance of each connection relation in the incidence matrix I according to the evaluation matrix W to realize the explanation of the global visual field network system based on deep learning.
The invention provides an interpretation method of a global visual field network system based on deep learning, which has the characteristics and advantages that:
according to the deep learning-based global visual field network system interpretation method, the output of the deep learning-based global visual field network system is converted into an equivalent hypergraph, and the importance of each connection relation in the hypergraph is quantitatively shown, so that the understanding difficulty of the original deep learning-based global visual field network system is greatly reduced, and a network administrator can conveniently understand the decision process of the result. When the interpretation method is deployed on an actual system, a network administrator is facilitated to understand and correct the decision process of the original global view network system.
Drawings
Fig. 1 shows a global routing result O of a network routing system S constructed in an embodiment of the method of the present invention.
FIG. 2 is a hypergraph H converted from a global routing result O in an embodiment of the present invention.
FIG. 3 shows the result of the important connection relationship obtained in the embodiment of the present invention.
Detailed Description
The invention provides an interpretation method of a global visual field network system based on deep learning, which comprises the following steps:
(1) inputting resources and requests into a global view network system S to be interpreted, outputting to obtain a global configuration result set, and recording the global configuration result set as O; for a software-defined network route optimization system based on deep learning, a set O of configuration results is what path traffic at any two points in the network should be forwarded along;
(2) a global view computer network system hypergraph H is constructed which includes resources and requests, and the network system with the global view is the allocation of the resources and requests, such as the allocation of link resources to traffic requests, the allocation of physical machine resources to virtual machine service requests, and the like. Thus, resources and requests can be represented as points and hyper-edges in the hyper-graph, respectively. The global-view computer network system hypergraph H has the following four forms:
a. when the global visual field network system S is a software defined network routing optimization system, physical links in the hypergraph H are constructed into points, and routing paths are constructed into hyperedges;
b. when the global visual field network system S is used for placing an optimization system for a virtual network function, a physical server in the hypergraph H is constructed into a point, and the virtual network function is a hyperedge;
c. when the global visual field network system S is a super-dense cellular network optimization system, mobile users in the hypergraph H are constructed into points, and a base station is constructed into a super edge;
d. when the global visual field network system S is a cluster task scheduling optimization system, the request tasks in the hypergraph H are constructed into points, and the dependency relationship among the tasks is constructed into a hypergraph;
a correlation matrix I is used for representing a hypergraph H of the global visual field computer network system, the correlation matrix I is a V multiplied by E dimensional 0-1 matrix, when the position (V, E) in the correlation matrix I is 1, the point V and the hypergraph E are represented to have a connection relation, and when the position (V, E) is 0, the point V and the hypergraph E are represented to have no connection relation, as shown in FIG. 1;
(3) and (3) performing feature calculation on the hypergraph H of the global-view computer network system in the step (2), wherein the feature calculation comprises the following steps:
(3-1) constructing an evaluation matrix W for representing the importance of each non-zero element in the incidence matrix I;
(3-2) calculating the performance loss of the evaluation matrix W:
judging the global configuration result set O in the step (1), and if the output result of the global configuration result in the step (1) is a discrete variable, calculating KL divergence D (W, I) of the incidence matrix I and the evaluation matrix W by using the following formula; if the output result of the global configuration result in the step (1) is a continuous variable, calculating the mean square error D (W, I) of the global configuration result correlation matrix I and the evaluation matrix W by using the following formula:
Figure BDA0002536027210000041
in the above formula, f (W) is the evaluation result of the evaluation matrix W by the evaluation function of the global view network system S to be interpreted, and f (I) is the evaluation result of the correlation matrix I of the global view network system S to be interpreted;
(3-3) calculating the conciseness of the evaluation matrix W in the step (3-1) by using the following formula, and defining the conciseness of the evaluation matrix W as:
Figure BDA0002536027210000042
in the above formula, WevTo evaluate the element value of the matrix W at the (e, v) position of the incidence matrix I;
(3-4) representing the certainty of the evaluation matrix W by adopting the entropy H (W) of the evaluation matrix W: at the same time, it is also expected that importance should be clustered towards 0 or 1 to indicate the dichotomous nature of importance: either important or not.
Figure BDA0002536027210000051
(3-5) establishing an optimization model for solving the evaluation matrix W, wherein an objective function of the optimization model is as follows:
minD(W,I)+λ1||W||+λ2H(W)
the constraint conditions of the optimization model are as follows:
Figure BDA0002536027210000052
wherein D (W, I) is the mean square error in step (3-2), | | W | | is the simplicity in step (3-3), H (W) is the entropy of the evaluation matrix W in step (3-4), and λ1And λ2Respectively calculating parameters of simplicity and entropy; lambda [ alpha ]1And λ2Can be set according to the preference of a network administrator for the corresponding property, and in one embodiment of the invention, lambda1And λ2Is 0.5 and 0.5;
(4) and (3) solving the optimization model in the step (3-5) by adopting a gradient descent method to obtain an evaluation matrix W, and obtaining the representation of the importance of each connection relation in the incidence matrix I according to the evaluation matrix W to realize the explanation of the global visual field network system based on deep learning.
The content of the method of the invention is described in detail below with reference to the accompanying drawings:
setting a software-defined network routing system S based on deep learning, and generating a routing decision result as shown in FIG. 1, wherein in FIG. 1, a to g are routers, 1 to 8 are physical links, and blue and red are routing paths from a to e and from a to g, which are calculated by the network routing system S. Constructing a hypergraph H of a network routing system S, wherein physical links are constructed as points, and routing paths are constructed as hyperedges; as shown in fig. 2. And calculating the evaluation matrix W to obtain the key influence on the final optimization target in the point-hyper-edge connection in the hyper-graph. For example, the algorithm finds that the path selection link 6 has a large influence on the global optimization result, and may prompt the network administrator to: note that link 6 acts as a performance bottleneck. Similar operations can be done for larger scale topologies, with the final possible visualization shown in fig. 3.

Claims (1)

1. A deep learning-based interpretation method of a global visual field network system is characterized by comprising the following steps:
(1) inputting resources and requests into a global view network system S to be interpreted, outputting to obtain a global configuration result set, and recording the global configuration result set as 0;
(2) constructing a global-view computer network system hypergraph H comprising resources and requests, the global-view computer network system hypergraph H having the following four forms:
a. when the global visual field network system S is a software defined network routing optimization system, physical links in the hypergraph H are constructed into points, and routing paths are constructed into hyperedges;
b. when the global visual field network system S is used for placing an optimization system for a virtual network function, a physical server in the hypergraph H is constructed into a point, and the virtual network function is a hyperedge;
c. when the global visual field network system S is a super-dense cellular network optimization system, mobile users in the hypergraph H are constructed into points, and a base station is constructed into a super edge;
d. when the global visual field network system S is a cluster task scheduling optimization system, the request tasks in the hypergraph H are constructed into points, and the dependency relationship among the tasks is constructed into a hypergraph;
representing a hypergraph H of the global visual field computer network system by using an incidence matrix I, wherein the incidence matrix I is a V multiplied by E dimensional 0-1 matrix, when the position (V, E) in the incidence matrix I is 1, the incidence matrix I represents that a point V and a hyperedge E have a connection relation, and when the position (V, E) is 0, the incidence matrix I represents that the point V and the hyperedge E have no connection relation;
(3) and (3) performing feature calculation on the hypergraph H of the global-view computer network system in the step (2), wherein the feature calculation comprises the following steps:
(3-1) constructing an evaluation matrix W for representing the importance of each non-zero element in the incidence matrix I;
(3-2) calculating the performance loss of the evaluation matrix W:
judging the global configuration result set 0 in the step (1), and if the output result of the global configuration result in the step (1) is a discrete variable, calculating KL divergence D (W, I) of the incidence matrix I and the evaluation matrix W by using the following formula; if the output result of the global configuration result in the step (1) is a continuous variable, calculating the mean square error D (W, I) of the global configuration result correlation matrix I and the evaluation matrix W by using the following formula:
Figure FDA0002536027200000011
in the above formula, f (W) is the evaluation result of the evaluation matrix W by the evaluation function of the global view network system S to be interpreted, and f (I) is the evaluation result of the correlation matrix I of the global view network system S to be interpreted;
(3-3) calculating the conciseness of the evaluation matrix W in the step (3-1) by using the following formula, and defining the conciseness of the evaluation matrix W as:
Figure FDA0002536027200000021
in the above formula, WevFor evaluating the elements of the matrix W at the (e, v) positions of the correlation matrix IThe prime value;
(3-4) representing the certainty of the evaluation matrix W by adopting the entropy H (W) of the evaluation matrix W:
Figure FDA0002536027200000022
(3-5) establishing an optimization model for solving the evaluation matrix W, wherein an objective function of the optimization model is as follows:
minD(W，I)+λ1||W||+λ2H(W)
the constraint conditions of the optimization model are as follows:
Figure FDA0002536027200000023
wherein D (W, I) is the mean square error in step (3-2), | | W | | is the simplicity in step (3-3), H (W) is the entropy of the evaluation matrix W in step (3-4), and λ1And λ2Respectively calculating parameters of simplicity and entropy;
(4) and (3) solving the optimization model in the step (3-5) by adopting a gradient descent method to obtain an evaluation matrix W, and obtaining the representation of the importance of each connection relation in the incidence matrix I according to the evaluation matrix W to realize the explanation of the global visual field network system based on deep learning.
CN202010532906.3A 2020-06-12 2020-06-12 Deep learning-based interpretation method of global visual field network system Active CN111753892B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010532906.3A CN111753892B (en) 2020-06-12 2020-06-12 Deep learning-based interpretation method of global visual field network system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010532906.3A CN111753892B (en) 2020-06-12 2020-06-12 Deep learning-based interpretation method of global visual field network system

Publications (2)

Publication Number Publication Date
CN111753892A true CN111753892A (en) 2020-10-09
CN111753892B CN111753892B (en) 2022-10-18

Family

ID=72676054

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010532906.3A Active CN111753892B (en) 2020-06-12 2020-06-12 Deep learning-based interpretation method of global visual field network system

Country Status (1)

Country Link
CN (1) CN111753892B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170116520A1 (en) * 2015-10-23 2017-04-27 Nec Laboratories America, Inc. Memory Efficient Scalable Deep Learning with Model Parallelization
CN108171735A (en) * 2017-12-27 2018-06-15 清华大学 1,000,000,000 pixel video alignment schemes and system based on deep learning
CN109190626A (en) * 2018-07-27 2019-01-11 国家新闻出版广电总局广播科学研究院 A kind of semantic segmentation method of the multipath Fusion Features based on deep learning
CN109460501A (en) * 2018-11-15 2019-03-12 成都傅立叶电子科技有限公司 A kind of global search Battle Assistant Decision-making system and method
CN109919319A (en) * 2018-12-31 2019-06-21 中国科学院软件研究所 Deeply learning method and equipment based on multiple history best Q networks

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170116520A1 (en) * 2015-10-23 2017-04-27 Nec Laboratories America, Inc. Memory Efficient Scalable Deep Learning with Model Parallelization
CN108171735A (en) * 2017-12-27 2018-06-15 清华大学 1,000,000,000 pixel video alignment schemes and system based on deep learning
CN109190626A (en) * 2018-07-27 2019-01-11 国家新闻出版广电总局广播科学研究院 A kind of semantic segmentation method of the multipath Fusion Features based on deep learning
CN109460501A (en) * 2018-11-15 2019-03-12 成都傅立叶电子科技有限公司 A kind of global search Battle Assistant Decision-making system and method
CN109919319A (en) * 2018-12-31 2019-06-21 中国科学院软件研究所 Deeply learning method and equipment based on multiple history best Q networks

Also Published As

Publication number Publication date
CN111753892B (en) 2022-10-18

Similar Documents

Publication Publication Date Title
Almasan et al. Digital twin network: Opportunities and challenges
JP2001236337A (en) Predicting device using neural network
US20180062958A1 (en) Neural network learning methods to identify network ports responsible for packet loss or delay
CN113360349A (en) Information optimization method based on big data and cloud service and artificial intelligence monitoring system
US7991617B2 (en) Optimum design management apparatus from response surface calculation and method thereof
Moradi et al. Comparison of machine learning techniques for VNF resource requirements prediction in NFV
CN116455820A (en) Multi-transmission path adjustment system and method based on congestion avoidance
CN116737436A (en) Root cause positioning method and system for micro-service system facing mixed deployment scene
CN114401516A (en) 5G slice network anomaly detection method based on virtual network traffic analysis
Morais Machine learning in multi-layer optical networks: why and how
CN111753892B (en) Deep learning-based interpretation method of global visual field network system
US11487747B2 (en) Anomaly location identification device, anomaly location identification method, and program
CN100356726C (en) Method and device for designing a data network
Salmen et al. Using firefly and genetic metaheuristics for anomaly detection based on network flows
CN110166362B (en) Service function graph low-delay mapping method based on node screening
Bolodurina et al. Model control of traffic by using data flows classification of the cloud applications in software-defined infrastructure of virtual data center
CN115437858A (en) Edge node abnormity positioning method, device, equipment and computer program product
JP6539237B2 (en) Communication network analysis device, communication network analysis method and communication network analysis program
JP2010134926A (en) Multi-modal green computing fusion system, computer program, method, and developing method of system (multi-modal green computing fusion using problem analytics)
Nguyen et al. Jointly optimize coding and node selection for distributed computing over wireless edge networks
KR20200126212A (en) Deep learning-based dynamic routing technology to achieve maximum user request throughput and minimum intra-communication latency in high performance computing environments with interconnection network based multi-node cluster
KR102199432B1 (en) Methods and devices for selecting optimal controller in Software Defined Networking
CN114819328B (en) Label prediction method, device, equipment and storage medium
Queiroz et al. Survivable SCADA Systems: An analytical framework using performance modelling
CN118037514B (en) Regional development difference analysis method, system and terminal based on analog network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant