CN103501346A - Non-structured P2P (Peer-to-Peer) resource searching method based on machine learning and network node reputation - Google Patents
Non-structured P2P (Peer-to-Peer) resource searching method based on machine learning and network node reputation Download PDFInfo
- Publication number
- CN103501346A CN103501346A CN201310479076.2A CN201310479076A CN103501346A CN 103501346 A CN103501346 A CN 103501346A CN 201310479076 A CN201310479076 A CN 201310479076A CN 103501346 A CN103501346 A CN 103501346A
- Authority
- CN
- China
- Prior art keywords
- network node
- resource
- value
- node
- target resource
- 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
Images
Abstract
The invention relates to a non-structured P2P (Peer-to-Peer) resource searching method based on machine learning and network node reputation. The method comprises the following steps: I, setting time to live (TTL); II, judging whether a target resource exists or not in a network node; III, judging the TTL of a request message; IV, judging whether a Q value relevant to the target resource in a table Q is empty or not; V, calculating a reward; VI, searching for the target resource; searching for resources from neighboring network nodes till the target resource is found or the TTL becomes zero. According to the method, the machine learning speed can be increased, the resource searching time is reduced, and the searching success rate is increased; meanwhile the safety and reliability of services are ensured.
Description
Technical field
The present invention relates to the network service searching method, be specifically related to a kind of non-structural P 2 P resource search method based on machine learning and network node prestige.
Technical background
Develop rapidly and the extensive use of Internet in the social life every field and universal rapidly along with network technology, amount of information and number of users grow with each passing day, making existing C/S model (is Client/Server, client/server) can't meet and support large-scale network application, produce thus the P2P technology.P2P is Peer-to-Peer, also referred to as equity, calculates or peer-to-peer network, and be a kind of overlay network built on physical network.The P2P technology has been widely used in the numerous areas such as file-sharing, reciprocity calculating, collaborative work, instant messaging at present.P2P is as one of key technology of future network, and key problem is exactly for a large amount of users provides reliable service, realizes that Internet resources share.But want to utilize fully the various resources in the P2P network, key issue is fast and effeciently to find resource.At present, in non-structural P 2 P, resource search method can be divided into two kinds of blind search and information searches:
So-called blind search is under the complete prerequisite without any relevant information, attempts to meet the resource query request by the network node of access sufficient amount; This searching method can produce bulk redundancy message, the waste bandwidth resource.
So-called information search is to utilize obtain for information about various, adopts heuristic to instruct current search, makes system pointed when query resource, overcomes the deficiency of blind search, and the method improves resource query efficiency to a certain extent; But the search success rate of this simple information search method is low, search efficiency can not get ensureing.
The topological structure of non-structural P 2 P is random, and this substantive characteristics has also caused this system to have larger uncertainty when resource query, therefore, how to improve search efficiency significant for non-structural P 2 P network.
Existing in the majority based on information search method in the searching method of non-structural P 2 P, as ant group algorithm, mobile agent and the Semantic Clustering algorithm based on Small World Model, network node prestige method, etc.; From the size of message of search success rate, generation, the aspects such as the adaptability of network dynamic change and network operation are studied respectively these methods and shown, can improve the search efficiency of resource under certain condition.But these methods also have various defects, for example set up the searching method of Semantic Small World, before search, integration cluster to resource just needs a large amount of cost of cost, in addition when resource query, the P2P system bandwidth waste problem caused due to a large amount of redundancy messages does not solve yet, and search efficiency also can not get ensureing.Therefore, for problems such as resource searching efficiency in non-structural P 2 P are low, needing further research to find resource searching mechanism more efficiently is the research emphasis of current P2P technology.
CN 101364958A disclosed " a kind of searching method based on non-structural P 2 P network ", its search procedure is: at first any one website on network calculates respectively one group of ttl value and offered load value Load according to formula TTL=round (lognN)+2, Load=nTTL, and calculates the priority of each class value; Count n and message bag time-to-live ttl value of neighbor station corresponding to maximum in many class values of priority is defined as to final argument; Then generate a query messages according to query contents, send to the determined n that finally will send in abutting connection with website; J neighbor station receiving this query messages named a person for a particular job after ttl value subtracts j, and search our station resource, if our station has the message that will inquire about, return to the transmission website, otherwise, this query messages is transmitted to other in abutting connection with website; And so forth until ttl value become 0 or search desired resource after stop.The method has advantages of that the search success rate is high and offered load is little, is used in the destructuring network and searches for the resource informations such as document, music, film.
Q study is a kind of typical machine learning method, and its thought source, in conditioned reflex theory and animal learning theory, is widely used in every field.As disclosed as CN 101634995 A " a kind of network connection speed predicting method based on machine learning ".The method comprises the following steps: 1) utilize custom browser, the connection speed of recording user and the website of browsing, as training set and test set; 2) utilize the website connection speed obtained, use neural metwork training and predict the connection speed of each website in this user and training set; 3) reduce situation according to the predicated error of all neural nets, or perform step 4), or training set is divided into to less training set and each training set is returned to execution step 2); 4) use the estimated performance of a decision tree test neural net; 5) use decision tree and neural net, the connection speed of predictive user and unknown website.The present invention utilizes artificial intelligence technology, and the applied for machines learning method is carried out the connection speed of predictive user and each website, promotes the precision of critic network situation, takes full advantage of user bandwidth, for the user provides better the Internet, experiences.
Have the researcher that the method for Q study is applied in the P2P resource searching based on grid, the method requires each network node all to safeguard a Q table; The Q value of each neighbor networks node that the Q table has comprised this network node, the resource accessibility of this neighbor networks node of Q value representation.Originally, the Q value of Q matrix section resource is empty, and network node is set up gradually and enrich Q table information separately by learning network state constantly.When network node can not meet searching request at self, the neighbor networks node of Q table Information Selection down hop according to self, be forwarded to the next-hop network node by request message; With additive method, compare, although the method has adaptivity preferably, the method has learning rate defect slowly; ' it is specially: due to the Q table when initial for empty, network node is not about the relevant information of target resource, and in this case, the network node selection strategy of the method is to adopt random selection strategy, forwarding to searching request has certain blindness, and search efficiency is lower; In addition; the dynamic of P2P network, self-organization and anonymity can not guarantee that all response to network nodes all provide honest service and reliable resource; some network node even provides the swindle service of malice; make requestor and the normal Gains resources of other network node of service, the method is not considered this problem yet.
Summary of the invention
Initially adopt random blindly Forward-reques for Q study in inquiry, the problems such as learning rate slowly and service safe is poor, the purpose of this invention is to provide a kind of non-structural P 2 P resource search method based on machine learning and network node prestige, the method can be accelerated machine learning speed, reduce the resource query time, improve the search success rate, ensure the security reliability of service simultaneously.
A kind of non-structural P 2 P resource search method based on machine learning and network node prestige of the present invention, comprise the steps:
Step 1, arrange life span; When a certain network node is received the resource request message that comprises source network node ID, resource name and life span TTL, the initial value that life span TTL is set is 10;
Step 2, judge whether network node exists target resource; Whether network node inspection this locality exists target resource, if Y, there is target resource in this locality, directly returns to the source network node resource response message; If N, there is not target resource in this locality, using resource request message as new information, joins the present networks node;
Step 3, the life span of judgement request message; Whether the life span TTL that judges request message is 0, if Y, search procedure finishes; If N, be forwarded to request message the neighbor networks node;
Step 4, judge in the Q table, whether the Q value about target resource is empty; At first network node checks present networks node Q table, if be judged as N, the Q value of target resource is arranged, and selects the neighbor networks node with maximum Q value, and, to its Forward-reques message, the life span ttl value in request message reduces 1 simultaneously; If be judged as Y, there is no the Q value of target resource, network node is checked self prestige table, select network node that credit value is the highest as the neighbor networks node, and, to its Forward-reques message, the life span ttl value in request message reduces 1 simultaneously; At a period of time week after date, according to the number of times of neighbor networks node success downloaded resources, calculate and upgrade credit value;
Step 5, calculate award; Request message, to the neighbor networks node, calculates award, upgrades the Q value of previous dive network node;
Step 6, the search target resource; And start in neighbor networks node searching resource, until find target resource or TTL, be 0 end.
The present invention is at the search initial stage, and in the Q of network node table, the Q value of part resource is empty, and network node is the unceasing study resource information in search procedure, calculates reward value and upgrades the Q table.Along with the time increases, the information of network node Q table can get more and more, network node is according to Q table guidance search process, therefore, in the target node resource request, increased judgment condition, judge in the Q table, whether the Q value about target resource is empty, when the Q value of target resource is empty in judgement Q table simultaneously, utilize the credit worthiness of network node; The credit worthiness of each network node is the overall merit of making according to the experience of the historical behavior of this network node and information interaction, its concrete selection course is: network node receives certain resource request, the life span TTL(Time To Live that it comprises source network node ID (Identity), resource name and resource request message), if the present networks node has target resource, return to the source network node response message, and finish inquiry request; If present networks node driftlessness resource forwards resource request message by following strategy:
At first network node checks present networks node Q table, judges in the Q table, whether the Q value about target resource is empty, if not empty, selects neighbor networks node corresponding to maximum Q value as next-hop network node Forward-reques; If the Q table about target resource is sky, network node is checked self prestige table, selects the neighbor networks node of credit value maximum, and resource request is forwarded, at a period of time week after date, according to the number of times of neighbor networks node success downloaded resources, calculate and upgrade credit value.
In P2P, each network node is safeguarded a Q table, network node forwards resource request message according to the Q table information of self, request message of every forwarding, to from this hunting action, obtain an award (reward), when request message arrives the next-hop network node, the value of the Q about this resource and award according to new network node, feed back to the upper hop network node, upgrades the Q value of upper hop network node.But the Q value of working as the Q value matrix section resource of network node is empty, the network node credit mechanism that the present invention introduces, changed original random forwarding strategy, network node is safeguarded a prestige table, because the network node credit worthiness is the overall merit of making according to the experience of the historical behavior of network node and information interaction, degree of belief is higher, illustrate that between network node, information success interaction times is more, satisfaction to service is larger, therefore, the present invention on the basis of Q study in conjunction with faith mechanism, not only can solve the fraud of hostile network node in network, can utilize the trusting relationship between network node simultaneously, the selection interaction experiences is many, the neighbor networks node forwarding inquiries request that trust value is higher, to make up the defect of Q table information deficiency, accelerate the Q learning process, thereby reduce the resource query time, improve the resource searching success rate.
The present invention has analyzed the deficiency of existing P2P system resource searching method, and proposition improves and optimizates scheme on this basis, mainly combine Q machine learning method and network node credit mechanism, at the search initial stage, network node is by judging whether the Q table is that sky is selected concrete forwarding strategy about the Q value of resource, the neighbor networks node Forward-reques that this method enjoys a good reputation by selection, avoided original random forwarding strategy, this method is being accelerated the Q learning process, reduce the query time of resource, improve the search success rate, and the fail safe aspect that ensures inquiry service, there is effect preferably.
The present invention be directed to and traditional based on Q study learning speed, slowly and not consider what this problem of security reliability of resource proposed, because the network node credit worthiness is the overall merit of making according to the experience of the historical behavior of network node and information interaction, degree of belief is higher, illustrate that between network node, information success interaction times is more, satisfaction to service is larger, therefore, the present invention on the basis of Q study in conjunction with faith mechanism, not only can solve the fraud of hostile network node in network, can utilize the trusting relationship between network node simultaneously, the selection interaction experiences is many, the neighbor networks node forwarding inquiries request that trust value is higher, not only can make up the defect of Q table information deficiency, accelerate the Q learning process, thereby reduce the resource query time, improve the resource searching success rate, and ensured that the user can obtain reliable service when resource-sharing, improve fail safe.
The accompanying drawing explanation
Fig. 1 is resource searching flow chart of the present invention;
Fig. 2 is the Q value table of network node resource;
Fig. 3 is trusting relationship schematic diagram between network node.
Embodiment
Below in conjunction with accompanying drawing, the present invention is further illustrated:
Shown in Figure 1: a kind of non-structural P 2 P resource search method based on machine learning and network node prestige comprises the steps:
Step 1, arrange life span; When a certain network node is received the resource request message that comprises source network node ID, resource name and life span TTL, the initial value that life span TTL is set is 10;
Step 2, judge whether network node exists target resource; Whether network node inspection this locality exists target resource, if Y, there is target resource in this locality, directly returns to the source network node resource response message; If N, there is not target resource in this locality, using resource request message as new information, joins the present networks node;
Step 3, the life span of judgement request message; Whether the life span TTL that judges request message is 0, if Y, search procedure finishes; If N, be forwarded to request message the neighbor networks node;
Step 4, judge in the Q table, whether the Q value about target resource is empty; At first network node checks present networks node Q table, if be judged as N, the Q value of target resource is arranged, and selects the neighbor networks node with maximum Q value, and, to its Forward-reques message, the life span ttl value in request message reduces 1 simultaneously; If be judged as Y, there is no the Q value of target resource, network node is checked self prestige table, select network node that credit value is the highest as the neighbor networks node, and, to its Forward-reques message, the life span ttl value in request message reduces 1 simultaneously; At a period of time week after date, according to the number of times of neighbor networks node success downloaded resources, calculate and upgrade credit value;
Step 5, calculate award; Request message, to the neighbor networks node, calculates award, upgrades the Q value of previous dive network node; When calculating reward value, the present invention has replaced original link length with the network node number of degrees, and because the high network node of the number of degrees has higher link number, and the information interaction between other network nodes is extensive.Therefore, preferentially select the neighbor networks node that the number of degrees are high, can obtain higher award.
Step 6, the search target resource; And start in neighbor networks node searching resource, until find target resource or TTL, be 0 end.
As shown in Figure 2, network node has k neighbours to the Q table that above-mentioned each network node is safeguarded, is respectively n
1', n
2', n
3' ... ..n
k', the resource that each neighbor networks node comprises is called r
1, r
2, r
3... r
m, the Q (r in table
m, n
k') be k neighbours n
k' about resource r
mthe Q value, Q value representation network node has the possibility of resource, it is larger that the Q value shows that more greatly network node has the possibility of resource, and when the resource searching request arrives, network node is according to the Q table information Forward-reques message of present networks node, network node is jumped in every forwarding one, to from this hunting action, obtain an award, Q value and award according to new state, feed back to the upper hop network node, upgrade the Q value of upper hop network node, query steps is carried out successively according to same steps as.
The problem of upgrading about reward value Rew, its computing formula is as follows:
Rew?=?D
n′+β(load
n,n′+length
n,n′+U
n,n′)
In above formula, D
n 'for an enhanced signal of Q study, its value be 0 or 100(have target resource, D when network node
n 'value 100, otherwise D
n 'be 0); Load
n, n ': the average load of network node and its neighbor networks node; Length
n, n ': the length that means the neighbor networks node link; U
n, n ': for leaving due to network node or the not availability of the resource that fault causes.β: be the award weighted factor, value is (1,0).
Due to P2P logical topology feature, between network node, forwarding messages carries out between the neighbor networks node, therefore without the length of considering link, in the present invention, this condition is changed into to the number of degrees of network node, because the network node that the number of degrees are high has higher link number, and the information interaction between other network nodes is extensive.Therefore, preferentially select the neighbor networks node that the number of degrees are high, can obtain higher award, the correction formula of calculating about award in the present invention is as follows:
Rew=D
n′+β(load
n,n′+degree
n,n′+U
n,n′)
In search procedure, the Q value of network node is constantly updated according to the state information of next-hop network node, and following formula has provided Q value iteration update method:
Q
n(r,n′)=Q
n(r,n′)+α[Rew+γmax
n′Q
n′(r,n′)-Q
n(r,n′)]
Wherein: α is the study factor, and span is (0,1), and γ is discount factor, and span is (0,1).
Above formula has shown the method that the Q value is upgraded, that is: the maximum Q value of the selection of the network node of new state and certain resource dependency, be multiplied by discount factor γ, then, after deducting former Q value, add reward value Rew, and whole and study factor-alpha multiplies each other, then gets algebraical sum with former Q value.
In the searching method based on Q study, traditional query script is after network node is received resource request, in the situation that present networks node No Assets, request message need to be forwarded to the neighbor networks node, its strategy is to check Q value table, the maximum neighbor networks node of Q value that selection has this resource forwards, and when the Q of part resource value is sky, chooses at random a neighbor networks node Forward-reques.For improving this blindness forwarding strategy, the present invention introduces the network node credit mechanism, prestige has been embodied a concentrated reflection of between network node in history is mutual the overall merit of service satisfaction, and lot of documents is divided into two classes by the trusting relationship between network node: directly trust and recommendation trust.
Trusting relationship below in conjunction with accompanying drawing 3 explanation network nodes, have 4 network nodes to represent 4 users in network in figure, be respectively: A, M, N, B, its network node A directly is connected with network node B, network node M and network node N are intermediate network node, and three paths are arranged between A and B, are respectively: A → B, A → M → B, A → N → B, A → B is that prestige between two network nodes is direct trust so, its computational methods can be expressed as:
Wherein, n be network node A and B a period of time (0, t) the mutual number of times of internal information, Score
a(B) be illustrated in the marking value of the mutual middle A network node of primary information to certain behavior of B network node, value is [1,1].
Recommendation trust refers between A and B and did not carry out directly link so, but a kind of trusting relationship of setting up according to the recommendation in the middle of other.In Fig. 3, A → M → B, A → N → B is A, the recommendation trust between B, A to the recommendation trust of B is:
RR
A(B)=α*(DR
A(M)*DR
M(B))+β(DR
A(N)*DR
N(B))
DR wherein
a(M), DR
m(B), DR
a(N), DR
n(B) be respectively path A M, MB, AN, the direct trust value of NB, α and β are respectively the weight of two sections sub-paths, and meet alpha+beta=1.
From the above, credit value Reputation Value(A, the B of network node A to network node B) be the weighted sum of direct trust value and recommendation trust, computing formula is described as:
Reputation?Value（A，B）=λ*DR
A(B)+(1-λ)*RR
A(B)
Wherein, λ directly trusts coefficient, and value is (0,1).
Claims (1)
1. the non-structural P 2 P resource search method based on machine learning and network node prestige, comprise the steps:
Step 1, arrange life span; When a certain network node is received the resource request message that comprises source network node ID, resource name and life span TTL, the initial value that life span TTL is set is 10;
Step 2, judge whether network node exists target resource; Whether network node inspection this locality exists target resource, if Y, there is target resource in this locality, directly returns to the source network node resource response message; If N, there is not target resource in this locality, using resource request message as new information, joins the present networks node;
Step 3, the life span of judgement request message; Whether the life span TTL that judges request message is 0, if Y, search procedure finishes; If N, be forwarded to request message the neighbor networks node;
Step 4, judge in the Q table, whether the Q value about target resource is empty; At first network node checks present networks node Q table, if be judged as N, the Q value of target resource is arranged, and selects the neighbor networks node with maximum Q value, and, to its Forward-reques message, the life span ttl value in request message reduces 1 simultaneously; If be judged as Y, there is no the Q value of target resource, network node is checked self prestige table, select network node that credit value is the highest as the neighbor networks node, and, to its Forward-reques message, the life span ttl value in request message reduces 1 simultaneously; At a period of time week after date, according to the number of times of neighbor networks node success downloaded resources, calculate and upgrade credit value;
Step 5, calculate award; Request message, to the neighbor networks node, calculates award, upgrades the Q value of previous dive network node;
Step 6, the search target resource; And start in neighbor networks node searching resource, until find target resource or TTL, be 0 end.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310479076.2A CN103501346B (en) | 2013-10-14 | 2013-10-14 | A kind of non-structural P 2 P resource search method based on machine learning and network node prestige |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310479076.2A CN103501346B (en) | 2013-10-14 | 2013-10-14 | A kind of non-structural P 2 P resource search method based on machine learning and network node prestige |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103501346A true CN103501346A (en) | 2014-01-08 |
| CN103501346B CN103501346B (en) | 2016-08-10 |
Family
ID=49866516
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310479076.2A Active CN103501346B (en) | 2013-10-14 | 2013-10-14 | A kind of non-structural P 2 P resource search method based on machine learning and network node prestige |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103501346B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104219148A (en) * | 2014-09-25 | 2014-12-17 | 四川大学 | Game theory normalization based generous dynamic routing selection method |
| CN109547351A (en) * | 2019-01-22 | 2019-03-29 | 西安电子科技大学 | Method for routing based on Q study and trust model in Ad Hoc network |
| CN110049530A (en) * | 2019-05-24 | 2019-07-23 | 西安电子科技大学 | Method for routing based on Q study and trust model in vehicular ad hoc network |
| CN113507738A (en) * | 2021-06-29 | 2021-10-15 | 成都金融梦工场投资管理有限公司 | Routing decision method for mobile ad hoc network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1941719A (en) * | 2006-09-27 | 2007-04-04 | 北京航空航天大学 | Method for managing network resource |
| CN101702675A (en) * | 2009-11-20 | 2010-05-05 | 西安电子科技大学 | System for managing P2P network security and trust based on path optimization and finding |
| CN101895474A (en) * | 2010-07-29 | 2010-11-24 | 江苏大学 | Information diffusion-based P2P network route control method |
| WO2011133209A2 (en) * | 2010-04-23 | 2011-10-27 | Thomson Licensing | Method and system for providing recommendations in a social network |
-
2013
- 2013-10-14 CN CN201310479076.2A patent/CN103501346B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1941719A (en) * | 2006-09-27 | 2007-04-04 | 北京航空航天大学 | Method for managing network resource |
| CN101702675A (en) * | 2009-11-20 | 2010-05-05 | 西安电子科技大学 | System for managing P2P network security and trust based on path optimization and finding |
| WO2011133209A2 (en) * | 2010-04-23 | 2011-10-27 | Thomson Licensing | Method and system for providing recommendations in a social network |
| CN101895474A (en) * | 2010-07-29 | 2010-11-24 | 江苏大学 | Information diffusion-based P2P network route control method |
Non-Patent Citations (1)
| Title |
|---|
| WEINA LIU等: "A reputation-based market model in grid environment", 《FIRST INTERNATIONAL SYMPOSIUM ON DATA, PRIVACY AND E-COMMERCE,ISDPE2007》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104219148A (en) * | 2014-09-25 | 2014-12-17 | 四川大学 | Game theory normalization based generous dynamic routing selection method |
| CN104219148B (en) * | 2014-09-25 | 2017-06-23 | 四川大学 | Based on the normalized generous dynamics route selection method of theory of games |
| CN109547351A (en) * | 2019-01-22 | 2019-03-29 | 西安电子科技大学 | Method for routing based on Q study and trust model in Ad Hoc network |
| CN110049530A (en) * | 2019-05-24 | 2019-07-23 | 西安电子科技大学 | Method for routing based on Q study and trust model in vehicular ad hoc network |
| CN113507738A (en) * | 2021-06-29 | 2021-10-15 | 成都金融梦工场投资管理有限公司 | Routing decision method for mobile ad hoc network |
| CN113507738B (en) * | 2021-06-29 | 2023-12-12 | 成都金融梦工场投资管理有限公司 | Routing decision method for mobile ad hoc network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103501346B (en) | 2016-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| He et al. | Green resource allocation based on deep reinforcement learning in content-centric IoT | |
| CN105049326B (en) | Social content caching method in a kind of edge network region | |
| Chen et al. | Edge intelligent networking optimization for internet of things in smart city | |
| CN103501346A (en) | Non-structured P2P (Peer-to-Peer) resource searching method based on machine learning and network node reputation | |
| Napoli et al. | Improving files availability for bittorrent using a diffusion model | |
| GB2557674A (en) | Automated decision making apparatus and methods | |
| Rolim et al. | Situation awareness and computational intelligence in opportunistic networks to support the data transmission of urban sensing applications | |
| CN115066694A (en) | Computation graph optimization | |
| Malektaji et al. | Deep reinforcement learning-based content migration for edge content delivery networks with vehicular nodes | |
| Karami et al. | RPRTD: Routing protocol based on remaining time to encounter nodes with destination node in delay tolerant network using artificial neural network | |
| Zhang et al. | Collaborative task offloading with computation result reusing for mobile edge computing | |
| Li et al. | An incentive mechanism for message relaying in unstructured peer-to-peer systems | |
| CN105243080A (en) | ESB (Enterprise Service Bus) framework based resource perception adaptive system | |
| Rui et al. | Location‐dependent sensing data collection and processing mechanism in vehicular network | |
| US8996427B2 (en) | Method for a system that solves multimedia resource discovery and retrieval problem sets | |
| CN116582407A (en) | Containerized micro-service arrangement system and method based on deep reinforcement learning | |
| Nguyen et al. | Multi-agent task assignment in vehicular edge computing: A regret-matching learning-based approach | |
| Noghabi et al. | Optimized query forwarding for resource discovery in unstructured peer-to-peer grids | |
| Wu et al. | PECCO: A profit and cost‐oriented computation offloading scheme in edge‐cloud environment with improved Moth‐flame optimization | |
| Cai et al. | Cooperative content caching and delivery in vehicular networks: A deep neural network approach | |
| Lamaazi et al. | Trust-3DM: Trustworthiness-Based Data-Driven Decision-Making Framework Using Smart Edge Computing for Continuous Sensing | |
| Tang et al. | Multi-User Layer-Aware Online Container Migration in Edge-Assisted Vehicular Networks | |
| Liu et al. | Product search algorithm based on improved ant colony optimization in a distributed network | |
| Wang et al. | Bayesian Approach with Maximum Entropy Principle for trusted quality of Web service metric in e‐commerce applications | |
| KR20220055363A (en) | A method for controlling a state control parameter for adjusting a state of a network of a base station using one of a plurality of models and an electronic device performing the same |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |