CN110009113B - Internet of things equipment autonomous learning method, device, equipment and storage medium - Google Patents
Internet of things equipment autonomous learning method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110009113B CN110009113B CN201810001639.XA CN201810001639A CN110009113B CN 110009113 B CN110009113 B CN 110009113B CN 201810001639 A CN201810001639 A CN 201810001639A CN 110009113 B CN110009113 B CN 110009113B
- Authority
- CN
- China
- Prior art keywords
- equipment
- node
- data
- internet
- things
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The embodiment of the invention discloses an autonomous learning method, an autonomous learning device, an autonomous learning equipment and a storage medium of Internet of things equipment, wherein the method comprises the following steps: determining first equipment meeting a first preset condition according to the attribute of second equipment, wherein the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is the existing terminal equipment in the Internet of things; acquiring first device data X of the first device and second device data Y of the second device which meet a second preset condition; acquiring an initial labeling set S for labeling each piece of data in first equipment data X of the first equipment; and determining a training model phi' of the second equipment according to the initial labeling set S and the second equipment data Y.
Description
Technical Field
The invention relates to the field of machine intelligence, in particular to an autonomous learning method, an autonomous learning device, an autonomous learning equipment and a storage medium of Internet of things equipment.
Background
The development of the internet of things is on an accelerating trend, and according to the prediction of many international consulting companies, 250 hundred million connected internet of things devices are predicted globally in 2020. Due to the mobility of the equipment, the real-time performance of the service and the like, the equipment of the internet of things is in a highly dynamic changing environment, and new equipment is added into the environment of the internet of things at any time. The newly added equipment often has the problem of lack of training data or training samples, the traditional solution is to manually re-mark new equipment data, the re-marking process of the data is time-consuming and labor-consuming, and due to the need of manual intervention, the intelligent degree of the whole system is very low, and the intelligent requirement in the era of the internet of things cannot be met.
Disclosure of Invention
In view of this, embodiments of the present invention provide an autonomous learning method, apparatus, device, and storage medium for internet of things devices to solve at least one problem in the prior art, so that when a new device is added to an internet of things system, relearning without manual intervention can be implemented, thereby improving the intelligence degree of the device.
The technical scheme of the embodiment of the invention is realized as follows:
the embodiment of the invention provides an autonomous learning method for Internet of things equipment, which comprises the following steps:
determining first equipment meeting a first preset condition according to the attribute of second equipment, wherein the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is the existing terminal equipment in the Internet of things;
acquiring first device data X of the first device and second device data Y of the second device which meet a second preset condition;
acquiring an initial labeling set S for labeling each piece of data in first equipment data X of the first equipment;
and determining a training model phi' of the second equipment according to the initial labeling set S and the second equipment data Y.
The embodiment of the invention provides an autonomous learning device for Internet of things equipment, which comprises:
the device comprises a first determining unit, a second determining unit and a third determining unit, wherein the first determining unit is used for determining first equipment meeting a first preset condition according to the attribute of second equipment, the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is the existing terminal equipment in the Internet of things;
a first obtaining unit, configured to obtain first device data X of the first device and second device data Y of the second device that satisfy a second preset condition;
a second obtaining unit, configured to obtain an initial labeling set S ═ S [ S ] for labeling each piece of data in the first device data X of the first device1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, wherein
And the second determining unit is used for determining a training model phi' of the second equipment according to the initial labeling set S and the second equipment data Y.
The embodiment of the invention provides an autonomous learning device of equipment of the Internet of things, which comprises a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor executes the program to realize the steps in the autonomous learning method of the equipment of the Internet of things.
An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps in the method for autonomous learning of internet of things devices.
In the embodiment of the invention, first equipment meeting a first preset condition is determined according to the attribute of second equipment, the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is the existing terminal equipment in the Internet of things; acquiring first device data X of the first device and second device data Y of the second device which meet a second preset condition; acquiring an initial labeling set S ═ S [ S ] for labeling each piece of data in first device data X of the first device1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, wherein
Determining a training model phi' of the second equipment according to the initial labeling set S and the second equipment data Y; therefore, when new equipment is added into the Internet of things system, relearning without manual intervention can be realized, and the intelligent degree of the equipment is improved.
Drawings
Fig. 1 is a schematic flow chart illustrating an implementation process of an autonomous learning method of internet of things equipment according to an embodiment of the invention;
fig. 2 is a schematic flow chart illustrating an implementation process of a method for autonomous learning by an internet of things device according to another embodiment of the present invention;
fig. 3 is a schematic diagram of autonomous learning of an internet of things device according to an embodiment of the present invention;
FIG. 4 is a diagram of an undirected graph model in accordance with an embodiment of the present invention;
FIG. 5 is a schematic diagram of an implementation flow of automatic label updating according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a component of an autonomous learning apparatus of an internet of things device according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a hardware entity of the terminal device of the internet of things according to the embodiment of the invention.
Detailed Description
With the development of society, the number of terminal devices in the internet of things is increasing, and due to the mobility of the devices, the real-time performance of services and the like, the devices in the internet of things are in a highly dynamic changing environment, and new devices are added into the environment of the internet of things at any time. The traditional solution is to manually re-mark new device data, wherein marking is a process of classifying the training data or the training samples to obtain a classification result. Because of the new internet of things equipment, the equipment data of the new internet of things equipment needs to be marked, the data re-marking process is time-consuming and labor-consuming, and the intelligent degree of the whole system is very low due to the need of manual intervention, so that the intelligent requirement in the internet of things era can not be met. Therefore, an autonomous learning algorithm of the Internet of things equipment is designed to enable the Internet of things equipment to have autonomous learning capability. When the environment changes, namely new equipment with the same function is added into the system, the relearning without manual intervention is realized, and the method has important significance for improving the intelligent degree of the equipment.
The existing method for automatically learning the Internet of things equipment is to utilize a method for supervising learning in machine learning to classify test data by learning the training data on the basis of marked training data, the method is only suitable for the condition that the data are marked, and the Internet of things system is always in high dynamic change due to the practical conditions of network connection, power consumption, system updating and the like of the Internet of things equipment, namely new and untrained equipment is possibly added into the system at any time, but the equipment does not have identification capability due to the fact that the equipment is not trained, namely marked data, aiming at the condition, the traditional solution is to manually re-mark new equipment data, the re-marking of the data consumes time and labor, and the intelligence degree of the whole system is very low due to the need of manual intervention, the intelligent learning method can not meet the requirement of the intelligence of the internet of things era, and no self-learning method which is simple and convenient and is suitable for the internet of things equipment exists at present.
The technical solution of the present invention is further elaborated below with reference to the drawings and the embodiments.
In this embodiment, the functions implemented by the method may be implemented by calling a program code through a processor in the internet of things terminal device, and the program code may be stored in a computer storage medium.
Fig. 1 is a schematic view of an implementation flow of an autonomous learning method of an internet of things device in an embodiment of the present invention, and as shown in fig. 1, the method includes:
step S101, determining first equipment meeting a first preset condition according to attributes of second equipment, wherein the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is existing terminal equipment in the Internet of things;
the attribute of the device includes a function and a category implemented by the device, and the first preset condition is that the attribute is the same or the attribute is similar, for example, if the functions of the two devices are the same or similar, the two devices are considered to satisfy the first preset condition.
Step S102, acquiring first device data X of the first device and second device data Y of the second device which meet a second preset condition;
step S103, acquiring first device data of the first deviceAnd (4) marking each piece of data in X by an initial marking set S ═ S1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, wherein
n represents the total number of categories;
and step S104, determining a training model phi' of the second equipment according to the initial labeling set S and the second equipment data Y.
From the above, it can be seen that the model Φ of the device a is transmitted to the device B through synchronization for a period of time, so as to obtain the model Φ ', and the device B has the working capability by using the model Φ', thereby realizing the transmission and improvement of the machine learning capability between the devices and completing the improvement of the intelligent degree of the whole internet of things system.
In the implementation process, the main execution body of the method is a terminal newly accessed to the internet of things, for example, the newly accessed terminal is a device B, and all or part of the method may be run on the device B. However, it should be noted that, if all the information is executed on the device B, a preset relationship list needs to be imported into the device B (the second device), where the relationship list is used to indicate a mapping relationship between device attributes and devices (e.g., device identifiers) that have access to the internet of things, so that the device B may determine the device a (the first device) that has access to the internet of things through step S101. Of course, step S101 may also be performed by a network device in the internet of things, such as a server or a gateway of the internet of things. The first device data and the initial annotation set S in steps S102 and S103 may be a request from device B to device a or a network device.
In the above embodiment, determining the training model Φ' of the second device according to the initial labeling set S and the second device data Y includes:
step S11, constructing an undirected graph according to the second device data Y, wherein each piece of data Y in the second device data YjCorresponding to the node v in the undirected graphjNode v in the undirected graphjThe similarity between the two is corresponding to the edge in the undirected graph;
step S12, determining each node v in the undirected graphjDetermining a reliable node set Ψ in the initial labeling set S according to the eccentricity;
wherein the determining each node v in the undirected graphjComprises: for each class CiBy the formula
Determining a node vjWherein node vjEccentricity of (d) indicates the node point vjAnd class CiMaximum of distances of all nodes in the data YjRepresenting a node vj，YkAny piece of data representing the second device data Y, max representing taking the maximum value, and d representing the calculated distance;
correspondingly, determining a reliable node set Ψ in the initial labeling set S according to the eccentricity, including: by the formula
Calculation class CiMost reliable node R inCiWherein, class CiMost reliable node R inCiIs in the same class as class CiThe node with the minimum node eccentricity in the node list is min, and the minimum value is taken; the most reliable node of all classes is determined as the set Ψ of reliable nodes.
Step S13, determining an updated label set RL according to the reliable node set Ψ and the initial label set;
wherein, the completing the updating of the initial label set according to the reliable node set Ψ to obtain an updated label set RL includes:
by passing
Updating the initial label set S to obtain an updated label set RL ═ RL1,rl2,…,rln]Wherein the symbol
The representation does not belong to, and the symbol e represents belonging to.
And step S14, training the second equipment data Y according to the updated label set RL, and obtaining a training model phi' of the second equipment.
The determining an updated label set RL according to the reliable node set Ψ and the initial label set comprises:
by passing
Initializing RL to obtain initialized label set RL ═ RL1,rl2,…,rln]Wherein the symbol
The symbol belongs to a symbol;
and according to the reliable node set psi, completing the updating of the initialized labeling set RL with the lowest traversal cost through a minimum spanning tree algorithm to obtain an updated labeling set RL.
Wherein, according to the reliable node set Ψ, the updating of the initialized label set RL is completed with the lowest traversal cost through the minimum spanning tree algorithm, so as to obtain an updated label set RL, and the updating includes:
step S21, selecting a node v with rl as zerop；
Step S22, Slave node vpStarting, selecting the side with the maximum weight in the undirected graph G for diffusion according to the undirected graph G;
step S23, determining whether rl of another node adjacent to the edge with the largest weight is zero, if so, continuing to the slave node vpStarting, selecting the side with the maximum side weight for path diffusion according to the undirected graph G; if not, determining the non-zero node as an end node, and ending the traversal process;
step S24, setting rl of all nodes on the traversed diffusion path as rl of the end node, and stopping until rl of all nodes is not zero.
In other embodiments, the method further comprises: determining the similarity between nodes in the undirected graph as the weight of the edge in the undirected graph, and assuming that a node v existsiAnd node vjBy the formula
Calculating similarity η (v) between nodesi,vj)。
In this embodiment of the present invention, the obtaining an initial labeling set S for labeling each piece of data in first device data X of the first device includes:
labeling each piece of data in first equipment data X of the first equipment by using a training model phi of the first equipment to obtain an initial labeling set S ═ S1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, wherein
A set of categories C.
The invention provides a method for autonomously learning Internet of things equipment, which is used for systematically and integrally researching the autonomous learning of the Internet of things equipment through a probabilistic undirected graph and a minimum spanning tree algorithm and aims to provide the method for autonomously learning the Internet of things equipment, and comprises the following steps:
firstly, working the trained equipment A with the same function as the new equipment B for a period of time t, and labeling the equipment data (data expressed as A + B) of the equipment A and the equipment B acquired in the period of time t according to the model phi because the equipment A has working capacity, namely a trained model phi, so as to obtain a labeled soft _ label;
note that, the initial label set S corresponding to soft _ label is labeled, and any label in the initial label set S is labeled as si.
Secondly, constructing a probability undirected graph according to the data of the equipment B, wherein the data of the equipment B corresponds to nodes in the undirected graph, and the similarity between the nodes corresponds to edges in the undirected graph;
thirdly, proposing a concept of node eccentricity for each node, namely calculating the eccentricity of each node, and determining the most reliable node set Ψ in soft _ label according to the eccentricity;
thirdly, according to the obtained reliable node set psi, updating the soft _ label with the lowest traversal cost by using a minimum spanning tree algorithm for reference;
and finally, training the data of the equipment B according to the updated soft _ label result to obtain a model phi ' so as to finish the transfer of the learning capacity from the equipment A to the equipment B, namely, transferring the model phi of the equipment A to the equipment B through synchronization for a period of time to obtain the model phi ' of the equipment B, and utilizing the working capacity of the model phi ' of the equipment B, thereby realizing the transfer and improvement of the machine learning capacity among the equipment and finishing the improvement of the intelligent degree of the whole Internet of things system.
An embodiment of the present invention provides an autonomous learning method for an internet of things device, and fig. 2 is a schematic view of an implementation flow of the autonomous learning method for the internet of things device according to another embodiment of the present invention, as shown in fig. 2, the method includes the following steps:
step S1, setting a device synchronization mechanism, realizing the initial transfer of labels among devices, and obtaining an initial label set S;
here, in the implementation process, the device synchronization mechanism may be determined according to the attribute information of the device, where the attribute information of the device includes a function implemented by the device, a category to which the device belongs, and the like; for example, if at least one of the functions and the categories of the two devices has the same attribute, the two devices in the device synchronization mechanism may be set, for example, the device a is a temperature sensor, if the device B is also a temperature sensor, the device a and the device B belong to the same category or the implemented devices have the same function, the device a and the device B may implement data synchronization, and if the device a is a trained device, the device a has marked data and an existing trained model Φ; if device B is a new device, the model of device B can be trained using the labeled data of device A and the existing trained model Φ.
Step S2, constructing an undirected graph model G according to the data of the newly added equipment B;
step S3, calculating the eccentricity of each node in the undirected graph model G, and determining the most reliable node set Ψ according to the eccentricity;
step S4, updating the initial label by using a minimum spanning tree algorithm for reference according to the reliable node set psi and the initial label set S to obtain a refined _ label;
it should be noted that, the initial label set RL corresponding to the label defined _ label, any label in the initial label set RL is labeled rli,
step S5, according to the refined _ label result, training the data of the equipment B to obtain a model phi';
and step S6, realizing the construction of the capability of the equipment B according to the model phi' of the equipment B.
In step S1, it is assumed that the ith data X of the apparatus a is acquiredi：Xi＝{xi1,xi2,…,xik}; manually marking each piece of data (namely marking process) to obtain a mark
The set of all labels is S ═ S1,s2,…,st]Wherein, in the step (A),
the symbol represents the ith data, and the ith data may be data that can be acquired at time i, CiClass represents class, i.e. classification result, CiMay be any one of the class sets C ═ { C1, C2, …, Cn }, siDenotes s1To stI is an integer of 1 to t inclusive;by comparing the above data XiAnd training with all labeled sets S to obtain the model phi. The function F can be realized by using the model phi, and when the equipment B is newly added into the Internet of things system, the equipment B can only acquire the ith data Y of the equipment Bi＝{yi1,yi2,…,yimBut does not have any functionality, i.e. no annotation on the acquired data.
Setting a device synchronization mechanism, namely enabling the device A and the device B to coexist in the Internet of things system for a period of time t, and obtaining data Hi＝{Xi,Yi}＝{xi1,xi2,…,xik,yi1,yi2,…,yim1,2, … t, since device a can implement function F using model Φ, for new data HiA part of data X in (2)iAccording to phi (X)i) The original annotation, i.e. s, of the data can be obtainedi。
Fig. 3 is a schematic diagram of autonomous learning of internet-of-things devices in the embodiment of the present invention, assuming that a device a in fig. 3 is a motion sensing sensor, a device B is also a motion sensing sensor, and the device a is a trained device and has marked data and an existing trained model Φ; the device B is a new device, and the model of the device B can be trained using the marked data of the device a and the existing trained model Φ.
In step S2, the new data H is usediAnother part of the data in (1), that is, the data Y acquired by the newly joining device BiAnd constructing an undirected graph model. Let us assume that an undirected graph is denoted by G ═ V, E, where V denotes a set of nodes V ═ V in the undirected graph1,v2,…,vn]Data YiCorresponding to node v in the undirected graphiAccording to the synchronization mechanism of step S1, the node V in the node set ViWith an initial label siAnd E represents the edge set in the undirected graph, and the similarity between nodes corresponds to the weight of the edges in the undirected graph. Suppose there is a node viAnd node vjThe similarity η (v) between nodesi,vj) Calculated from the following formula (1):
two nodes v in equation (1)iAnd node vjWhere i is not equal to j.
FIG. 4 is a diagram of an undirected graph model according to an embodiment of the invention, where v is shown in FIG. 41To v6Representing a set of nodes V, the connecting line between two nodes representing an edge, V1To v6The set of edges between any two nodes is E, and the edges between two nodes have no direction, and therefore are called undirected graphs.
In step S3, the concept of node eccentricity is proposed, and each class C is assignediNode vjEccentricity of (d) indicates the node point vjAnd class CiThe maximum value of the distances of all nodes in the node is represented by the following formula (2):
in the formula (2), YjThe data representing the jth device B is the node vj，YkRepresents class CiAny one of all nodes in the network.
For each class CiCan acquire the most reliable node in the category, category CiMost reliable node R inCiIs in the same class as class CiThe node having the smallest node eccentricity is represented by the following formula (3):
therefore, the most reliable node set Ψ for all classes is shown in equation (4) below:
in step S4, the method includes two steps, one is a process of initializing the label set RL according to the reliable node set Ψ and the initial label set S, and the other is a process of updating the label set RL by using the minimum spanning tree prim algorithm for reference.
The process of initializing the label set RL according to the reliable node set Ψ and the initial label set S includes: initializing an label set RL according to the reliable node set Ψ and the initial label set S to obtain a label set RL ═ RL1,rl2,…,rln]The following procedure (5) was followed:
wherein, for reference of the minimum spanning tree prim algorithm, the mechanism for updating the label set RL is that:
step 1) randomly selecting a node with rl being zero;
step 2) starting from the node, selecting the side with the maximum side weight for diffusion according to the undirected graph G constructed in the step S2;
and 3) judging whether rl of another node adjacent to the edge with the maximum weight is zero, if so, returning to the step 2) to continue traversing, if not, marking the node which is not zero as an end node, ending the traversing process, setting all the traversed nodes as the rl of the end node, and stopping until all the traversed nodes are not zero.
From the above, it can be seen that the label set RL is updated by using the minimum spanning tree prim algorithm, that is, starting from a node with RL being zero, an edge with the largest edge weight is selected in the undirected graph G for path diffusion (this path is called a diffusion path), and ends at a node with RL being not zero, and RL of the end node (i.e. the node with RL being not zero) is assigned to all nodes on the diffusion path.
Fig. 5 is a schematic diagram of an implementation process of automatically updating annotations according to an embodiment of the present invention, and as shown in fig. 5, the process includes:
step S501, obtain V ═ V1, V2, …, vn ];
wherein V ═ V1, V2, …, vn is obtained]I.e. obtaining the set of nodes V ═ V1, V2, …, vn in the undirected graph](ii) a Data Y of device BiCorresponding to node v in the undirected graphi；
Step S502, determining a zero element set in the RL set as zero _ set;
step S503, determine whether the zero element set zero _ set is empty? If yes, go to step S512, if no, go to step S504;
step S504, randomly selecting element v in zero _ seti；
In step S505, the minimum spanning tree _ set is [ v ═ vi]，U＝V-tree_set；
The set U is a complement of the minimum spanning tree set, and the node set V is a full set.
Step S506, determine whether the element in U is empty? If yes, go to step S502; if not, go to step S507;
step S507, selecting the element v most similar to the element v in tree _ set in Uj；
Here, the element v in U most similar to that in tree _ set is selected according to the similarityjSuppose there is a node viAnd node vjThe similarity η (v) between nodesi,vj) Calculated from the following formula (1):
two nodes v in equation (1)iAnd node vjWhere i is not equal to j.
Step S508, judging vjIs there present? If yes, go to step S509; if not, go to step S502;
step S509, judging rljIs it 0? If yes, go to step S510; if not, go to step S511;
wherein v isjIs of the class rljIf rl isjIf it is zero, the process proceeds to step S510.
Step S510, tree _ set [ tree _ set, v ═ v-j]，U＝V-tree_set；
Step S511, tree _ set [ tree _ set, v ]j]All rls of nodes in tree _ set are updated to rlj；
And step S512, ending.
In step S5, training data of the device B is implemented according to the result in the RL, and a corresponding model Φ' is obtained according to the features of the respective algorithms by selecting a naive bayes algorithm, an SVM algorithm, a kNN algorithm, and other classical classification algorithms in the training process;
in step S6, the capability of the device B is constructed according to the model Φ' of the device B, that is, when the device B newly collects a set of data Zi＝{zi1,zi2,…,zimAt this time, the model Φ' (Z) is usedi) The class label of the new data can be obtained, that is, the transmission of the identification capability of the device a to the identification capability of the device B is realized, and under the condition that the device a is not provided, the function similar to that of the device a can be realized only by the device B.
In the embodiment, by using a graph theory and a minimum spanning tree algorithm for reference, the autonomous learning of the internet of things equipment is systematically and integrally researched, and aiming at the problem that newly added equipment in an internet of things scene is lack of training data, a synchronous autonomous learning method of the internet of things equipment is provided, firstly, the equipment A which has the same function as the new equipment B and is trained together works for a period of time t, and as the equipment A has the working capacity, namely a trained model phi exists, the data of A + B acquired in the period of time t is labeled according to the model phi to obtain a label S; secondly, according to the data of the B device, constructing an undirected graph, wherein the data of the B device corresponds to nodes in the graph, and the similarity between the nodes corresponds to edges in the graph; thirdly, a concept of labeling eccentricity is provided for each node, and the most reliable node set psi in the S is determined according to the eccentricity; thirdly, according to the obtained node set, updating the S with the lowest traversal cost by using a minimum spanning tree algorithm for reference; and finally, training the data of the equipment B according to the updated RL result to obtain a model phi', thereby completing the transfer of the learning capacity from the equipment A to the equipment B and enabling the equipment B to have the working capacity. The method is an autonomous learning method which is realized without the help of manpower, can effectively improve the intelligent degree of the whole Internet of things system, and is particularly important for unattended Internet of things equipment with highly dynamic environment changes.
Compared with the prior art, the embodiment has the following advantages: the invention related to the existing method for autonomously learning the Internet of things equipment describes how the Internet of things equipment autonomously learns the preference of a user according to the habit and record of the user using the equipment, for example, an air conditioner autonomously learns the optimum temperature of the user, an electric cooker autonomously learns the diet preference of the user, and the intelligent control of the Internet of things equipment is realized according to the autonomous learning result. Firstly, the invention provides a synchronous Internet of things equipment autonomous learning method aiming at the problem that newly added equipment in an Internet of things scene is lack of training data, the method is based on graph theory and minimum spanning tree algorithm, two Internet of things equipment A and B with the same identification function are considered, wherein the equipment A is trained, namely the equipment A has marking training data, the other equipment B is not trained, if a user plans to abandon the trained equipment A and start the new equipment B, the method can provide the Internet of things equipment autonomous learning method, the identification capability of the equipment A can be transferred to the equipment B after the equipment A and the equipment B are used simultaneously for a short period of time, namely the equipment B has identification capability similar to the equipment A, the method is an autonomous learning method which is realized without the help of manpower, and the method can be used for unattended equipment autonomous learning, The internet of things equipment with highly dynamic environment change is particularly important, the intelligent degree of the internet of things equipment can be improved, and the invention focuses on an autonomous learning mechanism rather than sensing, learning and controlling the whole process.
Based on the foregoing embodiments, an embodiment of the present invention provides an autonomous learning apparatus for internet of things devices, where the apparatus includes units and modules included in the units, and can be implemented by a processor in an internet of things terminal device; of course, the implementation can also be realized through a specific logic circuit; in implementation, the processor may be a Central Processing Unit (CPU), a Microprocessor (MPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), or the like.
Fig. 6 is a schematic structural diagram of an autonomous learning apparatus of an internet of things device according to an embodiment of the present invention, and as shown in fig. 6, the apparatus 600 includes a first determining unit 601, a first obtaining unit 602, a second obtaining unit 603, and a second determining unit 604, where:
a first determining unit 601, configured to determine, according to an attribute of a second device, a first device that meets a first preset condition, where the second device is a terminal device that newly joins the internet of things, and the first device is an existing terminal device in the internet of things;
a first obtaining unit 602, configured to obtain first device data X of the first device and second device data Y of the second device, where a second preset condition is met;
a second obtaining unit 603, configured to obtain an initial labeling set S ═ S for labeling each piece of data in the first device data X of the first device1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, wherein
A second determining unit 604, configured to determine a training model Φ' of the second device according to the initial labeling set S and the second device data Y.
The second determining unit 604 includes:
a construction module, configured to construct an undirected graph according to the second device data Y, where each piece of the second device data Y isjCorresponding to the node v in the undirected graphjNode v in the undirected graphjThe similarity between the two is corresponding to the edge in the undirected graph;
first determining moduleFor determining each node v in said undirected graphjDetermining a reliable node set Ψ in the initial labeling set S according to the eccentricity;
a second determining module, configured to determine an updated label set RL according to the reliable node set Ψ and the initial label set;
and the training module is used for training the second equipment data Y according to the updated label set RL to obtain a training model phi' of the second equipment.
Wherein the determining each node v in the undirected graphjComprises: for each class CiBy the formula
Determining a node vjWherein node vjEccentricity of (d) indicates the node point vjAnd class CiMaximum of distances of all nodes in the data YjRepresenting a node vj，YkAny piece of data representing the second device data Y, max representing taking the maximum value, and d representing the calculated distance; correspondingly, determining a reliable node set Ψ in the initial labeling set S according to the eccentricity, including: by the formula
Calculation class CiMost reliable node R inCiWherein, class CiMost reliable node R inCiIs in the same class as class CiThe node with the minimum node eccentricity in the node list is min, and the minimum value is taken; the most reliable node of all classes is determined as the set Ψ of reliable nodes.
The above description of the apparatus embodiments, similar to the above description of the method embodiments, has similar beneficial effects as the method embodiments. For technical details not disclosed in the embodiments of the apparatus according to the invention, reference is made to the description of the embodiments of the method according to the invention for understanding.
It should be noted that, in the embodiment of the present invention, if the above-mentioned method for autonomously learning internet of things is implemented in the form of a software functional module, and is sold or used as an independent product, it may also be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an internet of things terminal device to execute all or part of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read Only Memory (ROM), a magnetic disk, or an optical disk. Thus, embodiments of the invention are not limited to any specific combination of hardware and software.
Correspondingly, an embodiment of the present invention provides an autonomous learning device for an internet of things device (an internet of things terminal device), which includes a memory and a processor, where the memory stores a computer program operable on the processor, and the processor implements the steps in the autonomous learning method for the internet of things device when executing the program.
Correspondingly, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps in the method for autonomous learning of an internet of things device described above.
Here, it should be noted that: the above description of the storage medium and device embodiments is similar to the description of the method embodiments above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and the apparatus according to the invention, reference is made to the description of the embodiments of the method according to the invention.
It should be noted that fig. 7 is a schematic diagram of a hardware entity of an internet of things terminal device according to an embodiment of the present invention, and as shown in fig. 7, the hardware entity of the internet of things terminal device 700 includes: a processor 701, a communication interface 702, and a memory 703, wherein
The processor 701 generally controls the overall operation of the internet of things terminal device 700.
The communication interface 702 may enable the internet of things terminal device to communicate with other terminals or servers through a network.
The Memory 703 is configured to store instructions and applications executable by the processor 701, and may also buffer data (for example, image data, audio data, voice communication data, and video communication data) to be processed or already processed by each module in the to-be-processed processor 701 and the terminal device 700 of the internet of things, and may be implemented by a FLASH Memory (FLASH) or a Random Access Memory (RAM).
It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present invention, the sequence numbers of the above-mentioned processes do not mean the execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above-described device embodiments are merely illustrative, for example, the division of the unit is only a logical functional division, and there may be other division ways in actual implementation, such as: multiple units or components may be combined, or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the coupling, direct coupling or communication connection between the components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between the devices or units may be electrical, mechanical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units; can be located in one place or distributed on a plurality of network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, all the functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately regarded as one unit, or two or more units may be integrated into one unit; the integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.
Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium includes: various media that can store program codes, such as a removable Memory device, a Read Only Memory (ROM), a magnetic disk, or an optical disk.
Alternatively, the integrated unit of the present invention may be stored in a computer-readable storage medium if it is implemented in the form of a software functional module and sold or used as a separate product. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an internet of things terminal device to execute all or part of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, a ROM, a magnetic or optical disk, or other various media that can store program code.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the present invention, and all such changes or substitutions are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (9)
1. An autonomous learning method for Internet of things equipment is characterized by comprising the following steps:
determining first equipment meeting a first preset condition according to the attribute of second equipment, wherein the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is the existing terminal equipment in the Internet of things;
acquiring first device data X of the first device and second device data Y of the second device which meet a second preset condition;
acquiring an initial labeling set S for labeling each piece of data in first equipment data X of the first equipment;
constructing an undirected graph according to the second equipment data Y, wherein each piece of the second equipment data YjCorresponding to the node v in the undirected graphjNode v in the undirected graphjThe similarity between the two is corresponding to the edge in the undirected graph;
determining each node v in the undirected graphjDetermining a reliable node set Ψ in the initial labeling set S according to the eccentricity;
determining an updated label set RL according to the reliable node set Ψ and the initial label set;
and training the second equipment data Y according to the updated label set RL to obtain a training model phi' of the second equipment.
2. The method according to claim 1, wherein said determining each node v in said undirected graphjComprises:
for each class CiBy the formula
Determining a node vjWherein node vjEccentricity of (d) indicates the node point vjAnd class CiMaximum of distances of all nodes in the data YjRepresenting a node vj，YkAny piece of data representing the second device data Y, max representing taking the maximum value, and d representing the calculated distance;
correspondingly, determining a reliable node set Ψ in the initial labeling set S according to the eccentricity, including: by the formula
Calculation class CiMost reliable node R inCiWherein, class CiMost reliable node R inCiIs in the same class as class CiThe node with the minimum node eccentricity in the node list is min, and the minimum value is taken;
the most reliable node of all classes is determined as the set Ψ of reliable nodes.
3. The method of claim 1, wherein the determining an updated label set RL from the reliable node set Ψ and the initial label set comprises:
by passing
Initializing RL to obtain initialized label set RL ═ RL1,rl2,…,rln]Wherein the symbol
The symbol belongs to a symbol;
and according to the reliable node set psi, completing the updating of the initialized labeling set RL with the lowest traversal cost through a minimum spanning tree algorithm to obtain an updated labeling set RL.
4. The method according to claim 3, wherein said updating the initialized label set RL with the lowest traversal cost according to the reliable node set Ψ by a minimum spanning tree algorithm to obtain an updated label set R comprises:
selecting a node v with rl of zerop；
Slave node vpStarting, selecting the side with the maximum weight in the undirected graph G for diffusion according to the undirected graph G;
judging whether rl of another node adjacent to the edge with the maximum weight is zero or not, if so, continuing to use the slave node vpStarting, selecting the side with the maximum side weight for path diffusion according to the undirected graph G; if not, determining the non-zero node as an end node, and ending the traversal process;
setting the rl of all nodes on the traversed diffusion path as the rl of the end node, and stopping until the rl of all nodes is not zero.
5. The method of claim 4, further comprising:
determining the similarity between nodes in the undirected graph as the weight of the edge in the undirected graph, and assuming that a node v existsiAnd node vjBy the formula
Calculating similarity η (v) between nodesi,vj)。
6. The method according to any one of claims 1 to 5, wherein the obtaining of the initial labeling set S for labeling each piece of data in the first device data X of the first device comprises:
labeling each piece of data in first equipment data X of the first equipment by using a training model phi of the first equipment to obtain an initial labeling set S ═ S1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, where SjE the set of categories C, n represents the total number of categories.
7. An autonomous learning device for internet of things equipment, the device comprising:
the device comprises a first determining unit, a second determining unit and a third determining unit, wherein the first determining unit is used for determining first equipment meeting a first preset condition according to the attribute of second equipment, the second equipment is terminal equipment newly added into the Internet of things, and the first equipment is the existing terminal equipment in the Internet of things;
a first obtaining unit, configured to obtain first device data X of the first device and second device data Y of the second device that satisfy a second preset condition;
a second obtaining unit, configured to obtain an initial labeling set S ═ S [ S ] for labeling each piece of data in the first device data X of the first device1,s2,…,sn]Wherein s isjIndicates any one of the labeled results in S, where Sj∈{C1,C2,…,Cn}；
A second determining unit, configured to construct an undirected graph according to the second device data Y, where each piece of the second device data Y isjCorresponding to the node v in the undirected graphjNode v in the undirected graphjThe similarity between the two is corresponding to the edge in the undirected graph; determining each node v in the undirected graphjDetermining a reliable node set Ψ in the initial labeling set S according to the eccentricity; according to the reliable node set Ψ and the initializationThe label set determines an updated label set RL; and training the second equipment data Y according to the updated label set RL to obtain a training model phi' of the second equipment.
8. An autonomous learning device for internet of things devices, comprising a memory and a processor, wherein the memory stores a computer program operable on the processor, and the processor executes the program to implement the steps of the autonomous learning method for internet of things devices according to any one of claims 1 to 6.
9. A computer-readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps in the method for autonomous learning by a device of the internet of things of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810001639.XA CN110009113B (en) | 2018-01-02 | 2018-01-02 | Internet of things equipment autonomous learning method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810001639.XA CN110009113B (en) | 2018-01-02 | 2018-01-02 | Internet of things equipment autonomous learning method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110009113A CN110009113A (en) | 2019-07-12 |
| CN110009113B true CN110009113B (en) | 2021-05-11 |
Family
ID=67164129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810001639.XA Active CN110009113B (en) | 2018-01-02 | 2018-01-02 | Internet of things equipment autonomous learning method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110009113B (en) |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8001121B2 (en) * | 2006-02-27 | 2011-08-16 | Microsoft Corporation | Training a ranking function using propagated document relevance |
| US8019763B2 (en) * | 2006-02-27 | 2011-09-13 | Microsoft Corporation | Propagating relevance from labeled documents to unlabeled documents |
| US20090132561A1 (en) * | 2007-11-21 | 2009-05-21 | At&T Labs, Inc. | Link-based classification of graph nodes |
| CN102163285A (en) * | 2011-03-09 | 2011-08-24 | 北京航空航天大学 | Cross-domain video semantic concept detection method based on active learning |
| CN102768670B (en) * | 2012-05-31 | 2014-08-20 | 哈尔滨工程大学 | Webpage clustering method based on node property label propagation |
| CN102890698B (en) * | 2012-06-20 | 2015-06-24 | 杜小勇 | Method for automatically describing microblogging topic tag |
| CN105893382A (en) * | 2014-12-23 | 2016-08-24 | 天津科技大学 | Priori knowledge based microblog user group division method |
| CN106778878B (en) * | 2016-12-21 | 2020-06-26 | 东方网力科技股份有限公司 | Character relation classification method and device |
-
2018
- 2018-01-02 CN CN201810001639.XA patent/CN110009113B/en active Active
Non-Patent Citations (2)
| Title |
|---|
| A Low-Cost and Accurate Indoor Localization Algorithm Using Label Propagation Based Semi-supervised Learning;Shaoshuai Liu 等;《2009 Fifth International Conference on Mobile Ad-hoc and Sensor Networks》;20100129;第108-111页 * |
| Improving semi-supervised learning through optimum connectivity;Willian P.Amorim 等;《Pattern Recognition》;20161231;第60卷;第72-85页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110009113A (en) | 2019-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109919316B (en) | Method, device and equipment for acquiring network representation learning vector and storage medium | |
| US10482375B2 (en) | Deep graph representation learning | |
| US10204090B2 (en) | Visual recognition using social links | |
| Cieslewski et al. | Efficient decentralized visual place recognition using a distributed inverted index | |
| US10661438B2 (en) | Robot apparatus, methods and computer products | |
| WO2021089013A1 (en) | Spatial graph convolutional network training method, electronic device and storage medium | |
| CN110929046B (en) | Knowledge entity recommendation method and system based on heterogeneous network embedding | |
| CN111310074B (en) | Method and device for optimizing labels of interest points, electronic equipment and computer readable medium | |
| CN109684456B (en) | Scene ability intelligent question-answering system based on Internet of things ability knowledge graph | |
| US20150339591A1 (en) | Collegial Activity Learning Between Heterogeneous Sensors | |
| CN106845644A (en) | A kind of heterogeneous network of the contact for learning user and Mobile solution by correlation | |
| CN108696453B (en) | Lightweight SDN service flow notification method and system | |
| US11593665B2 (en) | Systems and methods driven by link-specific numeric information for predicting associations based on predicate types | |
| CN113190754B (en) | Recommendation method based on heterogeneous information network representation learning | |
| CN111400504A (en) | Method and device for identifying enterprise key people | |
| Chikkamannur | Semantic Annotation of IoT Resource with ontology orchestration | |
| CN115686868B (en) | Cross-node-oriented multi-mode retrieval method based on federated hash learning | |
| CN114943017A (en) | Cross-modal retrieval method based on similarity zero sample hash | |
| CN110009113B (en) | Internet of things equipment autonomous learning method, device, equipment and storage medium | |
| US9819752B2 (en) | Method and system for discovery of services in interconnected smart spaces | |
| WO2022162677A1 (en) | Distributed machine learning with new labels using heterogeneous label distribution | |
| Jeon et al. | Ontology-based user intention recognition for proactive planning of intelligent robot behavior | |
| JP2012155394A (en) | Document classification learning control device, document classification device and computer program | |
| Lim et al. | A modular approach to landmark detection based on a Bayesian network and categorized context logs | |
| CN112003907A (en) | Deterministic forwarding method for network resource demand and computational power demand thereof |
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 |