CN104808977B - A kind of method and device for the state for determining user - Google Patents
A kind of method and device for the state for determining user Download PDFInfo
- Publication number
- CN104808977B CN104808977B CN201410040098.3A CN201410040098A CN104808977B CN 104808977 B CN104808977 B CN 104808977B CN 201410040098 A CN201410040098 A CN 201410040098A CN 104808977 B CN104808977 B CN 104808977B
- Authority
- CN
- China
- Prior art keywords
- state
- rbm
- user
- electronic equipment
- aobvious layer
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
The invention discloses a kind of method and device for the state for determining user.Methods described includes：Obtain the N class parameters on the electronic equipment of the user；Wherein, N is positive integer；The N classes parameter is separately fixed on constraint Boltzmann machine RBM N number of aobvious layer neuron, the input as the RBM；The RBM is handled the N classes parameter；By the N+1 aobvious layer neuron output results of the RBM, the result represents the state of the user.
Description
Technical field
The present invention relates to electronic technology field, more particularly to a kind of method and device for the state for determining user.
Background technology
The One function of existing instant communication software is the state that user can set instant communication software manually, optionally
State includes：Online, busy, stealthy, offline four basic status.Tencent QQ is further, by detecting mouse-keyboard attonity
Time, after the attonity time is more than X minutes, i.e., the state of instant communication software is set away from, it is any when monitoring
The state of instant communication software is arranged to online or busy during the action of mouse or keyboard and resets timer.Such as Fig. 1 institutes
Show, be Tencent QQ（2013 editions）For setting the interface of attonity timer.
What online, busy, stealthy, offline four kinds of states substantially reflected is not the state of chat software, but reflection is chatted
The state of its software user, i.e. user, when user enters busy condition from presence, user usually not and
When the custom for manually changing instant communication software state, generally whether judge user by detecting the method for attonity time
Offline, but stealthy state always needs user oneself to set, because software can not judge whether user needs to be in stealthy shape
State.
In addition, in some other application scenarios, it is also desirable to the state of user is obtained, such as：Electronic equipment judge be
It is no whether to judge user in the state using electronic equipment when entering battery saving mode, if in busy state,
Represent using electronic equipment, then it is working condition that will keep electronic equipment.Under the usage scenario, generally also by
User is configured by user setup interface, such as sets the parameter and time interval of monitoring.
Therefore, in the prior art, typically user oneself judges the state of oneself, then changes electronic equipment manually again
State or communication software state；Either electronic equipment is set to need the parameter monitored by user, then electronic equipment
It is monitored, the state of electronic equipment or the state of communication software is just changed when monitoring result meets condition.Therefore, it is existing
Determine that the method for the state of user is required to user and set manually in technology, so less efficient and not intelligent enough.
The content of the invention
The embodiment of the present invention provides a kind of method and device for the state for determining user, to solve to determine in the prior art
The method of the state of user is required to user and set manually, thus it is less efficient, and not intelligent enough technical problem.
One aspect of the present invention provides a kind of method for the state for determining user, and methods described includes：
Obtain the N class parameters on the electronic equipment of the user；Wherein, N is positive integer；The N classes parameter is solid respectively
It is scheduled on constraint Boltzmann machine RBM N number of aobvious layer neuron, the input as the RBM；The RBM is to the N classes parameter
Handled；By the N+1 aobvious layer neuron output results of the RBM, the result represents the state of the user.
Optionally, operation has an instant communication software on the electronic equipment, and methods described also includes：According to the result
Change the state of the current account of the instant communication software.
Optionally, after user changes the RBM state of N+1 aobvious layer neuron manually, methods described also includes：Will
State and the N classes parameter after the change sample new as one, re -training RBM.
Another aspect of the present invention also provides a kind of device for the state for determining user, including：
Acquiring unit, the N class parameters on the electronic equipment for obtaining the user；Wherein, N is positive integer；Processing is single
Member, for the N classes parameter to be separately fixed to constraint Boltzmann machine RBM N number of aobvious layer neuron, as the RBM
Input；The RBM, for handling the N classes parameter；And exported by the N+1 aobvious layer neuron of the RBM
As a result, the result represents the state of the user.
Optionally, the processing unit is additionally operable to when operation has an instant communication software on the electronic equipment, according to
The result changes the state of the current account of the instant communication software.
Optionally, the RBM is additionally operable to after user changes the RBM state of N+1 aobvious layer neuron manually, will more
State and the N classes parameter after the changing sample new as one, re -training RBM.
The one or more technical schemes provided in the embodiment of the present invention, have at least the following technical effects or advantages：
In an embodiment of the present invention, the N class parameters on the electronic equipment of user are obtained；Wherein, N is positive integer；By N classes
Parameter is separately fixed on constraint Boltzmann machine RBM N number of aobvious layer neuron, the input as RBM；RBM enters to N class parameters
Row processing；By RBM N+1 aobvious layer neuron output results, the state of user is as a result represented.It can thus be seen that
The N class parameters on electronic equipment are separately fixed on RBM aobvious layer neuron in the present embodiment, then pass through RBM operation
Principle is handled N class parameters, then the N+1 aobvious layer neuron output results from RBM, then this result can is used
To represent the state of user, so the method in the present embodiment does not need user to set manually first, but electronic equipment is intelligent
The real state of judgement user so that electronic equipment is more intelligent.Further, the method in the present embodiment employs
RBM judges the state of user, and RBM neuron can flexibly increase, so the parameter that can be monitored can be a lot, such as
This one, the state of the user judged by this method is just closer to the real state of user, so accuracy can be significantly
Increase.
Further, in one embodiment, if operation has an instant communication software on electronic equipment, then further may be used also
To change the state of the current account of instant communication software according to the result, it is possible to intelligently judge instant communication software
" online " or " busy " state of user, instant messaging is then set automatically when user does not select " offline " or " stealthy "
The state of software so that the state of instant communication software can really reflect the state of its user.
Further, in one embodiment, after user have changed the state of itself manually, then one will be formed
New sample, and the new sample re -training RBM is utilized, so the method in the present embodiment enables electronic equipment continuous
Study, Grow Your Own is more intelligent, while also causes the result of output to be increasingly able to accurately reflect the true shape of user
State.
Brief description of the drawings
Fig. 1 is the interface schematic diagram for setting attonity timer in the prior art；
Fig. 2 is the simplification structural representation of the RBM in one embodiment of the invention；
Fig. 3 is the structural representation of the RBM in an instantiation of the invention；
Fig. 4 is the flow chart of the method for the state of the determination user in one embodiment of the invention；
Fig. 5 is the functional block diagram of the electronic equipment of the state of the determination user in one embodiment of the invention.
Embodiment
The embodiment of the present invention provides a kind of method and device for the state for determining user, to solve to determine in the prior art
The method of the state of user is required to user and set manually, thus it is less efficient, and not intelligent enough technical problem.
Technical scheme in the embodiment of the present invention is the above-mentioned technical problem of solution, and general thought is as follows：
In an embodiment of the present invention, the N class parameters on the electronic equipment of user are obtained；Wherein, N is positive integer；By N classes
Parameter is separately fixed on constraint Boltzmann machine RBM N number of aobvious layer neuron, the input as RBM；RBM enters to N class parameters
Row processing；By RBM N+1 aobvious layer neuron output results, the state of user is as a result represented.It can thus be seen that
The N class parameters on electronic equipment are separately fixed on RBM aobvious layer neuron in the present embodiment, then pass through RBM operation
Principle is handled N class parameters, then the N+1 aobvious layer neuron output results from RBM, then this result can is used
To represent the state of user, so the method in the present embodiment does not need user to set manually first, but electronic equipment is intelligent
The real state of judgement user so that electronic equipment is more intelligent.Further, the method in the present embodiment employs
RBM judges the state of user, and RBM neuron can flexibly increase, so the parameter that can be monitored can be a lot, such as
This one, the state of the user judged by this method is just closer to the real state of user, so accuracy can be significantly
Increase.
In order to be better understood from above-mentioned technical proposal, constraint Boltzmann machine is introduced first（Restricted
Boltzmann Machine；RBM）, RBM is a kind of neutral net based on statistical mechanics, be refer to shown in Fig. 2, is an allusion quotation
The RBM structure charts of type.In fig. 2, each node 10 is a neuron, and all neurons are divided into two layers, are in
One layer of lower section, such as neuron v1With neuron v2Referred to as aobvious layer neuron, one layer in top, such as neuron h1, neuron
h2With neuron h3Referred to as hidden neuron.Further, from figure 2 it can be seen that the input of aobvious layer neuron is all from hidden
Layer neuron, and the input of hidden neuron is all from aobvious layer neuron.
The state of each neuron has two, and 1 or -1,1 represents state of activation, and -1 represents unactivated state, that aobvious layer
Neuronal activation or inactive probability refer to formula（1）：
Wherein, P (vi| h) expression shows layer neuronal activation i-th in the case of known all hidden neuron states
Conditional probability（vi=1）, the wherein h of runic represents the state vector of all hidden neurons, viRepresent i-th and show layer neuron
State；wijRepresent the weights between j-th of hidden neuron and i-th of aobvious layer neuron；hjRepresent j-th hidden neuron
State；T can regard a constant as.
And hidden neuron activation or inactive probability refer to formula（2）：
Wherein, P (hj| v) represent j-th of hidden neuron activation in the case of known all aobvious layer neuron states
The v of conditional probability, wherein runic represents the state vector of all aobvious layer neurons.
Then weight matrix { the w of neutral net is trainedij, specific training method can be：Assuming that there is a training sample：
D={ d1,d2,...,dN,dN+1, all elements are+1 or -1 in D.Carry out that D is fixed on into RBM's first during network training
On aobvious layer neuron.So-called " fixation " refers to the state of RBM i-th of neuron of aobvious layer being set as di.In known D situation
Under multiple sampling is carried out to all hidden neuron h, calculate average dihj, dihjRepresent diWith hjMultiple sampling value product
Average.
Then the locking to showing layer neuron is removed（Non-locking state is also referred to as free state）, by all hidden neurons
Original state be set as -1, then according to the state and formula of hidden neuron（1）The activation probability of aobvious layer neuron is calculated,
Show the state of layer neuron according to the activation probability sampling.Obtain recycling formula after the state of aobvious layer neuron（2）Calculate
The activation probability of all hidden neurons, according to the state of the activation probability sampling hidden neuron.Above procedure is referred to as once
Iteration.After some wheel iteration, RBM networks will enter equilibrium state（RBM iterative process can be construed to Markov
The state migration procedure of chain, so markovian equilibrium state will be entered after successive ignition）.Into equilibrium state
Afterwards, then some wheel iteration is carried out and calculate average vihj, vihjRepresent the v under equilibrium stateiWith hjMultiple sampling value product
Average.
According to formula（3）Update the weights of network（More new formula is derived according to gradient descent method for this）.
wij←wij+<dihj>-<vihj>（3）
If D is called an AD HOC, then RBM training result be：RBM shows layer neuron in a free state
Generation pattern D maximization.If RBM training sample is the set { D of multiple patterns1,D2,...,Dm, then RBM
The result of training is：RBM shows the maximization of pattern in layer neuron productive set in a free state.
RBM has the function of Pattern completion.Such as by D1Top n element be locked in RBM top n and show layer neuron
On, the aobvious layer neurons of N+1 are free, after some wheel sampling iteration N+1 aobvious layer neuron by with close to 1 it is general
Rate is equal to D1The N+1 element.Now claim RBM in known mode D1A part when complete D1Remainder.The present invention
It is such a characteristic that make use of RBM.
Below in conjunction with the side of the state of the determination user of Figure of description and specific embodiment to the present embodiment
Method is described in detail.
It refer to shown in Fig. 3, be the network structure of the RBM employed in the present embodiment, the aobvious layer in neutral net is total to
There is N+1 neuron, wherein, N is positive integer；Hidden layer shares M neuron, in practice, the number of hidden neuron
It could be arranged to 2 times or so of aobvious layer neuron number, the requirement do not fixed, but should not too much also should not be very little.
The busy-idle condition of the state representation user of N+1 aobvious layer neurons, neuron vN+1, and neuron vN+1With two
Kind state, the busy and another state representation of one of which state representation of neuron can be used idle（It is general to use activation shape
State represents busy, and unactivated state represents idle）.It is generally necessary to initial training is carried out to network.Process is trained to collect first
Sample { D1,D2,...,Dm, then according to formula（3）The method training network of offer.It will not be described here.
Then the N class parameters on electronic equipment are separately fixed at into the 1st to show on layer neuron to n-th, such as neuron
v1To neuron vN, have per class parameter be with no two aspects, correspond to activation and the unactivated state of each neuron respectively.It is logical
After crossing manner described above progress successive ignition, sample neuron vN+1State, and by the state output to represent user
Busy-idle condition.
If user disagrees the output of network（Such as network output represents that user does, but the state of user is not busy, or
The output of person's network represents that user is not busy, but the state of user is busy）, then user can manually set the busy-idle condition of oneself,
Once user is operated manually, RBM just obtains a new sample Dx, wherein top n element is in current electronic device
N class parameters, the N+1 element is the busy-idle condition of oneself that user manually sets.Now can be by DxIt is added to sample set
Remove an earliest sample in conjunction and from sample set.Then RBM is carried out again according to new sample set to network
Training.
Referring next to Fig. 4, Fig. 4 is the flow chart of the method for the state of the determination user in the present embodiment, this method
Including：
Step 101：Obtain the N class parameters on the electronic equipment of user；
Step 102：N class parameters are separately fixed on RBM N number of aobvious layer neuron, the input as RBM；
Step 103：RBM is handled N class parameters；
Step 104：By RBM N+1 aobvious layer neuron output results, the result represents the state of user.
Specifically, in a step 101, can be by user to electricity although the state of user can not be obtained directly
The mode of operation of sub- equipment to determine the state of user indirectly, it is possible to the use of operation electronic equipment is represented with electronic equipment
Family, so the electronic equipment of user here is the electronic equipment operated by user.
Further, in a step 101, the N class parameters on the electronic equipment of user are obtained, this N classes parameter can be electronics
Equipment arbitrarily can embody user's possible operation or not operate the parameter of electronic equipment, such as in past predetermined amount of time
Whether keyboard action is had；Whether the office software on electronic equipment, such as word are active；Or on electronic equipment
Whether system time is day off.
And step 101 can periodically obtain N class parameters, such as：Acquisition per minute is once.
Specific acquisition modes can be different for inhomogeneous parameter, such as are for the office software on electronic equipment
As long as the no explorer can for being active detecting on electronic equipment is got；And for being on electronic equipment
Whether the system time is day off, can be by checking that system clock can is got.
After N class parameters are got, step 102 is carried out, i.e., N class parameters are separately fixed to RBM N number of aobvious layer god
Through in member, the input as RBM.
Specifically, from the foregoing aobvious layer neuron to RBM state description, each neuron has two shapes
State, wherein, state of activation is represented with 1, and unactivated state use -1 represents, then N class parameters whether two aspects are just corresponding N number of
The two states of aobvious layer neuron, so these final parameters are used as RBM input in the form of 1 or -1.Such as：Electronics is set
Standby upper system time is day off, the state of activation of the corresponding aobvious layer neuron fixed；System time on electronic equipment
For non-day off, then the unactivated state of the corresponding aobvious layer neuron fixed；If this parameter obtained is " electronics is set
Standby upper system time is day off ", then the input for the aobvious layer neuron fixed is 1；Similar, by N classes parameter institute
The input of fixed aobvious layer neuron is determined and inputted.
Then step 103 is performed, i.e. RBM is handled N class parameters, is specially：According to the input in step 102, profit
Successive ignition computing is carried out with RBM operating mechanism.Specifically, N class parameters are locked in the 1st to N number of aobvious layer nerve first
In member, and the original state of all hidden neurons is set as -1, then uses formula（1）Calculate N+1 aobvious layer neuron
Activation probability and sample obtain its state（1st is locked out to the state of N number of aobvious layer neuron, it is not necessary to it is updated, this
When obtained N+1 aobvious layer neuron state be that intermediate result is not last output）.Thus owned
The state of aobvious layer neuron, hidden neuron have input, then just utilize formula（2）Calculate the activation of all hidden neurons
Probability is simultaneously sampled and obtains state, after the completion of renewal, result is exported to N+1 aobvious layer neurons, then this N+1 aobvious layers nerve
Member has new input again, then proceedes to according to formula（1）Calculate N+1 aobvious layer neuronal activation probability and sample and obtain shape
State（Other aobvious layer neurons are in the lock state）, after the completion of renewal, export to hidden neuron, then proceed to utilize formula
（2）Update the state of hidden neuron；Above procedure is repeated until reaching an equilibrium state, now, N+1 aobvious layer nerve
The result of member output, you can to represent the state of user that whole neutral net exported.
The N class parameters on electronic equipment are separately fixed at the aobvious of RBM in the present embodiment it can be seen from foregoing description
On layer neuron, then N class parameters are handled by RBM operation logic, then the N+1 aobvious layer neuron from RBM
Output result, then this result can is used for representing the state of user, so the method in the present embodiment first need not
User is set manually, but the real state of judgement user of electronic equipment intelligence, so that electronic equipment is more intelligent.
Further, the method in the present embodiment employs RBM to judge the state of user, and RBM neuron can flexibly increase,
So the parameter that can be monitored can be a lot, consequently, it is possible to which the state of the user judged by this method is just closer to user
Real state, so accuracy can significantly increase.
Further, after the state of user is obtained, because being required to determine the state of user, Ran Hougen in many scenes
Subsequent treatment is carried out according to the state of user, such as：Operation has an instant communication software on electronic equipment, then this method is also wrapped
Include：The state of the current account of instant communication software is changed according to the result so that the state of current account can be with user's
Time of day is consistent, so the shape of the current account for setting instant communication software that can be intelligent by the method in the present embodiment
State, go to set manually without user.
Certainly, in practice, although defining the state of user by the method described in Fig. 4, user is also
State can be changed manually, such as is modified by an input block；Or user can also change instant communication software
The state of current account.As an example it is assumed that the state of the user determined is busy, but user have selected to be non-busy
State, then RBM will be using the state after change and N classes the parameter sample new as one, re -training RBM.Therefore, pass through
Method in the present embodiment so that electronic equipment can constantly learn, and Grow Your Own is more intelligent, while also cause after
The result of continuous output increasingly accurately reflects the time of day of user.
A specific example will be lifted below to illustrate the implementation process of the method in the present embodiment, wherein N values are 12,
Between corresponding relation, N classes parameter and N number of aobvious layer neuron between User Status and the state of N+1 aobvious layer neurons
Corresponding relation is as follows：
Aobvious layer neuron v13：Activation --- user is in busy condition；It is inactive --- user is in online（It is non-busy）
State；Aobvious layer neuron v1：Activation --- there is keyboard action within past A minutes；It is inactive --- within past A minutes
On-keyboard acts；Aobvious layer neuron v2：Activation --- there is keyboard action within past B minutes；It is inactive --- in past B
On-keyboard action in minute；Aobvious layer neuron v3：Activation --- there is mouse action within past C minutes；It is inactive ---
Without mouse action in past C minutes；Aobvious layer neuron v4：Activation --- there is mouse action within past D minutes；It is non-to swash
It is living --- without mouse action within past D minutes；Aobvious layer neuron v5：Activation --- keyboarding speed test is more than E beats/min
Clock；It is inactive --- keyboarding speed test is less than or equal to E beats/min；Aobvious layer neuron v6：Activation --- mouse action speed is big
In F beats/min；It is inactive --- mouse action speed is less than or equal to F beats/min；Aobvious layer neuron v7：Activation --- network is clear
Device of looking at is active；It is inactive --- web browser is in unactivated state；Aobvious layer neuron v8：Activation ---
Office Word softwares are active；It is inactive --- Office Word softwares are in unactivated state；Aobvious layer nerve
First v9：Activation --- Office Excel softwares are active；It is inactive --- Office Excel softwares are in non-and swashed
State living；Aobvious layer neuron v10：Activation --- Office Outlook softwares are active；It is inactive --- Office
Outlook softwares are in unactivated state；Aobvious layer neuron v11：Activation --- system time is day off；It is inactive --- be
The system time is non-day off；Aobvious layer neuron v12：Activation --- system time is the time of having a rest on working day；It is inactive --- system
Time is the working time on working day；Wherein, A and B are differed, and C and D are differed.
After these parameters all fix, the great amount of samples of this 13 variables is just gathered, then as input variable pair
RBM is trained, and RBM is by memory neuron v after training13With neuron v1~v12Between relation.RBM starts working
When neuron v will be gathered from electronic equipment1~v12As RBM input, start sampling renewal to every other neuron, treat
Neuron v after RBM is run to equilibrium state13Output of the state of output as RBM.
Certainly, user is possible to disagree RBM judgement, and user can select to correct RBM output, and now RBM is by root
The weights for the whole neutral net of result re -training corrected according to user.
Based on same inventive concept, the embodiment of the present application also provides a kind of device for the state for determining user, refer to Fig. 5
Shown, for the functional block diagram of the device of the state of the determination user in the present embodiment, the device includes：
Acquiring unit 201, the N class parameters on the electronic equipment for obtaining user；Wherein, N is positive integer；Processing unit
202, for N class parameters to be separately fixed to constraint Boltzmann machine RBM203 N number of aobvious layer neuron, as RBM203's
Input；RBM203, for handling N class parameters；And pass through RBM203 N+1 aobvious layer neuron output results, knot
Fruit represents the state of user.
Further, processing unit 202 is additionally operable to when operation has an instant communication software on electronic equipment, according to result more
Change the state of the current account of instant communication software.
Further, RBM203 be additionally operable to when by an input block change user state after, by the shape after change
State and the N classes parameter sample new as one, re -training RBM203.
The method of the state of the determination user in the device and previous embodiment of the state of user is determined in the present embodiment is
Based on the invention under same design, pass through the method for the state of foregoing couple of determination user and its retouching in detail for various change form
Stating, those skilled in the art can be apparent from determining the implementation process of the device of the state of user in the present embodiment, so
It is succinct for specification, it will not be repeated here.
By one embodiment in above-described embodiment in the present invention or multiple embodiments, following skill can be at least realized
Art effect：
In an embodiment of the present invention, the N class parameters on the electronic equipment of user are obtained；Wherein, N is positive integer；By N classes
Parameter is separately fixed on constraint Boltzmann machine RBM N number of aobvious layer neuron, the input as RBM；RBM enters to N class parameters
Row processing；By RBM N+1 aobvious layer neuron output results, the state of user is as a result represented.It can thus be seen that
The N class parameters on electronic equipment are separately fixed on RBM aobvious layer neuron in the present embodiment, then pass through RBM operation
Principle is handled N class parameters, then the N+1 aobvious layer neuron output results from RBM, then this result can is used
To represent the state of user, so the method in the present embodiment does not need user to set manually first, but electronic equipment is intelligent
The real state of judgement user so that electronic equipment is more intelligent.Further, the method in the present embodiment employs
RBM judges the state of user, and RBM neuron can flexibly increase, so the parameter that can be monitored can be a lot, such as
This one, the state of the user judged by this method is just closer to the real state of user, so accuracy can be significantly
Increase.
Further, in one embodiment, if operation has an instant communication software on electronic equipment, then further may be used also
To change the state of the current account of instant communication software according to the result, it is possible to intelligently judge instant communication software
" online " or " busy " state of user, instant messaging is then set automatically when user does not select " offline " or " stealthy "
The state of software so that the state of instant communication software can really reflect the state of its user.
Further, in one embodiment, after user have changed the state of itself manually, then one will be formed
New sample, and the new sample re -training RBM is utilized, so the method in the present embodiment enables electronic equipment continuous
Study, Grow Your Own is more intelligent, while also causes the result of output to be increasingly able to accurately reflect the true shape of user
State.
Obviously, those skilled in the art can carry out the essence of various changes and modification without departing from the present invention to the present invention
God and scope.So, if these modifications and variations of the present invention belong to the scope of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to comprising including these changes and modification.
Claims (6)
- A kind of 1. method for the state for determining user, it is characterised in that methods described includes：Obtain the N class parameters on the electronic equipment of the user；Wherein, N is positive integer；The N classes parameter is separately fixed on constraint Boltzmann machine RBM N number of aobvious layer neuron, as the defeated of the RBM Enter, for the N classes parameter to embody the parameter whether user operates the electronic equipment, N number of aobvious layer neuron there are two kinds of shapes State, the result for whether operating the electronic equipment of the N classes parameter correspond to the two states of N number of aobvious layer neuron；The RBM is handled the N classes parameter；By the N+1 aobvious layer neuron output results of the RBM, the result represents the state of the user.
- 2. the method as described in claim 1, it is characterised in that operation has an instant communication software, institute on the electronic equipment Stating method also includes：The state of the current account of the instant communication software is changed according to the result.
- 3. method as claimed in claim 1 or 2, it is characterised in that when the user changes N+1 of the RBM manually During the state of aobvious layer neuron, methods described also includes：Using the state after change and the N classes parameter sample new as one, RBM described in re -training.
- A kind of 4. device for the state for determining user, it is characterised in that including：Acquiring unit, the N class parameters on the electronic equipment for obtaining the user；Wherein, N is positive integer；Processing unit, for the N classes parameter to be separately fixed to constraint Boltzmann machine RBM N number of aobvious layer neuron, make For the input of the RBM, whether the N classes parameter operates the parameter of the electronic equipment, N number of aobvious layer god for embodiment user There are two states through member, the result for whether operating the electronic equipment of the N classes parameter corresponds to N number of aobvious layer neuron Two states；The RBM, for handling the N classes parameter；And N+1 aobvious layer neuron output knots for passing through the RBM Fruit, the result represent the state of the user.
- 5. device as claimed in claim 4, it is characterised in that the processing unit is additionally operable to run on the electronic equipment When having an instant communication software, the state of the current account of the instant communication software is changed according to the result.
- 6. the device as described in claim 4 or 5, it is characterised in that the RBM is additionally operable to when the user changes institute manually When stating the RBM state of N+1 aobvious layer neurons, using the state after change and the N classes parameter sample new as one, RBM described in re -training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410040098.3A CN104808977B (en) | 2014-01-27 | 2014-01-27 | A kind of method and device for the state for determining user |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410040098.3A CN104808977B (en) | 2014-01-27 | 2014-01-27 | A kind of method and device for the state for determining user |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104808977A CN104808977A (en) | 2015-07-29 |
| CN104808977B true CN104808977B (en) | 2018-03-27 |
Family
ID=53693831
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410040098.3A Active CN104808977B (en) | 2014-01-27 | 2014-01-27 | A kind of method and device for the state for determining user |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104808977B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106059901A (en) * | 2016-06-24 | 2016-10-26 | 张水成 | Method and system for notifying usage status of instant messaging terminal |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101611602A (en) * | 2007-02-20 | 2009-12-23 | 斯凯普有限公司 | Instant messaging activity notification |
| CN103150595A (en) * | 2011-12-06 | 2013-06-12 | 腾讯科技（深圳）有限公司 | Automatic pair selection method and device in data processing system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5568449B2 (en) * | 2010-11-19 | 2014-08-06 | 株式会社日立製作所 | Passive optical network system, station side apparatus, and power consumption control method |
-
2014
- 2014-01-27 CN CN201410040098.3A patent/CN104808977B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101611602A (en) * | 2007-02-20 | 2009-12-23 | 斯凯普有限公司 | Instant messaging activity notification |
| CN103150595A (en) * | 2011-12-06 | 2013-06-12 | 腾讯科技（深圳）有限公司 | Automatic pair selection method and device in data processing system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104808977A (en) | 2015-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bornholdt et al. | Self-organized critical neural networks | |
| CN106709565A (en) | Neural network optimization method and device | |
| US11074503B2 (en) | Execution of a genetic algorithm having variable epoch size with selective execution of a training algorithm | |
| CN113361680B (en) | Neural network architecture searching method, device, equipment and medium | |
| CN110135637A (en) | Micro-capacitance sensor short-term load forecasting method based on shot and long term memory and adaptive boosting | |
| CN109190979A (en) | A kind of industry internet of things data analysis method, system and relevant device | |
| CN109189762A (en) | A kind of industry internet of things data analysis method, system and relevant device | |
| CN107103359A (en) | The online Reliability Prediction Method of big service system based on convolutional neural networks | |
| CN108108743A (en) | Abnormal user recognition methods and the device for identifying abnormal user | |
| Rummukainen et al. | Practical reinforcement learning-experiences in lot scheduling application | |
| CN113485801B (en) | Real-time DNN scheduling system and method based on neural network similarity modeling | |
| CN109389518A (en) | Association analysis method and device | |
| CN111416797A (en) | Intrusion detection method for optimizing regularization extreme learning machine by improving longicorn herd algorithm | |
| Prignano et al. | Extracting topological features from dynamical measures in networks of Kuramoto oscillators | |
| CN115510042A (en) | Power system load data filling method and device based on generation countermeasure network | |
| US20240095535A1 (en) | Executing a genetic algorithm on a low-power controller | |
| CN110866602A (en) | Method and device for integrating multitask model | |
| Valdez et al. | Parallel genetic algorithms for optimization of Modular Neural Networks in pattern recognition | |
| CN104808977B (en) | A kind of method and device for the state for determining user | |
| Wang et al. | Design of efficient hybrid neural networks for flexible flow shop scheduling | |
| CN109902921A (en) | Management method, device, equipment and the storage medium of user's growth system | |
| Franzosi et al. | Riemannian-geometric entropy for measuring network complexity | |
| CN104537224B (en) | Multi-state System Reliability analysis method and system based on adaptive learning algorithm | |
| CN110705756A (en) | Electric power energy consumption optimization control method based on input convex neural network | |
| CN113988399A (en) | Comprehensive energy scheduling method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |