CN109165738A - Optimization method and device, electronic equipment and the storage medium of neural network model - Google Patents
Optimization method and device, electronic equipment and the storage medium of neural network model Download PDFInfo
- Publication number
- CN109165738A CN109165738A CN201811093361.XA CN201811093361A CN109165738A CN 109165738 A CN109165738 A CN 109165738A CN 201811093361 A CN201811093361 A CN 201811093361A CN 109165738 A CN109165738 A CN 109165738A
- Authority
- CN
- China
- Prior art keywords
- model
- output
- full articulamentum
- student model
- student
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
Abstract
This disclosure relates to a kind of optimization method and device of neural network model, electronic equipment and storage medium.This method comprises: the selected part neuron from the full articulamentum before the output layer of student model；Input data is inputted to student model and teacher's model respectively, based on the partial nerve member chosen from the full articulamentum before the output layer of student model, obtains the output of student model；Whole neurons in full articulamentum before output layer based on teacher's model, obtain the output of teacher's model；The output of output and teacher's model based on student model optimizes student model.The embodiment of the present disclosure can stablize the accuracy rate for improving neural network model, and be capable of the generalization ability of strength neural network model under the premise of not increasing training data total amount and without re -training.
Description
Technical field
This disclosure relates to depth learning technology field more particularly to a kind of optimization method and device of neural network model,
Electronic equipment and storage medium.
Background technique
Currently, neural network model is applied to the various aspects such as computer vision and natural language processing (such as pedestrian
Retrieval and recognition of face), and achieve preferable effect.After neural network model training is completed, the accuracy rate of test is past
Toward the upper limit for the accuracy rate for limiting neural network model.
In the related technology, usually by increasing the total amount of training data, changing the structure of neural network model or to mould
Type carries out subtle adjustment and retraining, to improve the accuracy rate of neural network model.These methods all propose data volume non-
Often high requirement.The total amount and the different accuracy rate for surely improving neural network model of subtle adjustment for increasing training data, may
The reason of be the quality of data it is irregular lead to neural network model training difficulty increase.Change the structure of neural network model
A usually not specific standard, therefore not can guarantee the accuracy rate of neural network model after training.
Summary of the invention
The present disclosure proposes a kind of optimisation technique schemes of neural network model.
According to the one side of the disclosure, a kind of optimization method of neural network model is provided, comprising:
The selected part neuron from the full articulamentum before the output layer of student model；
Input data is inputted to the student model and teacher's model respectively, based on the output layer from the student model it
The partial nerve member chosen in preceding full articulamentum, obtains the output of the student model；
Whole neurons in full articulamentum before output layer based on teacher's model, obtain teacher's model
Output；
The output of output and teacher's model based on the student model, optimizes the student model.
In one possible implementation, the selected part nerve from the full articulamentum before the output layer of student model
Member, comprising:
The selected part neuron from the last one full articulamentum before the output layer of student model.
In one possible implementation, the selected part nerve from the full articulamentum before the output layer of student model
Member, comprising:
For each neuron in the first full articulamentum, a random number is generated in first interval respectively, wherein institute
Stating the first full articulamentum indicates that the output layer of the student model carries out the full articulamentum of neuron selection before；
If the corresponding random number of peripheral sensory neuron in the first full articulamentum belongs to second interval, from described first
The peripheral sensory neuron is chosen in full articulamentum, wherein the second interval is the subset of the first interval, and described second
Section is not equal to the first interval.
In one possible implementation, the output of output and teacher's model based on the student model, it is excellent
Change the student model, comprising:
Determine the mean square error between the output of the student model and the output of teacher's model；
Based on the mean square error, first-loss function is obtained；
Using student model described in the first-loss function optimization.
In one possible implementation, the output of output and teacher's model based on the student model, it is excellent
Change the student model, comprising:
Determine the relative entropy of the output of the output phase of the student model for teacher's model；
Based on the relative entropy, the second loss function is obtained；
The student model is optimized using second loss function.
In one possible implementation, it is based on the relative entropy, obtains the second loss function, comprising:
Determine the corresponding regulation coefficient of the relative entropy；
The product of the relative entropy and the regulation coefficient is determined as the second loss function.
In one possible implementation, the output is the decilog of decilog layer output.
According to the one side of the disclosure, a kind of optimization device of neural network model is provided, comprising:
Module is chosen, for the selected part neuron from the full articulamentum before the output layer of student model；
First determining module is based on for input data to be inputted to the student model and teacher's model respectively from described
The partial nerve member chosen in full articulamentum before the output layer of student model, obtains the output of the student model；
Second determining module, for all nerves in the full articulamentum before the output layer based on teacher's model
Member obtains the output of teacher's model；
Optimization module optimizes the student for the output of output and teacher's model based on the student model
Model.
In one possible implementation, the selection module is used for:
The selected part neuron from the last one full articulamentum before the output layer of student model.
In one possible implementation, the selection module includes:
Submodule is generated, for generating one in first interval respectively for each neuron in the first full articulamentum
A random number, wherein the first full articulamentum indicates that the output layer of the student model carries out the complete of neuron selection before
Articulamentum；
Submodule is chosen, if the corresponding random number of peripheral sensory neuron in the described first full articulamentum belongs to the secondth area
Between, then the peripheral sensory neuron is chosen from the described first full articulamentum, wherein the second interval is the first interval
Subset, and the second interval is not equal to the first interval.
In one possible implementation, the optimization module includes:
First determines submodule, equal between the output of the student model and the output of teacher's model for determining
Square error；
Second determines submodule, for being based on the mean square error, obtains first-loss function；
First optimization submodule, for using student model described in the first-loss function optimization.
In one possible implementation, the optimization module includes:
Third determines submodule, for determining the phase of the output of the output phase of the student model for teacher's model
To entropy；
4th determines submodule, for being based on the relative entropy, obtains the second loss function；
Second optimization submodule, for optimizing the student model using second loss function.
In one possible implementation, the described 4th determine that submodule includes:
First determination unit, for determining the corresponding regulation coefficient of the relative entropy；
Second determination unit, for the product of the relative entropy and the regulation coefficient to be determined as the second loss function.
In one possible implementation, the output is the decilog of decilog layer output.
According to the one side of the disclosure, a kind of electronic equipment is provided, comprising:
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to: execute the optimization method of above-mentioned neural network model.
According to the one side of the disclosure, a kind of computer readable storage medium is provided, computer program is stored thereon with
Instruction, the computer program instructions realize the optimization method of above-mentioned neural network model when being executed by processor.
In the embodiments of the present disclosure, based on the partial nerve chosen from the full articulamentum before the output layer of student model
Member obtains the output of student model, whole neurons in full articulamentum before the output layer based on teacher's model, obtains old
The output of teacher's model, and the output of the output based on student model and teacher's model optimize student model, and thus, it is possible to not increase
Add training data total amount and without stablizing the accuracy rate for improving neural network model, and for big portion under the premise of re -training
Neural network model and data is divided generally to be applicable in, and the part mind of the full articulamentum before the output layer by choosing student model
The output for effectively to be fitted student model through member, can effectively mitigate the over-fitting of neural network model, strength neural network mould
The generalization ability of type.
It should be understood that above general description and following detailed description is only exemplary and explanatory, rather than
Limit the disclosure.
According to below with reference to the accompanying drawings to detailed description of illustrative embodiments, the other feature and aspect of the disclosure will become
It is clear.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and those figures show meet this public affairs
The embodiment opened, and together with specification it is used to illustrate the technical solution of the disclosure.
Fig. 1 shows the flow chart of the optimization method of the neural network model according to the embodiment of the present disclosure.
Fig. 2 shows the illustrative streams according to the optimization method step S11 of the neural network model of the embodiment of the present disclosure
Cheng Tu.
Fig. 3 shows an illustrative stream of the optimization method step S14 according to the neural network model of the embodiment of the present disclosure
Cheng Tu.
Fig. 4 shows the another exemplary of the optimization method step S14 according to the neural network model of the embodiment of the present disclosure
Flow chart.
Fig. 5 shows an illustrative stream of the optimization method step S145 according to the neural network model of the embodiment of the present disclosure
Cheng Tu.
Fig. 6 shows the block diagram of the optimization device according to the neural network model of the embodiment of the present disclosure.
Fig. 7 shows an illustrative block diagram of the optimization device according to the neural network model of the embodiment of the present disclosure.
Fig. 8 is the block diagram of a kind of electronic equipment 800 shown according to an exemplary embodiment.
Fig. 9 is the block diagram of a kind of electronic equipment 1900 shown according to an exemplary embodiment.
Specific embodiment
Various exemplary embodiments, feature and the aspect of the disclosure are described in detail below with reference to attached drawing.It is identical in attached drawing
Appended drawing reference indicate element functionally identical or similar.Although the various aspects of embodiment are shown in the attached drawings, remove
It non-specifically points out, it is not necessary to attached drawing drawn to scale.
Dedicated word " exemplary " means " being used as example, embodiment or illustrative " herein.Here as " exemplary "
Illustrated any embodiment should not necessarily be construed as preferred or advantageous over other embodiments.
The terms "and/or", only a kind of incidence relation for describing affiliated partner, indicates that there may be three kinds of passes
System, for example, A and/or B, can indicate: individualism A exists simultaneously A and B, these three situations of individualism B.In addition, herein
Middle term "at least one" indicate a variety of in any one or more at least two any combination, it may for example comprise A,
B, at least one of C can indicate to include any one or more elements selected from the set that A, B and C are constituted.
In addition, giving numerous details in specific embodiment below in order to which the disclosure is better described.
It will be appreciated by those skilled in the art that without certain details, the disclosure equally be can be implemented.In some instances, for
Method, means, element and circuit well known to those skilled in the art are not described in detail, in order to highlight the purport of the disclosure.
Fig. 1 shows the flow chart of the optimization method of the neural network model according to the embodiment of the present disclosure.The neural network mould
The executing subject of the optimization method of type can be the optimization device of neural network model.For example, the optimization of the neural network model
Method can be executed by terminal device or server or other processing equipments.Wherein, terminal device can be user equipment (User
Equipment, UE), mobile device, user terminal, terminal, cellular phone, wireless phone, personal digital assistant (Personal
Digital Assistant, PDA), handheld device, calculate equipment, mobile unit or wearable device etc..The disclosure is implemented
Neural network model in example can be applied to the fields such as intelligent video analysis, safety monitoring or recognition of face.It is some can
In the implementation of energy, the optimization method of the neural network model can call the computer stored in memory by processor
The mode of readable instruction is realized.As shown in Figure 1, the method comprising the steps of S11 to step S14.
In step s 11, the selected part neuron from the full articulamentum before the output layer of student model.
In the embodiments of the present disclosure, student model and teacher's model are neural network model.Wherein, student model and old
Teacher's model can be identical neural network model, or different neural network models.In the embodiment of the present disclosure
Raw model can be used for the tasks such as processing feature extraction and/or Object identifying.Wherein, Object identifying can for person recognition or
Object identification.For example, person recognition can identify for recognition of face or pedestrian again；Object identification can be vehicle identification
Deng.Teacher's model in the embodiment of the present disclosure can be used for the tasks such as processing feature extraction and/or Object identifying.
In one possible implementation, it can be randomly selected from the full articulamentum before the output layer of student model
Partial nerve member.
In one possible implementation, the selected part nerve from the full articulamentum before the output layer of student model
Member, comprising: the selected part neuron from the last one full articulamentum before the output layer of student model.
It, can be from the full articulamentum of the last one before the output layer of student model as an example of the implementation
In randomly select partial nerve member.
In one possible implementation, the selected part nerve from the full articulamentum before the output layer of student model
Member, comprising: at least N number of neuron is chosen from the first full articulamentum, wherein the first full articulamentum indicates the output of student model
The full articulamentum of neuron selection is carried out before layer, the quantity for the neuron being selected in the first full articulamentum connects entirely less than first
The neuron population of layer is connect, N is equal to the neuron population of the first full articulamentum and the product of first threshold, and first threshold is greater than 0
And less than 1.For example, the output layer of student model before carry out neuron selection full articulamentum be student model output layer it
The full articulamentum of the last one preceding, then the last one full articulamentum before the first full articulamentum is the output layer of student model.
For example, first threshold is equal to 0.7.
In this implementation, by choosing at least N number of neuron from the first full articulamentum, thus, it is possible to avoid choosing
Very little neuron, so as to avoid due to caused by choosing very little neuron student model training difficulty greatly increase and/
Or student model the case where can not restraining.
In step s 12, input data is inputted to student model and teacher's model respectively, based on the output from student model
The partial nerve member chosen in full articulamentum before layer, obtains the output of student model.
In the embodiments of the present disclosure, input data can be training data.For example, when using pedestrian's weight identification mission training
When student model, input data can be the image of pedestrian.
In one possible implementation, the decilog for the output of decilog layer is exported.Wherein, decilog layer refers to
Logits layers, the decilog of decilog layer output refers to the logits of logits layers of output.
In alternatively possible implementation, the final output for neural network model is exported.For example, output is
Softmax layers of output.
In the embodiments of the present disclosure, pass through the selected part nerve from the full articulamentum before the output layer of student model
Member, and the defeated of student model is obtained based on the partial nerve member chosen from the full articulamentum before the output layer of student model
Out, thus, it is possible to mitigate the fitting degree of student model, student model over-fitting is avoided, so as to enhance the general of student model
Change ability.
In step s 13, whole neurons in the full articulamentum before output layer based on teacher's model, obtain teacher
The output of model.
In the embodiments of the present disclosure, the neuron of the full articulamentum of teacher's model is not chosen, is based on teacher's model
Whole neurons obtain the output of teacher's model.
In step S14, the output of output and teacher's model based on student model optimizes student model.
In one possible implementation, can output and teacher's model based on student model output, determine learn
The corresponding loss function of life model, and student model is optimized using the corresponding loss function of student model.In this implementation,
The corresponding loss function of student model can be used, and student model is optimized based on back-propagation algorithm.
In the embodiments of the present disclosure, based on the partial nerve chosen from the full articulamentum before the output layer of student model
Member obtains the output of student model, whole neurons in full articulamentum before the output layer based on teacher's model, obtains old
The output of teacher's model, and the output of the output based on student model and teacher's model optimize student model, and thus, it is possible to not increase
Add training data total amount and without stablizing the accuracy rate for improving neural network model, and for big portion under the premise of re -training
Neural network model and data is divided generally to be applicable in, and the part mind of the full articulamentum before the output layer by choosing student model
The output for effectively to be fitted student model through member, can effectively mitigate the over-fitting of neural network model, strength neural network mould
The generalization ability of type.
Fig. 2 shows the illustrative streams according to the optimization method step S11 of the neural network model of the embodiment of the present disclosure
Cheng Tu.As shown in Fig. 2, step S111 may include step S112.
In step S111, for each neuron in the first full articulamentum, one is generated in first interval respectively
Random number, wherein the first full articulamentum indicates that the output layer of student model carries out the full articulamentum of neuron selection before.
For example, first interval is [0,1].
In step S112, if the corresponding random number of peripheral sensory neuron in the first full articulamentum belongs to second interval,
Peripheral sensory neuron is chosen from the first full articulamentum, wherein second interval is the subset of first interval, and second interval is not equal to
First interval.
For example, second interval is [0,0.7].
In the embodiments of the present disclosure, whether belong to according to the corresponding random number of each neuron in the first full articulamentum
Two sections judge whether to choose each neuron in the first full articulamentum.For example, if a certain nerve in the first full articulamentum
The corresponding random number of member belongs to second interval, then chooses the neuron；If a certain neuron in the first full articulamentum is corresponding
Random number is not belonging to second interval, then does not choose the neuron.By for each neuron difference in the first full articulamentum
A random number is generated in first interval, if the corresponding random number of peripheral sensory neuron in the first full articulamentum belongs to the secondth area
Between, then peripheral sensory neuron is chosen from the first full articulamentum, if the corresponding random number of peripheral sensory neuron in the first full articulamentum
It is not belonging to second interval, then does not choose peripheral sensory neuron from the first full articulamentum, thus, it is possible to reach in the first full articulamentum
In randomly select the effect of partial nerve member.
Fig. 3 shows an illustrative stream of the optimization method step S14 according to the neural network model of the embodiment of the present disclosure
Cheng Tu.As shown in figure 3, step S14 may include step S141 to step S143.
In step s 141, the mean square error between the output of student model and the output of teacher's model is determined.
In the embodiments of the present disclosure, the defeated of the output of student model and teacher's model can be determined using related art method
Mean square error between out, details are not described herein.
In step S142, it is based on mean square error, obtains first-loss function.
It in one possible implementation, can be using mean square error as first-loss function.
It, can be using mean square error and the product of the first coefficient as first-loss letter in alternatively possible implementation
Number.
It should be noted that although the mode for describing to obtain first-loss function in a manner of implementation above is as above, this
Field technical staff it is understood that the disclosure answer it is without being limited thereto.Those skilled in the art can be according to practical application scene demand
And/or personal preference flexible setting obtains the specific implementation of first-loss function, as long as being based on mean square error
It can.
In step S143, using first-loss function optimization student model.
In one possible implementation, student model can be optimized using first-loss function and the first learning rate.
Fig. 4 shows the another exemplary of the optimization method step S14 according to the neural network model of the embodiment of the present disclosure
Flow chart.As shown in figure 4, step S14 may include step S144 to step S146.
In step S144, the relative entropy of the output of the output phase of student model for teacher's model is determined.
In the embodiments of the present disclosure, the output phase of student model can be determined for teacher's model using related art method
Output relative entropy, details are not described herein.
In step S145, it is based on relative entropy, obtains the second loss function.
It in one possible implementation, can be using the product of relative entropy and regulation coefficient as the second loss function.
It, can be using relative entropy as the second loss function in alternatively possible implementation.
It should be noted that although the mode for describing to obtain the second loss function in a manner of implementation above is as above, this
Field technical staff it is understood that the disclosure answer it is without being limited thereto.Those skilled in the art can be according to practical application scene demand
And/or personal preference flexible setting obtains the specific implementation of the second loss function, as long as being obtained based on relative entropy.
In step S146, student model is optimized using the second loss function.
In one possible implementation, student model can be optimized using the second loss function and the second learning rate.
Wherein, the second learning rate is less than the first learning rate.
Fig. 5 shows an illustrative stream of the optimization method step S145 according to the neural network model of the embodiment of the present disclosure
Cheng Tu.As shown in figure 5, step S145 may include step S1451 and step S1452.
In step S1451, the corresponding regulation coefficient of relative entropy is determined.
In one possible implementation, the corresponding regulation coefficient of relative entropy is greater than 1.
In step S1452, the product of relative entropy and regulation coefficient is determined as the second loss function.
By the way that the product of relative entropy and regulation coefficient is determined as the second loss function, can be avoided in back-propagation process
In since gradient is excessive lead to student model failure to train.
It should be noted that although the product of relative entropy and regulation coefficient is determined as the second loss function as example
It is as above to describe the mode that the second loss function is obtained based on relative entropy, it is understood by one of ordinary skill in the art that the disclosure is answered
It is without being limited thereto.Those skilled in the art can be based on opposite according to practical application scene demand and/or personal preference flexible setting
Entropy obtains the specific implementation of the second loss function.For example, the ratio of relative entropy and regulation coefficient can be determined as second
Loss function, wherein regulation coefficient is less than 1.
In the embodiments of the present disclosure, the parameter of teacher's model is fixed, and the parameter of student model is not fixed, that is, students are equal
In optimizable state, the parameter of student model is with training optimization.Hyper parameter during training, in student model
It can manually adjust.After training, student model be can be applied in other tasks different from training mission.
In the embodiments of the present disclosure, student model can be identical with teacher's model, and in other words, the embodiment of the present disclosure does not need
The very high student model of utility exercises supervision study, and the student model in the embodiment of the present disclosure can be steamed by knowledge
The method evaporated improves accuracy rate, and shows and stablize.
The embodiment of the present disclosure does not need to increase additional data, lower simultaneously for calculation amount and the requirement for calculating the time,
The performance of student model can be fast and effeciently promoted in the case where computing resource is less.
It is appreciated that above-mentioned each embodiment of the method that the disclosure refers to, without prejudice to principle logic,
To engage one another while the embodiment to be formed after combining, as space is limited, the disclosure is repeated no more.
It will be understood by those skilled in the art that each step writes sequence simultaneously in the above method of specific embodiment
It does not mean that stringent execution sequence and any restriction is constituted to implementation process, the specific execution sequence of each step should be with its function
It can be determined with possible internal logic.
In addition, the disclosure additionally provides optimization device, electronic equipment, the computer-readable storage medium of neural network model
Matter, program, the optimization method of the above-mentioned any neural network model that can be used to realize that the disclosure provides, corresponding technical solution
With description and referring to the corresponding record of method part, repeat no more.
Fig. 6 shows the block diagram of the optimization device according to the neural network model of the embodiment of the present disclosure.As shown in fig. 6, the mind
Optimization device through network model includes: to choose module 61, for selecting from the full articulamentum before the output layer of student model
Take partial nerve first；First determining module 62 is based on for input data to be inputted to student model and teacher's model respectively from
The partial nerve member chosen in full articulamentum before the output layer of raw model, obtains the output of student model；Second determines mould
Block 63 obtains the output of teacher's model for whole neurons in the full articulamentum before the output layer based on teacher's model；
Optimization module 64 optimizes student model for the output of output and teacher's model based on student model.
In one possible implementation, it chooses module 61 to be used for: last before the output layer of student model
Selected part neuron in a full articulamentum.
Fig. 7 shows an illustrative block diagram of the optimization device according to the neural network model of the embodiment of the present disclosure.Such as Fig. 7
It is shown:
In one possible implementation, choosing module 61 includes: to generate submodule 611, for connecting entirely for first
Each neuron in layer is connect, generates a random number in first interval respectively, wherein the first full articulamentum indicates student's mould
The full articulamentum of neuron selection is carried out before the output layer of type；Submodule 612 is chosen, if for the in the first full articulamentum
The corresponding random number of one neuron belongs to second interval, then chooses peripheral sensory neuron from the first full articulamentum, wherein the secondth area
Between be first interval subset, and second interval be not equal to first interval.
In one possible implementation, optimization module 64 includes: the first determining submodule 641, for determining student
Mean square error between the output of model and the output of teacher's model；Second determines submodule 642, for being based on mean square error,
Obtain first-loss function；First optimization submodule 643, for using first-loss function optimization student model.
In one possible implementation, optimization module 64 includes: that third determines submodule 644, for determining student
Relative entropy of the output phase of model for the output of teacher's model；4th determines submodule 645, for being based on relative entropy, obtains
Second loss function；Second optimization submodule 646, for optimizing student model using the second loss function.
In one possible implementation, the 4th determine that submodule 645 includes: the first determination unit, for determining phase
Regulation coefficient corresponding to entropy；Second determination unit, for the product of relative entropy and regulation coefficient to be determined as the second loss letter
Number.
In one possible implementation, the decilog for the output of decilog layer is exported.
In the embodiments of the present disclosure, based on the partial nerve chosen from the full articulamentum before the output layer of student model
Member obtains the output of student model, whole neurons in full articulamentum before the output layer based on teacher's model, obtains old
The output of teacher's model, and the output of the output based on student model and teacher's model optimize student model, and thus, it is possible to not increase
Add training data total amount and without stablizing the accuracy rate for improving neural network model, and for big portion under the premise of re -training
Neural network model and data is divided generally to be applicable in, and the part mind of the full articulamentum before the output layer by choosing student model
The output for effectively to be fitted student model through member, can effectively mitigate the over-fitting of neural network model, strength neural network mould
The generalization ability of type.
In some embodiments, the embodiment of the present disclosure provides the function that has of device or comprising module can be used for holding
The method of row embodiment of the method description above, specific implementation are referred to the description of embodiment of the method above, for sake of simplicity, this
In repeat no more.
The embodiment of the present disclosure also proposes a kind of computer readable storage medium, is stored thereon with computer program instructions, institute
It states when computer program instructions are executed by processor and realizes the above method.Computer readable storage medium can be non-volatile meter
Calculation machine readable storage medium storing program for executing.
The embodiment of the present disclosure also proposes a kind of electronic equipment, comprising: processor；For storage processor executable instruction
Memory；Wherein, the processor is configured to the above method.
The equipment that electronic equipment may be provided as terminal, server or other forms.
Fig. 8 is the block diagram of a kind of electronic equipment 800 shown according to an exemplary embodiment.For example, electronic equipment 800 can
To be mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, Medical Devices are good for
Body equipment, the terminals such as personal digital assistant.
Referring to Fig. 8, electronic equipment 800 may include following one or more components: processing component 802, memory 804,
Power supply module 806, multimedia component 808, audio component 810, the interface 812 of input/output (I/O), sensor module 814,
And communication component 816.
The integrated operation of the usual controlling electronic devices 800 of processing component 802, such as with display, call, data are logical
Letter, camera operation and record operate associated operation.Processing component 802 may include one or more processors 820 to hold
Row instruction, to perform all or part of the steps of the methods described above.In addition, processing component 802 may include one or more moulds
Block, convenient for the interaction between processing component 802 and other assemblies.For example, processing component 802 may include multi-media module, with
Facilitate the interaction between multimedia component 808 and processing component 802.
Memory 804 is configured as storing various types of data to support the operation in electronic equipment 800.These data
Example include any application or method for being operated on electronic equipment 800 instruction, contact data, telephone directory
Data, message, picture, video etc..Memory 804 can by any kind of volatibility or non-volatile memory device or it
Combination realize, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM) is erasable
Except programmable read only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, fastly
Flash memory, disk or CD.
Power supply module 806 provides electric power for the various assemblies of electronic equipment 800.Power supply module 806 may include power supply pipe
Reason system, one or more power supplys and other with for electronic equipment 800 generate, manage, and distribute the associated component of electric power.
Multimedia component 808 includes the screen of one output interface of offer between the electronic equipment 800 and user.
In some embodiments, screen may include liquid crystal display (LCD) and touch panel (TP).If screen includes touch surface
Plate, screen may be implemented as touch screen, to receive input signal from the user.Touch panel includes one or more touches
Sensor is to sense the gesture on touch, slide, and touch panel.The touch sensor can not only sense touch or sliding
The boundary of movement, but also detect duration and pressure associated with the touch or slide operation.In some embodiments,
Multimedia component 808 includes a front camera and/or rear camera.When electronic equipment 800 is in operation mode, as clapped
When taking the photograph mode or video mode, front camera and/or rear camera can receive external multi-medium data.It is each preposition
Camera and rear camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 810 is configured as output and/or input audio signal.For example, audio component 810 includes a Mike
Wind (MIC), when electronic equipment 800 is in operation mode, when such as call mode, recording mode, and voice recognition mode, microphone
It is configured as receiving external audio signal.The received audio signal can be further stored in memory 804 or via logical
Believe that component 816 is sent.In some embodiments, audio component 810 further includes a loudspeaker, is used for output audio signal.
I/O interface 812 provides interface between processing component 802 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include, but are not limited to: home button, volume button, start button and lock
Determine button.
Sensor module 814 includes one or more sensors, for providing the state of various aspects for electronic equipment 800
Assessment.For example, sensor module 814 can detecte the state that opens/closes of electronic equipment 800, the relative positioning of component, example
As the component be electronic equipment 800 display and keypad, sensor module 814 can also detect electronic equipment 800 or
The position change of 800 1 components of electronic equipment, the existence or non-existence that user contacts with electronic equipment 800, electronic equipment 800
The temperature change of orientation or acceleration/deceleration and electronic equipment 800.Sensor module 814 may include proximity sensor, be configured
For detecting the presence of nearby objects without any physical contact.Sensor module 814 can also include optical sensor,
Such as CMOS or ccd image sensor, for being used in imaging applications.In some embodiments, which may be used also
To include acceleration transducer, gyro sensor, Magnetic Sensor, pressure sensor or temperature sensor.
Communication component 816 is configured to facilitate the communication of wired or wireless way between electronic equipment 800 and other equipment.
Electronic equipment 800 can access the wireless network based on communication standard, such as WiFi, 2G or 3G or their combination.Show at one
In example property embodiment, communication component 816 receives broadcast singal or broadcast from external broadcasting management system via broadcast channel
Relevant information.In one exemplary embodiment, the communication component 816 further includes near-field communication (NFC) module, short to promote
Cheng Tongxin.For example, radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band can be based in NFC module
(UWB) technology, bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, electronic equipment 800 can be by one or more application specific integrated circuit (ASIC), number
Word signal processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, for executing the above method.
In the exemplary embodiment, a kind of non-volatile computer readable storage medium storing program for executing is additionally provided, for example including calculating
The memory 804 of machine program instruction, above-mentioned computer program instructions can be executed by the processor 820 of electronic equipment 800 to complete
The above method.
Fig. 9 is the block diagram of a kind of electronic equipment 1900 shown according to an exemplary embodiment.For example, electronic equipment 1900
It may be provided as a server.Referring to Fig. 9, electronic equipment 1900 includes processing component 1922, further comprise one or
Multiple processors and memory resource represented by a memory 1932, can be by the execution of processing component 1922 for storing
Instruction, such as application program.The application program stored in memory 1932 may include it is one or more each
Module corresponding to one group of instruction.In addition, processing component 1922 is configured as executing instruction, to execute the above method.
Electronic equipment 1900 can also include that a power supply module 1926 is configured as executing the power supply of electronic equipment 1900
Management, a wired or wireless network interface 1950 is configured as electronic equipment 1900 being connected to network and an input is defeated
(I/O) interface 1958 out.Electronic equipment 1900 can be operated based on the operating system for being stored in memory 1932, such as
Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or similar.
In the exemplary embodiment, a kind of non-volatile computer readable storage medium storing program for executing is additionally provided, for example including calculating
The memory 1932 of machine program instruction, above-mentioned computer program instructions can by the processing component 1922 of electronic equipment 1900 execute with
Complete the above method.
The disclosure can be system, method and/or computer program product.Computer program product may include computer
Readable storage medium storing program for executing, containing for making processor realize the computer-readable program instructions of various aspects of the disclosure.
Computer readable storage medium, which can be, can keep and store the tangible of the instruction used by instruction execution equipment
Equipment.Computer readable storage medium for example can be-- but it is not limited to-- storage device electric, magnetic storage apparatus, optical storage
Equipment, electric magnetic storage apparatus, semiconductor memory apparatus or above-mentioned any appropriate combination.Computer readable storage medium
More specific example (non exhaustive list) includes: portable computer diskette, hard disk, random access memory (RAM), read-only deposits
It is reservoir (ROM), erasable programmable read only memory (EPROM or flash memory), static random access memory (SRAM), portable
Compact disk read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical coding equipment, for example thereon
It is stored with punch card or groove internal projection structure and the above-mentioned any appropriate combination of instruction.Calculating used herein above
Machine readable storage medium storing program for executing is not interpreted that instantaneous signal itself, the electromagnetic wave of such as radio wave or other Free propagations lead to
It crosses the electromagnetic wave (for example, the light pulse for passing through fiber optic cables) of waveguide or the propagation of other transmission mediums or is transmitted by electric wire
Electric signal.
Computer-readable program instructions as described herein can be downloaded to from computer readable storage medium it is each calculate/
Processing equipment, or outer computer or outer is downloaded to by network, such as internet, local area network, wide area network and/or wireless network
Portion stores equipment.Network may include copper transmission cable, optical fiber transmission, wireless transmission, router, firewall, interchanger, gateway
Computer and/or Edge Server.Adapter or network interface in each calculating/processing equipment are received from network to be counted
Calculation machine readable program instructions, and the computer-readable program instructions are forwarded, for the meter being stored in each calculating/processing equipment
In calculation machine readable storage medium storing program for executing.
Computer program instructions for executing disclosure operation can be assembly instruction, instruction set architecture (ISA) instructs,
Machine instruction, machine-dependent instructions, microcode, firmware instructions, condition setup data or with one or more programming languages
The source code or object code that any combination is write, the programming language include the programming language-of object-oriented such as
Smalltalk, C++ etc., and conventional procedural programming languages-such as " C " language or similar programming language.Computer
Readable program instructions can be executed fully on the user computer, partly execute on the user computer, be only as one
Vertical software package executes, part executes on the remote computer or completely in remote computer on the user computer for part
Or it is executed on server.In situations involving remote computers, remote computer can pass through network-packet of any kind
It includes local area network (LAN) or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as benefit
It is connected with ISP by internet).In some embodiments, by utilizing computer-readable program instructions
Status information carry out personalized customization electronic circuit, such as programmable logic circuit, field programmable gate array (FPGA) or can
Programmed logic array (PLA) (PLA), the electronic circuit can execute computer-readable program instructions, to realize each side of the disclosure
Face.
Referring herein to according to the flow chart of the method, apparatus (system) of the embodiment of the present disclosure and computer program product and/
Or block diagram describes various aspects of the disclosure.It should be appreciated that flowchart and or block diagram each box and flow chart and/
Or in block diagram each box combination, can be realized by computer-readable program instructions.
These computer-readable program instructions can be supplied to general purpose computer, special purpose computer or other programmable datas
The processor of processing unit, so that a kind of machine is produced, so that these instructions are passing through computer or other programmable datas
When the processor of processing unit executes, function specified in one or more boxes in implementation flow chart and/or block diagram is produced
The device of energy/movement.These computer-readable program instructions can also be stored in a computer-readable storage medium, these refer to
It enables so that computer, programmable data processing unit and/or other equipment work in a specific way, thus, it is stored with instruction
Computer-readable medium then includes a manufacture comprising in one or more boxes in implementation flow chart and/or block diagram
The instruction of the various aspects of defined function action.
Computer-readable program instructions can also be loaded into computer, other programmable data processing units or other
In equipment, so that series of operation steps are executed in computer, other programmable data processing units or other equipment, to produce
Raw computer implemented process, so that executed in computer, other programmable data processing units or other equipment
Instruct function action specified in one or more boxes in implementation flow chart and/or block diagram.
The flow chart and block diagram in the drawings show system, method and the computer journeys according to multiple embodiments of the disclosure
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
One module of table, program segment or a part of instruction, the module, program segment or a part of instruction include one or more use
The executable instruction of the logic function as defined in realizing.In some implementations as replacements, function marked in the box
It can occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually be held substantially in parallel
Row, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram and/or
The combination of each box in flow chart and the box in block diagram and or flow chart, can the function as defined in executing or dynamic
The dedicated hardware based system made is realized, or can be realized using a combination of dedicated hardware and computer instructions.
The presently disclosed embodiments is described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or technological improvement to the technology in market for best explaining each embodiment, or lead this technology
Other those of ordinary skill in domain can understand each embodiment disclosed herein.
Claims (10)
1. a kind of optimization method of neural network model characterized by comprising
The selected part neuron from the full articulamentum before the output layer of student model；
Input data is inputted to the student model and teacher's model respectively, based on before the output layer of the student model
The partial nerve member chosen in full articulamentum, obtains the output of the student model；
Whole neurons in full articulamentum before output layer based on teacher's model, obtain the defeated of teacher's model
Out；
The output of output and teacher's model based on the student model, optimizes the student model.
2. the method according to claim 1, wherein being selected from the full articulamentum before the output layer of student model
Take partial nerve first, comprising:
The selected part neuron from the last one full articulamentum before the output layer of student model.
3. method according to claim 1 or 2, which is characterized in that the full articulamentum before the output layer of student model
Middle selected part neuron, comprising:
For each neuron in the first full articulamentum, a random number is generated in first interval respectively, wherein described the
One full articulamentum indicates that the output layer of the student model carries out the full articulamentum of neuron selection before；
If the corresponding random number of peripheral sensory neuron in the first full articulamentum belongs to second interval, connect entirely from described first
It connects and chooses the peripheral sensory neuron in layer, wherein the second interval is the subset of the first interval, and the second interval
Not equal to the first interval.
4. the method according to claim 1, which is characterized in that the output is that decilog layer exports
Decilog.
5. a kind of optimization device of neural network model characterized by comprising
Module is chosen, for the selected part neuron from the full articulamentum before the output layer of student model；
First determining module is based on for input data to be inputted to the student model and teacher's model respectively from the student
The partial nerve member chosen in full articulamentum before the output layer of model, obtains the output of the student model；
Second determining module is obtained for whole neurons in the full articulamentum before the output layer based on teacher's model
To the output of teacher's model；
Optimization module optimizes the student model for the output of output and teacher's model based on the student model.
6. device according to claim 5, which is characterized in that the selection module is used for:
The selected part neuron from the last one full articulamentum before the output layer of student model.
7. device according to claim 5 or 6, which is characterized in that the selection module includes:
Generate submodule, complete for for each neuron in the first articulamentum, generated in first interval respectively one with
Machine number, wherein the first full articulamentum indicates that the output layer of the student model carries out the full connection of neuron selection before
Layer；
Submodule is chosen, if the corresponding random number of peripheral sensory neuron in the described first full articulamentum belongs to second interval,
The peripheral sensory neuron then is chosen from the described first full articulamentum, wherein the second interval is the son of the first interval
Collection, and the second interval is not equal to the first interval.
8. the device according to any one of claim 5 to 7, which is characterized in that the output is that decilog layer exports
Decilog.
9. a kind of electronic equipment characterized by comprising
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to: perform claim require any one of 1 to 4 described in method.
10. a kind of computer readable storage medium, is stored thereon with computer program instructions, which is characterized in that the computer
Method described in any one of Claims 1-4 is realized when program instruction is executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811093361.XA CN109165738B (en) | 2018-09-19 | 2018-09-19 | Neural network model optimization method and device, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811093361.XA CN109165738B (en) | 2018-09-19 | 2018-09-19 | Neural network model optimization method and device, electronic device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109165738A true CN109165738A (en) | 2019-01-08 |
| CN109165738B CN109165738B (en) | 2021-09-14 |
Family
ID=64879584
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811093361.XA Active CN109165738B (en) | 2018-09-19 | 2018-09-19 | Neural network model optimization method and device, electronic device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109165738B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009052A (en) * | 2019-04-11 | 2019-07-12 | 腾讯科技（深圳）有限公司 | A kind of method of image recognition, the method and device of image recognition model training |
| CN110222705A (en) * | 2019-04-23 | 2019-09-10 | 华为技术有限公司 | A kind of training method and relevant apparatus of network model |
| CN110807434A (en) * | 2019-11-06 | 2020-02-18 | 威海若维信息科技有限公司 | Pedestrian re-identification system and method based on combination of human body analysis and coarse and fine particle sizes |
| WO2021047286A1 (en) * | 2019-09-12 | 2021-03-18 | 华为技术有限公司 | Text processing model training method, and text processing method and apparatus |
| CN112825143A (en) * | 2019-11-20 | 2021-05-21 | 北京眼神智能科技有限公司 | Deep convolutional neural network compression method, device, storage medium and equipment |
| WO2022052997A1 (en) * | 2020-09-09 | 2022-03-17 | Huawei Technologies Co.,Ltd. | Method and system for training neural network model using knowledge distillation |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030233335A1 (en) * | 2002-06-17 | 2003-12-18 | Mims Aj | Student neural network |
| CN106548190A (en) * | 2015-09-18 | 2017-03-29 | 三星电子株式会社 | Model training method and equipment and data identification method |
| CN106709565A (en) * | 2016-11-16 | 2017-05-24 | 广州视源电子科技股份有限公司 | Optimization method and device for neural network |
| CN106991440A (en) * | 2017-03-29 | 2017-07-28 | 湖北工业大学 | A kind of image classification algorithms of the convolutional neural networks based on spatial pyramid |
| CN107273976A (en) * | 2017-06-29 | 2017-10-20 | 广州日滨科技发展有限公司 | A kind of optimization method of neutral net, device, computer and storage medium |
| CN107341541A (en) * | 2016-04-29 | 2017-11-10 | 北京中科寒武纪科技有限公司 | A kind of apparatus and method for performing full articulamentum neural metwork training |
| CN108122003A (en) * | 2017-12-19 | 2018-06-05 | 西北工业大学 | A kind of Weak target recognition methods based on deep neural network |
| CN108140144A (en) * | 2016-03-31 | 2018-06-08 | 富士通株式会社 | A kind of method, apparatus being trained to neural network model and electronic equipment |
| US20180197425A1 (en) * | 2017-01-06 | 2018-07-12 | Washington State University | Self-monitoring analysis and reporting technologies |
-
2018
- 2018-09-19 CN CN201811093361.XA patent/CN109165738B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030233335A1 (en) * | 2002-06-17 | 2003-12-18 | Mims Aj | Student neural network |
| CN106548190A (en) * | 2015-09-18 | 2017-03-29 | 三星电子株式会社 | Model training method and equipment and data identification method |
| CN108140144A (en) * | 2016-03-31 | 2018-06-08 | 富士通株式会社 | A kind of method, apparatus being trained to neural network model and electronic equipment |
| CN107341541A (en) * | 2016-04-29 | 2017-11-10 | 北京中科寒武纪科技有限公司 | A kind of apparatus and method for performing full articulamentum neural metwork training |
| CN106709565A (en) * | 2016-11-16 | 2017-05-24 | 广州视源电子科技股份有限公司 | Optimization method and device for neural network |
| US20180197425A1 (en) * | 2017-01-06 | 2018-07-12 | Washington State University | Self-monitoring analysis and reporting technologies |
| CN106991440A (en) * | 2017-03-29 | 2017-07-28 | 湖北工业大学 | A kind of image classification algorithms of the convolutional neural networks based on spatial pyramid |
| CN107273976A (en) * | 2017-06-29 | 2017-10-20 | 广州日滨科技发展有限公司 | A kind of optimization method of neutral net, device, computer and storage medium |
| CN108122003A (en) * | 2017-12-19 | 2018-06-05 | 西北工业大学 | A kind of Weak target recognition methods based on deep neural network |
Non-Patent Citations (4)
| Title |
|---|
| JUNHO YIM等: ""A Gift from Knowledge Distillation: Fast Optimization, Network Minimization and Transfer Learning"", 《2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》 * |
| LIANG LU等: ""Knowledge Distillation for Small-footprint Highway Networks"", 《HTTPS://ARXIV.ORG/ABS/1608.00892》 * |
| YI WEI等: ""Quantization Mimic: Towards Very Tiny CNN for Object Detection"", 《HTTPS://ARXIV.ORG/ABS/1805.02152V3》 * |
| 赵胜伟等: ""基于增强监督知识蒸馏的交通标识分类"", 《中国科技论文》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009052A (en) * | 2019-04-11 | 2019-07-12 | 腾讯科技（深圳）有限公司 | A kind of method of image recognition, the method and device of image recognition model training |
| CN110009052B (en) * | 2019-04-11 | 2022-11-18 | 腾讯科技（深圳）有限公司 | Image recognition method, image recognition model training method and device |
| CN110222705A (en) * | 2019-04-23 | 2019-09-10 | 华为技术有限公司 | A kind of training method and relevant apparatus of network model |
| CN110222705B (en) * | 2019-04-23 | 2023-10-24 | 华为技术有限公司 | Training method of network model and related device |
| WO2021047286A1 (en) * | 2019-09-12 | 2021-03-18 | 华为技术有限公司 | Text processing model training method, and text processing method and apparatus |
| CN110807434A (en) * | 2019-11-06 | 2020-02-18 | 威海若维信息科技有限公司 | Pedestrian re-identification system and method based on combination of human body analysis and coarse and fine particle sizes |
| CN110807434B (en) * | 2019-11-06 | 2023-08-15 | 威海若维信息科技有限公司 | Pedestrian re-recognition system and method based on human body analysis coarse-fine granularity combination |
| CN112825143A (en) * | 2019-11-20 | 2021-05-21 | 北京眼神智能科技有限公司 | Deep convolutional neural network compression method, device, storage medium and equipment |
| WO2022052997A1 (en) * | 2020-09-09 | 2022-03-17 | Huawei Technologies Co.,Ltd. | Method and system for training neural network model using knowledge distillation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109165738B (en) | 2021-09-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109165738A (en) | Optimization method and device, electronic equipment and the storage medium of neural network model | |
| CN110210535A (en) | Neural network training method and device and image processing method and device | |
| CN109800744A (en) | Image clustering method and device, electronic equipment and storage medium | |
| CN109800737A (en) | Face recognition method and device, electronic equipment and storage medium | |
| CN109919300A (en) | Neural network training method and device and image processing method and device | |
| CN109829433A (en) | Facial image recognition method, device, electronic equipment and storage medium | |
| CN109614876A (en) | Critical point detection method and device, electronic equipment and storage medium | |
| CN110348537A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN110516745A (en) | Training method, device and the electronic equipment of image recognition model | |
| CN109658352A (en) | Optimization method and device, electronic equipment and the storage medium of image information | |
| CN109614613A (en) | The descriptive statement localization method and device of image, electronic equipment and storage medium | |
| EP3896587A1 (en) | Electronic device for performing user authentication and operation method therefor | |
| CN109543537A (en) | Weight identification model increment training method and device, electronic equipment and storage medium | |
| CN109635920A (en) | Neural network optimization and device, electronic equipment and storage medium | |
| CN109145970A (en) | Question and answer treating method and apparatus, electronic equipment and storage medium based on image | |
| CN109934275A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN109858614A (en) | Neural network training method and device, electronic equipment and storage medium | |
| CN110458218A (en) | Image classification method and device, sorter network training method and device | |
| CN109711546A (en) | Neural network training method and device, electronic equipment and storage medium | |
| CN109615006A (en) | Character recognition method and device, electronic equipment and storage medium | |
| CN109978891A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN109920016A (en) | Image generating method and device, electronic equipment and storage medium | |
| CN110245757A (en) | A kind of processing method and processing device of image pattern, electronic equipment and storage medium | |
| CN110458102A (en) | A kind of facial image recognition method and device, electronic equipment and storage medium | |
| CN109902738A (en) | Network module and distribution method and device, electronic equipment and storage medium |
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 |