CN108897045A - Deep learning model training method and seismic data noise attenuation method, device and equipment - Google Patents
Deep learning model training method and seismic data noise attenuation method, device and equipment Download PDFInfo
- Publication number
- CN108897045A CN108897045A CN201810985408.7A CN201810985408A CN108897045A CN 108897045 A CN108897045 A CN 108897045A CN 201810985408 A CN201810985408 A CN 201810985408A CN 108897045 A CN108897045 A CN 108897045A
- Authority
- CN
- China
- Prior art keywords
- data
- deep learning
- training
- noise
- model
- 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.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 163
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000013136 deep learning model Methods 0.000 title claims abstract description 22
- 238000013135 deep learning Methods 0.000 claims abstract description 83
- 230000006870 function Effects 0.000 claims abstract description 60
- 238000012545 processing Methods 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 9
- 230000008569 process Effects 0.000 description 12
- 238000005457 optimization Methods 0.000 description 8
- 238000013480 data collection Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000004590 computer program Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 239000000047 product Substances 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 230000003139 buffering effect Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011017 operating method Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 239000007795 chemical reaction product Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/36—Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/30—Noise handling
- G01V2210/32—Noise reduction
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The present invention provides a kind of deep learning model training method and seismic data noise attenuation method, device and equipment, wherein deep learning model training method includes:Obtain training sample data, wherein the training sample data include:Noise data in original earthquake data and the original earthquake data;The training sample data are input in deep learning network model and are trained, until the loss function value of the deep learning network model meets preset requirement;The loss function value of the deep learning network model is met to training pattern when preset requirement, the seismic data noise attenuation model obtained as training.In embodiments of the present invention, by being trained to deep learning network model, the seismic data noise attenuation model that training is obtained adaptively identifies the effect of various horizontal noises, do not need artificially to give noise-removed threshold value, enable seismic data noise attenuation it is more acurrate, it is efficient, intelligently realize.
Description
Technical field
The present invention relates to seismic data processing technology field, in particular to a kind of deep learning model training method and earthquake
Data de-noising method, device and equipment.
Background technique
Seismic prospecting is the basis of oil and gas development, mainly has data sampling and processing, explains three big links, each link
Mass data can be generated, existing useful signal in these data, while also being interfered comprising much noise, since the presence of noise makes
It obtains effective seismic data to be difficult to be identified, therefore removal noise jamming is in seismic processing chain in practical applications
Important step.
In order to remove the noise jamming in seismic data, a variety of denoising modes are currently provided, such as:Time-domain is to space
Domain, frequency domain to wave-number domain, frequency domain to bent wave zone etc. all establish different earthquake denoising models, and apply dictionary, dilute
The methods of expression is dredged to carry out the denoising of earthquake.
However, these above-mentioned denoising modes all assume that seismic data is low-rank and can linearly express in higher dimensional space
On the basis of seismic data, it is just able to achieve the purpose of denoising.Further, it needs that denoising threshold is manually set during denoising
Value, implements complex, and efficiency is lower.
In view of the above-mentioned problems, currently no effective solution has been proposed.
Summary of the invention
The present invention provides a kind of deep learning model training method and seismic data noise attenuation method, device and equipment, with
Realize the effect for efficiently, intelligently carrying out seismic data noise attenuation.
The embodiment of the invention provides a kind of deep learning model training method, this method includes:Obtain number of training
Include according to, wherein training sample data:Noise data in original earthquake data and the original earthquake data;By institute
It states training sample data and is input in deep learning network model and be trained, until the loss of the deep learning network model
Functional value meets preset requirement;The loss function value of the deep learning network model is met to training mould when preset requirement
Type, the seismic data noise attenuation model obtained as training.
In one embodiment, the loss function is:
Wherein, l (W, b) indicates that loss function, W are the weight in the deep learning network model;B is the depth
Practise the biasing in network model;‖·‖FFor F norm, the number of non-zero element in vector is indicated as F=0, is indicated as F=1
The sum of each element absolute value in vector indicates the value of the quadratic sum extraction of square root of vector each element as F=2;N is the training
The total number of sample data, n are positive integer;I is a variable, and value range is the integer from 1 to n;F(xi) it is i-th
The noise data of training sample data input deep learning network model reality output;For in i-th of training sample data not
The secondary power of the seismic data of Noise;xiFor noise-containing seismic data in i-th of training sample data;As
Noise data in training sample data.
In one embodiment, there is N layer network in the deep learning network model, wherein be provided in a layer network
One channel can jump to N-a+1 layer network from a layer network by the channel information, wherein N is positive integer, and a is not
Positive integer greater than N.
The embodiment of the present invention also provides a kind of seismic data noise attenuation method, including:
Obtain seismic data to be denoised;By the seismic data input denoising model that training obtains in advance to be denoised
In, obtain the noise data in the seismic data to be denoised;By the seismic data to be denoised and the noise data
Difference operation is done, the seismic data after being denoised.
In one embodiment, the denoising model is the model obtained based on deep learning network training.
In one embodiment, training obtains the denoising model in the following way:
Obtain training sample data, wherein the training sample data include:Original earthquake data and it is described primitively
Shake the noise data in data;The training sample data are input in deep learning network model and are trained, institute is obtained
State denoising model.
The embodiment of the present invention also provides a kind of seismic data noise attenuation device, including:Module is obtained, earthquake to be denoised is obtained
Data;Input module obtains in the seismic data input denoising model that training obtains in advance to be denoised described wait go
The noise data in seismic data made an uproar;The seismic data to be denoised and the noise data are done difference by processing module
Operation, the seismic data after being denoised.
In one embodiment, the denoising model is the model obtained based on deep learning network training.
In one embodiment, training obtains the denoising model in the following way:Training sample data are obtained,
In, the training sample data include:Noise data in original earthquake data and the original earthquake data;By the instruction
White silk sample data, which is input in deep learning network model, to be trained, and the denoising model is obtained.
The embodiment of the present invention also provides a kind of deep learning model training apparatus, including:Module is obtained, training sample is obtained
Data, wherein the training sample data include:Noise data in original earthquake data and the original earthquake data;Instruction
Practice module, the training sample data are input in deep learning network model and are trained, until the deep learning net
The loss function value of network model meets preset requirement;Processing module expires the loss function value of the deep learning network model
Training pattern when sufficient preset requirement, the seismic data noise attenuation model obtained as training.
In one embodiment, the loss function is:
Wherein, l (W, b) indicates that loss function, W are the weight in the deep learning network model;B is the depth
Practise the biasing in network model;‖·‖FFor F norm, the number of non-zero element in vector is indicated as F=0, is indicated as F=1
The sum of each element absolute value in vector indicates the value of the quadratic sum extraction of square root of vector each element as F=2;N is the training
The total number of sample data, n are positive integer;I is a variable, and value range is the integer from 1 to n;F(xi) it is i-th
The noise data of training sample data input deep learning network model reality output;For in i-th of training sample data not
The secondary power of the seismic data of Noise;xiFor noise-containing seismic data in i-th of training sample data;As
Noise data in training sample data.
In one embodiment, the training sample data are input to training in deep learning network model, including:
N layer network is shared in deep learning network model;
A layer network is provided with one article of channel N-2a layer network that directly jumps and is connected in the deep learning network model
N-a+1 layer network, so that information is easier to transmit;
Wherein, N is positive integer, and a is the positive integer no more than N.
The embodiment of the present invention also provides a kind of seismic data noise attenuation equipment, including processor and can for storage processor
The step of memory executed instruction, the processor realizes the seismic data noise attenuation method when executing described instruction.
The embodiment of the present invention also provides a kind of computer readable storage medium, is stored thereon with computer instruction, the finger
Order is performed the step of realizing the seismic data noise attenuation method.
In embodiments of the present invention, by being trained to deep learning network model, so that the earthquake number that training obtains
The effect that various horizontal noises can be adaptively identified according to denoising model, does not need artificially to give noise-removed threshold value, so that ground
Shake data de-noising more acurrate, efficient, can be realized intelligently.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The some embodiments recorded in application, for those of ordinary skill in the art, in the premise of not making the creative labor property
Under, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow chart according to a kind of seismic data noise attenuation method of the embodiment of the present application;
Fig. 2 is the Basic architecture figure of depth residual error network;
Fig. 3 a is to be carried out according to a kind of seismic data noise attenuation method of the embodiment of the present application based on depth residual error learning model
The noise-containing seismic data sectional view of denoising;
Fig. 3 b be according to a kind of seismic data noise attenuation method of the embodiment of the present application be based on depth residual error learning model into
Noise data sectional view in the noise-containing seismic data of row denoising;
Fig. 3 c be according to a kind of seismic data noise attenuation method of the embodiment of the present application be based on depth residual error learning model into
Seismic data sectional view after the denoising of row denoising;
Fig. 4 is the flow chart according to a kind of deep learning model training method of the embodiment of the present application;
Fig. 5 a be according to a kind of deep learning model training method of the embodiment of the present application to depth residual error learning model into
The noise-containing seismic data sectional view of row training;
Fig. 5 b be according to a kind of deep learning model training method of the embodiment of the present application to depth residual error learning model into
Noise data sectional view in the noise-containing seismic data of row training;
Fig. 5 c be according to a kind of deep learning model training method of the embodiment of the present application to depth residual error learning model into
The row obtained loss function of training with the number of iterations change curve;
Fig. 6 is the structural block diagram according to a kind of seismic data noise attenuation device of the embodiment of the present application;
Fig. 7 is the structural block diagram according to a kind of deep learning model training apparatus of the embodiment of the present application;
Fig. 8 is the structural schematic diagram according to a kind of seismic data noise attenuation electronic equipment of the embodiment of the present application.
Specific embodiment
The principle and spirit of the invention are described below with reference to several illustrative embodiments.It should be appreciated that providing this
A little embodiments are used for the purpose of making those skilled in the art can better understand that realizing the present invention in turn, and be not with any
Mode limits the scope of the invention.On the contrary, it is more thorough and complete to these embodiments are provided so that the application discloses, and
And the scope of the present disclosure can be completely communicated to those skilled in the art.
It will be apparent to one skilled in the art that embodiments of the present invention can be implemented as a kind of system, appliance arrangement, method
Or computer program product.Therefore, disclose can be with specific implementation is as follows, i.e., by the application:It is complete hardware, complete soft
The form that part (including firmware, resident software, microcode etc.) or hardware and software combine.
In view of needing artificially to give noise-removed threshold value, larger workload and standard in existing seismic data noise attenuation technology
The problem of true property not can guarantee, the application passes through training deep learning network, and is used for ground for trained deep learning network
Data de-noising is shaken, the effect for efficiently, intelligently carrying out seismic data noise attenuation may be implemented.Specifically, in the present embodiment, mentioning
A kind of seismic data noise attenuation method is gone out, as shown in Figure 1, may comprise steps of:
S101:Obtain seismic data to be denoised.
Seismic prospecting refers to that elastic wave caused by artificial excitation utilizes the difference of underground medium elasticity and density, passes through sight
It surveys and the seismic wave of analysis artificial earthquake generation is in the propagation law of underground, infer the property of subterranean strata and the earth object of form
Manage exploitation method, therefore available mass data during seismic prospecting, existing useful signal in the mass data of acquisition,
It simultaneously also include a large amount of interfering noise signals, these noise-containing seismic datas can be used as seismic data to be denoised.
Further, these seismic datas to be denoised can be the seismic data obtained in real time, be also possible in advance
It is acquired to arrive, store seismic data in the database.
S102:By in the seismic data input denoising model that training obtains in advance to be denoised, obtain described wait go
The noise data in seismic data made an uproar.
That is, training has obtained denoising model in advance, in this way using the seismic data to be denoised of above-mentioned acquisition as denoising mould
The input data of type is input in preparatory trained denoising model, so that it may which identification obtains in seismic data to be denoised
Noise data, that is to say, that the output of the denoising model is the noise data in seismic data to be denoised.
Wherein, above-mentioned denoising model can be is obtained based on the training of deep learning network model, and can be passed through
Following training method obtains denoising model:
S1:Training sample data are obtained, which may include:Noise-containing original earthquake data is gone
The noise data in seismic data and original earthquake data after making an uproar;
S2:The training sample data that will acquire, which are input in deep learning network model, to be trained, by multiple
Iteration optimization, training obtain denoising model.
Wherein, above-mentioned training sample can be and carry out the seismic data of denoising in advance, these data can be with
As training sample in a manner of data pair, what each data centering included is:Original earthquake data, and, original earthquake data
In noise data.Certainly, noise data therein be also possible to subtract by original earthquake data denoising seismic data it
It obtains afterwards.Multiple data are trained to being input in deep learning network model as training sample, so that it may obtain
Denoising model.
In order to enable the denoising model that training obtains can be continued to optimize, new training sample can be obtained always, and
New training sample is input in model and is trained, so that denoising model can be optimised always.
Wherein, above-mentioned deep learning network model can use depth residual error learning model, as shown in Fig. 2, depth residual error
Learning model (Deep Residual Learning, ResNet) assumes that the input of certain section of learning network is x, and desired output is
H (x), i.e. H (x) are the desired potential mappings of complexity, but learning difficulty is big;If input x is directly passed to output as initial
As a result, by " quick connection (shortcut connections) " in Fig. 2, then the target for needing to learn at this time is exactly F
(x)=H (x)-x, then ResNet, which is equivalent to, changes learning objective, is no longer study one complete output, but most
The difference of excellent solution H (X) and congruence mapping x, i.e. residual error F (x)=H (x)-x.No weight is fast connected, each module after x is transmitted
Only study residual error F (x), so that network stabilization and being easy to learn.In Fig. 2, Weight layer is weight layer, Identity is
Identical mapping, Relu are line rectification function, are common activation primitives in deep learning network.
Wherein, above-mentioned residual error can be understood as the difference between actual observation value and estimated value.It is basic related model has been contained
The important information of hypothesis, the thought of residual error is exactly to remove identical main part, so that small variation is protruded, so that depth
Habit becomes simple.
S103:The seismic data to be denoised and the noise data are done into difference operation, the earthquake after being denoised
Data.
In one embodiment of the invention, the input data of above-mentioned denoising model is seismic data to be denoised, output
Data are the noise data in seismic data to be denoised.It, can after obtaining the noise data in seismic data to be denoised
With seismic data and noise data progress difference operation that will be to be denoised, to obtain effective seismic data of not Noise.
Above-mentioned seismic data noise attenuation method is specifically described below with reference to a specific embodiment, in the present embodiment
In, denoising is carried out to the original earthquake data obtained in seismic prospecting process by above-mentioned seismic data noise attenuation method,
It is important to note, however, that the specific embodiment merely to the present invention is better described, is not constituted to of the invention improper
It limits.
The original earthquake data obtained in seismic prospecting process is denoised by above-mentioned seismic data noise attenuation method
Processing may comprise steps of:
S11:During seismic prospecting, the noise-containing seismic data that will acquire is stored;
S12:The noise-containing seismic data that will acquire is input to the denoising mould that training obtains in advance as input data
Noise number in type, by the processing for the denoising model that the preparatory training obtains, in available noise-containing seismic data
According to, wherein noise-containing seismic data section is as shown in Figure 3a, and the noise data section in noise-containing seismic data is such as
Shown in Fig. 3 b;
S13:Noise data in noise-containing seismic data and noise-containing seismic data as shown in Figure 3b carries out
Difference operation, the seismic data after available denoising, wherein the seismic data section after denoising is as shown in Figure 3c.Using this
The seismic data noise attenuation method proposed is invented, without artificially giving noise-removed threshold value during denoising, so that it may adaptively
The noise for identifying various levels, effectively improves the accuracy and efficiency of seismic data noise attenuation.
A kind of deep learning model training method is additionally provided in the embodiment of the present invention, as shown in figure 4, may include following
Step:
S401:Obtain training sample data collection, wherein individualized training sample data may include:The earthquake number of Noise
According to the noise data in the seismic data of, Noise.
Above-mentioned training sample data collection can be obtained from earthquake big data, and training sample data is made to have diversified spy
Property;Noise data in the seismic data of above-mentioned Noise can be used as the label data in training process.
S402:Training sample data are input in deep learning network model and are trained, until deep learning network
The loss function value of model meets preset requirement.
Above-mentioned training sample data can be used as input data and be input in model to be trained, and being somebody's turn to do model to be trained can
To use deep learning network model, before training starts, each numerical value in above-mentioned deep learning network model is carried out initial
Change setting.Wherein, the output data of above-mentioned deep learning network model is the ground of Noise in the training sample data identified
The noise data in data is shaken, which is compared with label data, and calculate loss function value, loss function table
The similarity degree between the output data and label data of deep learning network model is reached, the smaller expression similarity of value is higher, root
Iteration optimization is set for each numerical value in deep learning network model according to comparison result.
For example, above-mentioned loss function can be expressed as:
Wherein, l (W, b) indicates loss function, and W is weight;B is biasing, and it is positive and negative that biasing size has measured neuron generation
The complexity of excitation;‖·‖FFor F norm, the number of non-zero element in vector is indicated as F=0, indicates vector as F=1
The sum of middle each element absolute value indicates the value of the quadratic sum extraction of square root of vector each element as F=2;N is training sample data
Total number, n is positive integer;I is a variable, and value range is the integer from 1 to n;F(xi) it is i-th of training sample
Pass through the Feature Mapping of deep learning network model, the i.e. noise data of deep learning network model reality output in data;
For the not secondary power of the seismic data of Noise in i-th of training sample data;xiIt makes an uproar to contain in i-th of training sample data
The seismic data of sound;Noise data as in training sample data.
Wherein, as ‖ ‖FWhen for 0 norm, if we with 0 norm come one parameter matrix of regularization, be exactly uncommon
The most elements for hoping the parameter matrix are all 0, that is to say, that the parameter matrix is allowed to be sparse;As ‖ ‖FWhen for 1 norm,
Also may be implemented it is sparse, but 1 norm because have Optimization Solution characteristic more better than 0 norm and be widely used;As ‖ ‖F
When for 2 norm, the limitation to the model space may be implemented, to avoid over-fitting to a certain extent, over-fitting is exactly to learn
Ability is too strong, so that the less general characteristic for being included training sample all learns to arrive.
Wherein, above-mentioned loss function can be understood as:There can be a target letter in each algorithm of usual machine learning
Number, the solution procedure of algorithm is by the process to this objective function optimization.It is usually used in classification or regression problem
Loss function (cost function) is used as its objective function.Loss function is different for the predicted value and true value of evaluation model
Degree, loss function value is smaller, and the performance of usual model is better.
It should be noted, however, that above-mentioned cited loss function is only a kind of exemplary illustration, what is actually realized
When, this application can also be not construed as limiting using other functions as loss function, it can according to actual needs and feelings
Condition selects suitable loss function to carry out deep learning model training.
In order to enable network is more stable, deep learning process is highly efficient, in this example, it is contemplated that network can be used
The mode of layer jump carries out model training, such as:N layer network is shared in deep learning network model, it can be in deep learning net
Be provided with interface channel in network model, by the interface channel directly can directly jump from a layer network N-2a layer network connect
It is connected to N-a+1 layer network, wherein N is positive integer, and a is the positive integer no more than N.
S403:The loss function value of above-mentioned deep learning network model is met to training pattern when preset requirement, as
The seismic data noise attenuation model that training obtains.
Because deep learning model training is the process of the calculating that iterates, model knowledge not can guarantee in this process
It is not the result is that absolutely accurate, but it is the need to ensure that the accuracy of the recognition result of model training meets a preset condition
And it is just available when no longer improving.Therefore, after the training for having carried out certain number, if loss function value occurs one completely
The minimum of sufficient preset requirement, and not reduced in scheduled buffering the number of iterations internal loss functional value, i.e. model training
When the accuracy of recognition result does not improve, it is believed that the loss function value of deep learning network model tends towards stability, then
Just illustrate that deep learning network model has been trained to expected denoising level, error within the acceptable range, can will damage
Model when function reaches minimum is lost as seismic data noise attenuation model.If in scheduled the number of iterations internal loss functional value
Do not reach preset requirement, then then continuing iteration, until the loss function value that training obtains meets preset requirement.
Wherein, total the number of iterations and buffering the number of iterations can be wanted to reach according to the quantity and training of training sample data
To precision requirement determine that the application is not construed as limiting this
Above-mentioned deep learning model training method is specifically described below with reference to a specific embodiment, in this reality
It applies in example, depth residual error learning model is trained by above-mentioned deep learning model training method, however it is noticeable
It is that the specific embodiment does not constitute improper limitations of the present invention merely to the present invention is better described.
Depth residual error learning model is trained by above-mentioned deep learning model training method, may include following step
Suddenly:
S21:Training sample data collection is collected by earthquake big data, wherein individualized training sample data includes:Containing making an uproar
Noise data in the seismic data of sound and the seismic data containing noise data, making an uproar in above-mentioned noise-containing seismic data
Sound data can be used as label data in the training process, wherein noise-containing seismic data section is as shown in Figure 5 a, contains
Noise data section in the seismic data of noise is as shown in Figure 5 b;
S22:Noise-containing seismic data and label data are input in depth residual error learning model to be trained
It is trained, initializing set is carried out to each numerical value in above-mentioned depth residual error learning model to be trained first.It will training sample
Notebook data is input in depth residual error learning model to be trained, and the output data of depth residual error learning model is the instruction identified
Practice the noise data in sample data in the seismic data of Noise, which is compared with label data, and counts
Calculate the loss function in this training.
S23:After the iteration optimization multiple to loss function, the value of loss function meets preset requirement, deep at this time
Degree residual error learning model has been trained to reach expected denoising level, and error within the acceptable range, complete by training.Wherein,
The loss function of depth residual error learning model can be as shown in Figure 5 c with the change curve of the number of iterations, and abscissa indicates iteration time
Number, ordinate indicate the error amount being calculated by loss function.
A kind of seismic data noise attenuation device is additionally provided in the embodiment of the present invention, as shown in fig. 6, the seismic data noise attenuation fills
It sets and may include:Module 601, input module 602 are obtained, processing module 603 is below illustrated the structure.
Module 601 is obtained, can be used for obtaining seismic data to be denoised, using the seismic data to be denoised as ground
Shake the input data of data de-noising device.
Input module 602 can be used for the above-mentioned seismic data input denoising model that training obtains in advance to be denoised
In, data, by successively transformation, are sent to output layer, obtain the noise data in seismic data to be denoised from input layer.
In one embodiment, above-mentioned denoising model can be is obtained based on the training of deep learning network model, this hair
The depth residual error learning model that bright middle deep learning network can be proposed using doctor He Kaiming, but the present invention is to deep learning
The type of network model is not specifically limited.
In one embodiment, it can train in the following way and obtain denoising model:
Training sample data collection can be obtained from earthquake big data, wherein individualized training sample data may include noisy
Noise data in the seismic data of sound, the seismic data of Noise, wherein in individualized training sample data Noise earthquake
Noise data in data can be used as label data;Initializing set is carried out to numerical value each in deep learning network model;From
Training sample data concentration randomly selects one group of data and is input to deep learning network model, and provides target output data, i.e.,
Label data in training sample data;The reality output data of deep learning network model are compared with target output data
Compared with, and loss function value is calculated, the output data and the phase between label data that loss function expresses deep learning network model
Like degree, the smaller expression similarity of value is higher, is set for according to comparison result to each numerical value in deep learning network model
Iteration optimization;After the training for having carried out certain number, judge whether loss function value touches the mark, if conditions are not met, then after
It is continuous to be iterated optimization, if it is satisfied, then the set amount of each numerical value in finally obtained deep learning network model is protected
It deposits to get trained denoising model is arrived.
Processing module 603 can be used for the noise data of seismic data and denoising model output to be denoised doing difference
Operation, the seismic data after being denoised.
A kind of deep learning model training apparatus is additionally provided in the embodiment of the present invention, as shown in fig. 7, the deep learning mould
Type training device may include:Module 701, training module 702 are obtained, processing module 703 is below illustrated the structure.
Module 701 is obtained, can be used for obtaining training sample data collection from earthquake big data, there is sample data more
Sample.Wherein training sample data may include the noise data in the seismic data of the seismic data of Noise, Noise, on
Stating the noise data in the seismic data of Noise can be used as label data in training process.
Training module 702, can be used for for training sample data being input in deep learning network model and is trained.Contain
Noise data in the seismic data of noise and the seismic data of Noise is used as the input number of deep learning network model
According to the output of deep learning network model is the noise data in the seismic data of the Noise identified, by the output data
Be compared with label data, and calculate loss function, loss function express the output data of deep learning network model with
Similarity degree between label data, the smaller expression similarity of value is higher, according to comparison result to each in deep learning network model
Numerical value is set for iteration optimization.
Processing module 703, when can be used for the loss function value of above-mentioned deep learning network model meeting preset requirement
Training pattern, as the obtained seismic data noise attenuation model of training.
After the training for having carried out certain number, when the loss function value of deep learning network model meets preset requirement
When, illustrate that deep learning network model has been trained to expected denoising level, error within the acceptable range, is trained at this time
Terminate, the seismic data noise attenuation model for meeting desired effect can be obtained, the seismic data noise attenuation model that training obtains can be certainly
Adaptively identify the noise of various levels.
The application, specifically can be refering to shown in Fig. 8 embodiment further provides a kind of seismic data noise attenuation electronic equipment
Electronic equipment composed structure schematic diagram based on seismic data noise attenuation method provided by the embodiments of the present application, the electronic equipment tool
Body may include input equipment 81, processor 82, memory 83.Wherein, specifically can be used for will be wait go for the input equipment 81
The seismic data made an uproar inputs in trained equipment in advance;The processor 82 specifically can be used for the ground to be processed to input
Shake data are handled, and result is exported;The memory 83 specifically can be used for storing the number generated during processing
According to.
In the present embodiment, the input equipment, which specifically can be, carries out information exchange between user and computer system
One of main device.The input equipment may include keyboard, mouse, camera, scanner, light pen, writing input board, language
Sound input unit etc.;Input equipment is used to initial data be input in computer with the programs for handling these numbers.The input
Equipment, which can also obtain, receives the data that other modules, unit, equipment transmit.The processor can be by any appropriate
Mode is realized.For example, processor can take such as microprocessor or processor and storage that can be executed by (micro-) processor
Computer readable program code (such as software or firmware) computer-readable medium, logic gate, switch, specific integrated circuit
(Application Specific Integrated Circuit, ASIC), programmable logic controller (PLC) and insertion microcontroller
Form etc..The storage implement body can be in modern information technologies for protecting stored memory device.The storage
Device may include many levels, in digital display circuit, as long as can save binary data can be memory;In integrated electricity
The circuit with store function of Lu Zhong, a not no physical form are also memory, such as RAM, FIFO;In systems, have
There is the storage equipment of physical form to be also memory, such as memory bar, TF card.
In the present embodiment, the function and effect of electronic equipment specific implementation, can compare with other embodiment
It explains, details are not described herein.
A kind of computer storage medium based on seismic data noise attenuation method is additionally provided in the application embodiment, it is described
Computer storage medium is stored with computer program instructions, is performed realization in the computer program instructions:Adaptively
It identifies the noise data in noise-containing seismic data, and removes noise data in noise-containing seismic data, output is not
The seismic data of Noise.
In the present embodiment, above-mentioned storage medium includes but is not limited to random access memory (Random Access
Memory, RAM), read-only memory (Read-Only Memory, ROM), caching (Cache), hard disk (Hard Disk
Drive, HDD) or storage card (Memory Card).The memory can be used for storing computer program instructions.Network is logical
Letter unit can be according to standard setting as defined in communication protocol, for carrying out the interface of network connection communication.
In the present embodiment, the function and effect of the program instruction specific implementation of computer storage medium storage, can
To compare explanation with other embodiment, details are not described herein.
In above description, it can be seen that the embodiment of the present invention realizes following technical effect:By to deep learning network
Model is trained, and the seismic data noise attenuation model that training is obtained adaptively identifies the effect of various horizontal noises
Fruit does not need artificially to give noise-removed threshold value so that seismic data noise attenuation it is more acurrate, efficient, intelligently realize.
Although this application provides the method operating procedure as described in embodiment or flow chart, based on conventional or noninvasive
The means for the property made may include more or less operating procedure.The step of enumerating in embodiment sequence is only numerous steps
One of execution sequence mode, does not represent and unique executes sequence.It, can be with when device in practice or end product execute
It is executed according to embodiment or method shown in the drawings sequence or parallel executes (such as parallel processor or multiple threads
Environment, even distributed data processing environment).The terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that process, method, product or equipment including a series of elements are not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, product or equipment
Intrinsic element.In the absence of more restrictions, be not precluded include the process, method of the element, product or
There is also other identical or equivalent elements in person's equipment.
Unit, device or module that above-described embodiment illustrates etc. can specifically realize by computer chip or entity, or
It is realized by the product with certain function.For convenience of description, various modules point are divided into function when describing apparatus above
It does not describe.It certainly, when implementing the application can the function of each module is real in the same or multiple software and or hardware
It is existing, the module for realizing same function can also be realized by the combination of multiple submodule or subelement etc..Dress described above
Set that embodiment is only schematical, for example, the division of the unit, only a kind of logical function partition, in actual implementation
There may be another division manner, such as multiple units or components can be combined or can be integrated into another system or one
A little features can be ignored, or not execute.Another point, shown or discussed mutual coupling or direct-coupling or communication link
Connecing can be through some interfaces, the indirect coupling or communication connection of device or unit, can be electrical property, mechanical or other shapes
Formula.
Obviously, those skilled in the art should be understood that each module of the above-mentioned embodiment of the present invention or each step can be with
It is realized with general computing device, they can be concentrated on a single computing device, or be distributed in multiple computing devices
On composed network, optionally, they can be realized with the program code that computing device can perform, it is thus possible to by it
Store and be performed by computing device in the storage device, and in some cases, can be held with the sequence for being different from herein
The shown or described step of row, perhaps they are fabricated to each integrated circuit modules or will be multiple in them
Module or step are fabricated to single integrated circuit module to realize.In this way, the embodiment of the present invention be not limited to it is any specific hard
Part and software combine.
The application can be used in numerous general or special purpose computing system environments or configuration.Such as:Personal computer, clothes
Business device computer, handheld device or portable device, laptop device, multicomputer system, microprocessor-based system, set
Top box, programmable electronic equipment, network PC, minicomputer, mainframe computer, point including any of the above system or equipment
Cloth calculates environment etc..
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the embodiment of the present invention can have various modifications and variations.All within the spirits and principles of the present invention, made
Any modification, equivalent substitution, improvement and etc. should all be included in the protection scope of the present invention.
Claims (14)
1. a kind of deep learning model training method, which is characterized in that including:
Obtain training sample data, wherein the training sample data include:Original earthquake data and the original earthquake number
Noise data in;
The training sample data are input in deep learning network model and are trained, until the deep learning network mould
The loss function value of type meets preset requirement;
The loss function value of the deep learning network model is met to training pattern when preset requirement, is obtained as training
Seismic data noise attenuation model.
2. the method according to claim 1, wherein the loss function is:
Wherein, l (W, b) indicates that loss function, W are the weight in the deep learning network model;B is the deep learning net
Biasing in network model;‖·‖FFor F norm, the number of non-zero element in vector is indicated as F=0, indicates vector as F=1
The sum of middle each element absolute value indicates the value of the quadratic sum extraction of square root of vector each element as F=2;N is the training sample
The total number of data, n are positive integer;I is a variable, and value range is the integer from 1 to n;F(xi) it is i-th of training
The noise data of sample data input deep learning network model reality output;It is not noisy in i-th of training sample data
The secondary power of the seismic data of sound;xiFor noise-containing seismic data in i-th of training sample data;As train
Noise data in sample data.
3. the method according to claim 1, wherein have N layer network in the deep learning network model,
In, it is provided with one article of channel in a layer network, N-a+1 layer network can be jumped to from a layer network by the channel information,
Wherein, N is positive integer, and a is the positive integer no more than N.
4. a kind of seismic data noise attenuation method, which is characterized in that including:
Obtain seismic data to be denoised;
By in the seismic data input denoising model that training obtains in advance to be denoised, the earthquake number to be denoised is obtained
Noise data in;
The seismic data to be denoised and the noise data are done into difference operation, the seismic data after being denoised.
5. according to the method described in claim 4, it is characterized in that, the denoising model is to be obtained based on deep learning network training
The model arrived.
6. according to the method described in claim 5, it is characterized in that, trained in the following way obtain the denoising model:
Obtain training sample data, wherein the training sample data include:Original earthquake data and the original earthquake number
Noise data in;
The training sample data are input in deep learning network model and are trained, the denoising model is obtained.
7. a kind of seismic data noise attenuation device, which is characterized in that including:
Module is obtained, seismic data to be denoised is obtained;
Input module obtains in the seismic data input denoising model that training obtains in advance to be denoised described wait go
The noise data in seismic data made an uproar;
The seismic data to be denoised and the noise data are done difference operation, the earthquake after being denoised by processing module
Data.
8. device according to claim 7, which is characterized in that the denoising model is to be obtained based on deep learning network training
The model arrived.
9. device according to claim 8, which is characterized in that training obtains the denoising model in the following way:
Obtain training sample data, wherein the training sample data include:Original earthquake data and the original earthquake number
Noise data in;
The training sample data are input in deep learning network model and are trained, the denoising model is obtained.
10. a kind of deep learning model training apparatus, which is characterized in that including:
Module is obtained, obtains training sample data, wherein the training sample data include:Original earthquake data and the original
Noise data in beginning seismic data;
The training sample data are input in deep learning network model and are trained by training module, until the depth
The loss function value of learning network model reaches preset requirement;
The loss function value of the deep learning network model is reached training pattern when preset requirement by processing module, as
The seismic data noise attenuation model that training obtains.
11. device according to claim 10, which is characterized in that the loss function is:
Wherein, l (W, b) indicates that loss function, W are the weight in the deep learning network model;B is the deep learning net
Biasing in network model;‖·‖FFor F norm, the number of non-zero element in vector is indicated as F=0, indicates vector as F=1
The sum of middle each element absolute value indicates the value of the quadratic sum extraction of square root of vector each element as F=2;N is the training sample
The total number of data, n are positive integer;I is a variable, and value range is the integer from 1 to n;F(xi) it is i-th of training
The noise data of sample data input deep learning network model reality output;It is not noisy in i-th of training sample data
The secondary power of the seismic data of sound;xiFor noise-containing seismic data in i-th of training sample data;As train
Noise data in sample data.
12. device according to claim 10, which is characterized in that the training sample data are input to deep learning net
Training in network model, including:
N layer network is shared in deep learning network model;
A layer network is provided with one article of channel N-2a layer network that directly jumps and is connected to N- in the deep learning network model
A+1 layer network, so that information is easier to transmit;
Wherein, N is positive integer, and a is the positive integer no more than N.
13. a kind of seismic data noise attenuation equipment, including processor and for the memory of storage processor executable instruction, institute
State the step of realizing any one of claim 4 to 6 the method when processor executes described instruction.
14. a kind of computer readable storage medium is stored thereon with computer instruction, described instruction, which is performed, realizes that right is wanted
The step of seeking any one of 4 to 6 the method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810985408.7A CN108897045A (en) | 2018-08-28 | 2018-08-28 | Deep learning model training method and seismic data noise attenuation method, device and equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810985408.7A CN108897045A (en) | 2018-08-28 | 2018-08-28 | Deep learning model training method and seismic data noise attenuation method, device and equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108897045A true CN108897045A (en) | 2018-11-27 |
Family
ID=64358570
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810985408.7A Pending CN108897045A (en) | 2018-08-28 | 2018-08-28 | Deep learning model training method and seismic data noise attenuation method, device and equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108897045A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110007347A (en) * | 2019-04-09 | 2019-07-12 | 西南石油大学 | A kind of deep learning seismic data denoising method |
CN110008633A (en) * | 2019-04-19 | 2019-07-12 | 中国石油大学(华东) | Road noise drawing method and system based on artificial intelligence deep neural network |
CN111046737A (en) * | 2019-11-14 | 2020-04-21 | 吉林大学 | Efficient intelligent sensing acquisition method for microseism signal detection |
CN111368710A (en) * | 2020-02-27 | 2020-07-03 | 东北石油大学 | Seismic data random noise suppression method combined with deep learning |
CN112464725A (en) * | 2020-10-30 | 2021-03-09 | 中国石油天然气集团有限公司 | First arrival picking method and device based on deep learning network |
CN112578475A (en) * | 2020-11-23 | 2021-03-30 | 中海石油(中国)有限公司 | Compact reservoir dual-dessert identification method based on data mining |
CN113534251A (en) * | 2020-04-22 | 2021-10-22 | 中国石油天然气股份有限公司 | Seismic data fracture identification method and device based on convolutional neural network |
CN113900147A (en) * | 2020-07-06 | 2022-01-07 | 中国石油天然气股份有限公司 | Sand dune ringing suppression method and system |
CN113900146A (en) * | 2020-07-06 | 2022-01-07 | 中国石油天然气股份有限公司 | Surface wave pressing method and system |
CN114114421A (en) * | 2021-11-05 | 2022-03-01 | 中国石油大学(华东) | Deep learning-based guided self-learning seismic data denoising method and device |
GB2611874A (en) * | 2021-10-06 | 2023-04-19 | Cgg Services Sas | Method for seismic interference noise attenuation using DNN |
WO2023092795A1 (en) * | 2021-11-25 | 2023-06-01 | 上海帜讯信息技术股份有限公司 | Noise data recognition method and apparatus, and terminal and storage medium |
WO2023231954A1 (en) * | 2022-06-02 | 2023-12-07 | 华为技术有限公司 | Data denoising method and related device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106204467A (en) * | 2016-06-27 | 2016-12-07 | 深圳市未来媒体技术研究院 | A kind of image de-noising method based on cascade residual error neutral net |
CN106991646A (en) * | 2017-03-28 | 2017-07-28 | 福建帝视信息科技有限公司 | A kind of image super-resolution method based on intensive connection network |
WO2018026995A1 (en) * | 2016-08-03 | 2018-02-08 | Schlumberger Technology Corporation | Multi-scale deep network for fault detection |
WO2018148492A1 (en) * | 2017-02-09 | 2018-08-16 | Schlumberger Technology Corporation | Geophysical deep learning |
CN108416755A (en) * | 2018-03-20 | 2018-08-17 | 南昌航空大学 | A kind of image de-noising method and system based on deep learning |
-
2018
- 2018-08-28 CN CN201810985408.7A patent/CN108897045A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106204467A (en) * | 2016-06-27 | 2016-12-07 | 深圳市未来媒体技术研究院 | A kind of image de-noising method based on cascade residual error neutral net |
WO2018026995A1 (en) * | 2016-08-03 | 2018-02-08 | Schlumberger Technology Corporation | Multi-scale deep network for fault detection |
WO2018148492A1 (en) * | 2017-02-09 | 2018-08-16 | Schlumberger Technology Corporation | Geophysical deep learning |
CN106991646A (en) * | 2017-03-28 | 2017-07-28 | 福建帝视信息科技有限公司 | A kind of image super-resolution method based on intensive connection network |
CN108416755A (en) * | 2018-03-20 | 2018-08-17 | 南昌航空大学 | A kind of image de-noising method and system based on deep learning |
Non-Patent Citations (4)
Title |
---|
HAISHAN LI,ET AL.: "Deep learning for ground-roll noise attenuation", 《SEG INTERNATIONAL EXPOSITION AND 88TH ANNUAL MEETING》 * |
于四伟,等: "基于深度学习的地震噪声压制 ", 《CPS/SEG北京2018国际地球物理会议暨展览电子论文集》 * |
常德宽,等: "基于深度学习 的地震断裂检测方法 ", 《CPS/SEG北京2018国际地球物理会议暨展览电子论文集》 * |
张芬,等: "基于深度残差网络的地震资料随机噪声衰减方法", 《CPS/SEG北京2018国际地球物理会议暨展览电子论文集》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110007347B (en) * | 2019-04-09 | 2020-06-30 | 西南石油大学 | Deep learning seismic data denoising method |
CN110007347A (en) * | 2019-04-09 | 2019-07-12 | 西南石油大学 | A kind of deep learning seismic data denoising method |
CN110008633A (en) * | 2019-04-19 | 2019-07-12 | 中国石油大学(华东) | Road noise drawing method and system based on artificial intelligence deep neural network |
CN110008633B (en) * | 2019-04-19 | 2023-04-14 | 中国石油大学(华东) | Highway noise suppression method and system based on artificial intelligence deep neural network |
CN111046737B (en) * | 2019-11-14 | 2022-07-08 | 吉林大学 | Efficient intelligent sensing acquisition method for microseism signal detection |
CN111046737A (en) * | 2019-11-14 | 2020-04-21 | 吉林大学 | Efficient intelligent sensing acquisition method for microseism signal detection |
CN111368710A (en) * | 2020-02-27 | 2020-07-03 | 东北石油大学 | Seismic data random noise suppression method combined with deep learning |
CN111368710B (en) * | 2020-02-27 | 2022-03-04 | 东北石油大学 | Seismic data random noise suppression method combined with deep learning |
CN113534251A (en) * | 2020-04-22 | 2021-10-22 | 中国石油天然气股份有限公司 | Seismic data fracture identification method and device based on convolutional neural network |
CN113900147A (en) * | 2020-07-06 | 2022-01-07 | 中国石油天然气股份有限公司 | Sand dune ringing suppression method and system |
CN113900146A (en) * | 2020-07-06 | 2022-01-07 | 中国石油天然气股份有限公司 | Surface wave pressing method and system |
CN113900147B (en) * | 2020-07-06 | 2024-05-28 | 中国石油天然气股份有限公司 | Sand dune ringing suppression method and system |
CN112464725A (en) * | 2020-10-30 | 2021-03-09 | 中国石油天然气集团有限公司 | First arrival picking method and device based on deep learning network |
CN112578475A (en) * | 2020-11-23 | 2021-03-30 | 中海石油(中国)有限公司 | Compact reservoir dual-dessert identification method based on data mining |
GB2611874A (en) * | 2021-10-06 | 2023-04-19 | Cgg Services Sas | Method for seismic interference noise attenuation using DNN |
CN114114421A (en) * | 2021-11-05 | 2022-03-01 | 中国石油大学(华东) | Deep learning-based guided self-learning seismic data denoising method and device |
CN114114421B (en) * | 2021-11-05 | 2023-09-29 | 中国石油大学(华东) | Deep learning-based guided self-learning seismic data denoising method and device |
WO2023092795A1 (en) * | 2021-11-25 | 2023-06-01 | 上海帜讯信息技术股份有限公司 | Noise data recognition method and apparatus, and terminal and storage medium |
WO2023231954A1 (en) * | 2022-06-02 | 2023-12-07 | 华为技术有限公司 | Data denoising method and related device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108897045A (en) | Deep learning model training method and seismic data noise attenuation method, device and equipment | |
Härdle | Smoothing techniques: with implementation in S | |
KR20170091963A (en) | Gesture classification apparatus and method using electromyogram signals | |
US10121386B2 (en) | Learning model for dynamic component utilization in a question answering system | |
Zhang et al. | Learning of structured graph dictionaries | |
CN106030475B (en) | Optimization for the touch processing of Intrusion Detection based on host | |
CN107943946B (en) | Relevance method for digging between test item bank knowledge point based on Apriori algorithm | |
CN107067031B (en) | Calligraphy posture automatic identification method based on Wi-Fi signal | |
CN109446328A (en) | A kind of text recognition method, device and its storage medium | |
Gunawan et al. | A BPSO-based method for high-utility itemset mining without minimum utility threshold | |
CN115795282B (en) | Shock tube dynamic pressure reconstruction method and device, electronic equipment and storage medium | |
Sharbati et al. | Detection and extraction of velocity pulses of near-fault ground motions using asymmetric Gaussian chirplet model | |
Gu et al. | A study of hierarchical correlation clustering for scientific volume data | |
Eastman et al. | Simpsonian'evolution by jumps' in an adaptive radiation of Anolis lizards | |
ur Rehman et al. | Translation invariant multi-scale signal denoising based on goodness-of-fit tests | |
CN110059131A (en) | A kind of method for visualizing and device that time series data relationship develops | |
CN109100803A (en) | The determination method and apparatus of micro-fracture | |
Gan | Signal processing and image processing for acoustical imaging | |
Li et al. | A new denoising method based on decomposition mixing of hydro-acoustic signal | |
CN113836804A (en) | Animal identification model establishing method based on convolutional neural network and application system thereof | |
Ludkovski | mlOSP: Towards a unified implementation of regression Monte Carlo algorithms | |
Pillai et al. | Strengthening Cybersecurity using a Hybrid Classification Model with SCO Optimization for Enhanced Network Intrusion Detection System | |
CN111382837B (en) | Antagonistic sample generation method based on depth product quantization | |
Kuzume et al. | FPGA-based lifting wavelet processor for real-time signal detection | |
Meselhi et al. | Parallel evolutionary algorithm for eeg optimization problems |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181127 |
|
RJ01 | Rejection of invention patent application after publication |