CN108897614A - A kind of memory method for early warning and server-side based on convolutional neural networks - Google Patents
A kind of memory method for early warning and server-side based on convolutional neural networks Download PDFInfo
- Publication number
- CN108897614A CN108897614A CN201810512007.XA CN201810512007A CN108897614A CN 108897614 A CN108897614 A CN 108897614A CN 201810512007 A CN201810512007 A CN 201810512007A CN 108897614 A CN108897614 A CN 108897614A
- Authority
- CN
- China
- Prior art keywords
- memory
- neural networks
- convolutional neural
- predicted
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
- G06F9/5016—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals the resource being the memory
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0876—Network utilisation, e.g. volume of load or congestion level
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Environmental & Geological Engineering (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of memory method for early warning and server-side based on convolutional neural networks, server-side receive the memory information that client is sent, generate pictorial information data to be predicted；Call memory disaggregated model to treat predicted pictures information data and classify, obtain the corresponding memory of pictorial information data to be predicted as a result, and memory result is back to client, memory disaggregated model be the convolutional neural networks model for having trained completion；The present invention is by can effectively extract the pictorial information formed by memory information using convolutional neural networks, risk is overflowed to effectively predict the memory information of acquisition and whether overflow or exist, superior technique is provided for client, server-side, O&M and supports；Meanwhile the present invention can be cross-platform general and has good scalability, i.e., the present invention provides it is a kind of it is cross-platform it is general, with good scalability and more authentic and valid memory overflows early warning scheme.
Description
Technical field
The present invention relates to computer vision and deep learning field, in particular to a kind of memory based on convolutional neural networks
Method for early warning and server-side.
Background technique
Client device is when running application program, it may appear that the phenomenon that using without enough memory headrooms for it, needle
, there are a variety of memory early warning schemes in the prior art in the phenomenon that overflowing to this memory, one is by monitoring programme, monitoring
Whether Installed System Memory soon reaches the upper limit；The second is the mistake of monitoring system itself is to determine whether overflowing or having there are memory
Potential memory problem risk.
The method and device of a kind of memory usage monitoring of application number 201610903370.5, by applying journey to each
The memory service condition of sequence is monitored, and reminds user when being more than warning value, it is allowed voluntarily to select how to clear up.But this side
Formula is often that server memory has had the early warning just provided when obvious problem, i.e., traditional memory early warning scheme is not
Real early warning.
Summary of the invention
The technical problem to be solved by the present invention is to：A kind of memory method for early warning and clothes based on convolutional neural networks is provided
Business end, to effectively carry out memory early warning.
In order to solve the above-mentioned technical problem, the technical solution adopted by the present invention is：
A kind of memory method for early warning based on convolutional neural networks, including step：
S1, server-side receive the memory information that client is sent, and generate pictorial information data to be predicted；
S2, it calls memory disaggregated model to classify the pictorial information data to be predicted, obtains the figure to be predicted
The corresponding memory of piece information data as a result, and the memory result is back to client, the memory disaggregated model is to have instructed
Practice the convolutional neural networks model completed.
In order to solve the above-mentioned technical problem, the another technical solution that the present invention uses for：
A kind of memory Warning Service end based on convolutional neural networks, including memory, processor and it is stored in memory
Computer program that is upper and can running on a processor, the processor realize following steps when executing the computer program：
S1, the memory information that client is sent is received, generates pictorial information data to be predicted；
S2, it calls memory disaggregated model to classify the pictorial information data to be predicted, obtains the figure to be predicted
The corresponding memory of piece information data as a result, and the memory result is back to client, the memory disaggregated model is to have instructed
Practice the convolutional neural networks model completed.
The beneficial effects of the present invention are：The present invention can be extracted effectively using convolutional neural networks by memory information shape
At pictorial information, thus the memory information for effectively predicting acquisition whether overflow or exist overflow risk, be client,
Server-side, O&M provide superior technique and support；Simultaneously as client only needs to be communicated with server-side, from
And it enables the invention to cross-platform general；If memory information is replaced with the information such as network traffic data, flow also may be implemented
The corresponding warning function of early warning energy, so that the present invention be made to have good scalability, i.e., the present invention provides a kind of cross-platform
It is general, with good scalability and more authentic and valid memory overflows early warning scheme.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of memory method for early warning based on convolutional neural networks of the embodiment of the present invention；
Fig. 2 is a kind of structural schematic diagram at memory Warning Service end based on convolutional neural networks of the embodiment of the present invention；
Fig. 3 is that server-side collects the picture A to be predicted about memory information in the embodiment of the present invention；
Fig. 4 is that server-side collects the picture B to be predicted about memory information in the embodiment of the present invention；
Fig. 5 is that server-side collects the picture C to be predicted about memory information in the embodiment of the present invention；
Fig. 6 is that server-side treats the result schematic diagram of predicted pictures A, B, C after classification in the embodiment of the present invention.
Label declaration：
1, a kind of memory Warning Service section based on convolutional neural networks；2, memory；3, processor.
Specific embodiment
To explain the technical content, the achieved purpose and the effect of the present invention in detail, below in conjunction with embodiment and cooperate attached
Figure is explained.
The design of most critical of the present invention is：It can effectively be extracted using convolutional neural networks and to be formed by memory information
Pictorial information, so that whether the memory information for effectively predicting acquisition, which overflows or exist, overflows risk.
Before this, technical solution to facilitate the understanding of the present invention, for english abbreviation involved in the present invention, equipment
Etc. being described as follows：
(1),GRPC：It is in the present invention the abbreviation of Google Remote Procedure Call, Chinese is construed to
One high-performance of Google open source, the remote procedure call frame across language.
(2), Map data structure：For saving the data with mapping relations, therefore in store two in Map data structure
Kind value, one group is used to save key, and another group is used to save value.
(3),ReLU：It is in the present invention the abbreviation of Rectified Linear Units, Chinese explains line rectification
Function, also known as amendment linear unit, are common activation primitives in a kind of artificial neural network, generally refer to ramp function and
Its mutation is the nonlinear function of representative.
Please refer to Fig. 1 and Fig. 2, a kind of memory method for early warning based on convolutional neural networks, including step：
S1, server-side receive the memory information that client is sent, and generate pictorial information data to be predicted；
S2, it calls memory disaggregated model to classify the pictorial information data to be predicted, obtains the figure to be predicted
The corresponding memory of piece information data as a result, and the memory result is back to client, the memory disaggregated model is to have instructed
Practice the convolutional neural networks model completed.
As can be seen from the above description, the beneficial effects of the present invention are：Using convolutional neural networks can effectively extract by
The pictorial information that memory information is formed, so that whether the memory information for effectively predicting acquisition, which overflows or exist, overflows wind
Danger provides superior technique for client, server-side, O&M and supports；Simultaneously as client only needs to carry out with server-side
Communication, to enable the invention to cross-platform general；It, can also if memory information is replaced with the information such as network traffic data
With realize flow early warning can corresponding warning function, to make the present invention that there is good scalability, i.e., the present invention provides
It is a kind of it is cross-platform it is general, with good scalability and more authentic and valid memory overflows early warning scheme.
Further, pictorial information data to be predicted are generated in the step S1 is specially：
Pictorial information data to be predicted are generated according to memory information continuous in setting time.
Seen from the above description, continuous memory information in setting time is provided, to generate the curve tendency of memory variation,
To obtain pictorial information data to be predicted, data are generated into picture, convenient for the identification and classification of following model.
Further, pictorial information number to be predicted is generated according to memory information continuous in setting time in the step S1
According to specially：
Preset at the first time, the second time, judge whether acquisition time reached for the second time, if it is not, then at interval of
It is primary just to acquire a memory information at the first time, and accumulating operation is carried out to acquisition time；If so, acquisition time is returned
Zero, and acquisition time coordinate diagram corresponding with memory information is generated, obtain pictorial information data to be predicted.
Seen from the above description, the time interval of acquisition is indicated at the first time, the second time indicated the time span of acquisition,
Second time removed in being exactly at the first time times of collection, wherein can carry out according to the demand of user with the second time at the first time
Third party's setting, so that the present invention has good flexibility and scalability.
Further, the training step of the memory disaggregated model in the step S2 is as follows：
S21, memory information image data collection is collected, the memory information image data collection is divided into test set and training
Collection, and according to memory, normal, memory overflows, memory risk establishes file to classify respectively in two set；
S22, building convolutional neural networks model structure, the convolutional neural networks model structure includes input module, spy
Levy extraction module, perceptron module；
S23, the initialization convolutional neural networks model, using the memory information image data collection in the training set as
Input data, the training convolutional neural networks model, and after training with the memory information image data in the test set
Collection is used as input data, tests the convolutional neural networks model for the accuracy of picture classification, if accuracy reaches expected
Value then saves the convolutional neural networks model after training.
Seen from the above description, the mode that labels that the name of a sheet by a sheet picture is not had to and classifying by file, can
Let us is more convenient, significantly more efficient carry out data prediction.
Further, the step S22 is specially：
S221, building input module, the input module includes input layer, the input layer be provided with it is described to be predicted
The one-to-one neuron of the picture pixels of pictorial information data；
S222, construction feature extraction module, the characteristic extracting module are followed successively by from top to bottom：First convolutional layer, first
Activation primitive layer, the first pond layer, the second convolutional layer, the second activation primitive layer, the second pond layer, third convolutional layer, third swash
Function layer, third pond layer living；
S223, building perceptron module, the perceptron module includes the first full articulamentum and the second full articulamentum, described
Second full articulamentum includes the neuron with memory result same number.
Seen from the above description, seen from the above description, present invention uses alternatively distributed three-deckers can be fine
Study to characteristics of image, reach better recognition accuracy.
Please refer to Fig. 1 and Fig. 2, a kind of memory Warning Service end based on convolutional neural networks, including memory, processing
Device and storage on a memory and the computer program that can run on a processor, the processor execution computer program
Shi Shixian following steps：
S1, the memory information that client is sent is received, generates pictorial information data to be predicted；
S2, it calls memory disaggregated model to classify the pictorial information data to be predicted, obtains the figure to be predicted
The corresponding memory of piece information data as a result, and the memory result is back to client, the memory disaggregated model is to have instructed
Practice the convolutional neural networks model completed.
As can be seen from the above description, the beneficial effects of the present invention are：Using convolutional neural networks can effectively extract by
The pictorial information that memory information is formed, so that whether the memory information for effectively predicting acquisition, which overflows or exist, overflows wind
Danger provides superior technique for client, server-side, O&M and supports；Simultaneously as client only needs to carry out with server-side
Communication, to enable the invention to cross-platform general；It, can also if memory information is replaced with the information such as network traffic data
With realize flow early warning can corresponding warning function, to make the present invention that there is good scalability, i.e., the present invention provides
It is a kind of it is cross-platform it is general, with good scalability and more authentic and valid memory overflows early warning scheme.
Further, pictorial information data to be predicted are generated in the step S1 is specially：
Pictorial information data to be predicted are generated according to memory information continuous in setting time.
Seen from the above description, continuous memory information in setting time is provided, to generate the curve tendency of memory variation,
To obtain pictorial information data to be predicted, data are generated into picture, convenient for the identification and classification of following model.
Further, pictorial information number to be predicted is generated according to memory information continuous in setting time in the step S1
According to specially：
Preset at the first time, the second time, judge whether acquisition time reached for the second time, if it is not, then at interval of
It is primary just to acquire a memory information at the first time, and accumulating operation is carried out to acquisition time；If so, acquisition time is returned
Zero, and acquisition time coordinate diagram corresponding with memory information is generated, obtain pictorial information data to be predicted.
Seen from the above description, the time interval of acquisition is indicated at the first time, the second time indicated the time span of acquisition,
Second time removed in being exactly at the first time times of collection, wherein can carry out according to the demand of user with the second time at the first time
Third party's setting, so that the present invention has good flexibility and scalability.
Further, the training step of the memory disaggregated model in the step S2 is as follows：
S21, memory information image data collection is collected, the memory information image data collection is divided into test set and training
Collection, and according to memory, normal, memory overflows, memory risk establishes file to classify respectively in two set；
S22, building convolutional neural networks model structure, the convolutional neural networks model structure includes input module, spy
Levy extraction module, perceptron module；
S23, the initialization convolutional neural networks model, using the memory information image data collection in the training set as
Input data, the training convolutional neural networks model, and after training with the memory information image data in the test set
Collection is used as input data, tests the convolutional neural networks model for the accuracy of picture classification, if accuracy reaches expected
Value then saves the convolutional neural networks model after training.
Seen from the above description, the mode that labels that the name of a sheet by a sheet picture is not had to and classifying by file, can
Let us is more convenient, significantly more efficient carry out data prediction.
Further, the step S22 is specially：
S221, building input module, the input module includes input layer, the input layer be provided with it is described to be predicted
The one-to-one neuron of the picture pixels of pictorial information data；
S222, construction feature extraction module, the characteristic extracting module are followed successively by from top to bottom：First convolutional layer, first
Activation primitive layer, the first pond layer, the second convolutional layer, the second activation primitive layer, the second pond layer, third convolutional layer, third swash
Function layer, third pond layer living；
S223, building perceptron module, the perceptron module includes the first full articulamentum and the second full articulamentum, described
Second full articulamentum includes the neuron with memory result same number.
Seen from the above description, seen from the above description, present invention uses alternatively distributed three-deckers can be fine
Study to characteristics of image, reach better recognition accuracy.
Fig. 1 and Fig. 2 is please referred to, the embodiment of the present invention one is：
A kind of memory method for early warning based on convolutional neural networks, including step：
S1, server-side receive the memory occupation rate information that client is sent, and occupy according to memory continuous in setting time
Rate information generates pictorial information data to be predicted；
S2, calling memory disaggregated model treat predicted pictures information data and classify, and obtain pictorial information number to be predicted
According to corresponding memory as a result, and memory result is back to client, memory disaggregated model is to have trained the convolutional Neural completed
Network model.
Wherein, generating pictorial information data to be predicted according to memory occupation rate information continuous in setting time is specially：
Preset at the first time for 1 second, the second time be 10 seconds, judge whether acquisition time reaches 10 seconds, if it is not, then
A memory occupation rate information is just acquired within every 1 second, and accumulating operation is carried out to acquisition time；If so, acquisition time is zeroed,
And acquisition time coordinate diagram corresponding with memory occupation rate information is generated, obtain pictorial information data to be predicted.
It can be seen from the above, using memory occupation rate information as the longitudinal axis, each acquisition time institute is right using acquisition time as horizontal axis
The memory occupation rate information answered is coordinate points, and adjacent coordinate points carry out line, forms memory occupation rate information change curve, tool
Body can refer to Fig. 3 to Fig. 5.1 second in the present embodiment, 10 seconds be default configuration, every 10 seconds be one acquisition time, can effectively and
Early warning is timely carried out, while the above-mentioned time carries out third party's customization according to the actual conditions of user.
From the foregoing, it will be observed that the memory information of the present embodiment acquisition is memory occupation rate information.
In addition, using GRPC communication protocol between client and server-side, while the number that client allowing service end provides
According to specification for structure, server-side data structure specification is Map data structure, and wherein key storage is acquisition time, is acquired for the first time
Index since 0, value storage be acquisition memory occupation rate information.
Meanwhile server-side defines predict interface, client calls server-side after obtaining memory occupation rate information
Predict interface by corresponding memory occupation rate information be stored in Map data structure, later, server-side call drawing API generate
Pictorial information data to be predicted.
Wherein, input of the pictorial information data to be predicted as memory disaggregated model exports memory as a result, in the present embodiment
Memory result be divided into three kinds：Memory is normal, memory overflows, memory risk.
Wherein, memory risk refers to that memory does not overflow, but there is the risk overflowed, can be with during presetting
Consideration is defined memory risk, for example the memory result in a continuous hour is all memory risk, then is judged as memory
It overflows.
Wherein, the training step of the memory disaggregated model in step S2 is as follows：
S21, memory information image data collection is collected, wherein having collected nearly 100,000 memory occupation rate data informations in total
Memory information image data collection is divided into test set and training set by picture, and wherein test set and training set ratio are 1：9, i.e.,
Nearly 10,000 memory occupation rate data information pictures are test set, and nearly 90,000 memory occupation rate data information pictures are training
Collection, and according to memory, normal, memory overflows, memory risk establishes file to classify respectively in two set, wherein
Three kinds of memory results have 3 thousand sheets or so in test set, and three kinds of memory results have 30,000 or so in training set, three files
Normally corresponding file is named as normal to memory, memory overflows corresponding file and is named as overflow, memory in folder
The corresponding file of risk is named as risk, referring in particular to Fig. 6；
S221, building input module, input module includes input layer, and input layer is provided with 150*150 neuron；
S222, construction feature extraction module, characteristic extracting module are followed successively by from top to bottom：
First convolutional layer is provided with 32 convolution kernels, wherein the size of each convolution kernel is 3*3；
First activation primitive layer, uses ReLU activation primitive；
First pond layer, size are 2*2；
Second convolutional layer is provided with 32 convolution kernels, wherein the size of each convolution kernel is 3*3；
Second activation primitive layer, uses ReLU activation primitive；
Second pond layer, size are 2*2；
Third convolutional layer is provided with 64 convolution kernels, wherein the size of each convolution kernel is 3*3；
Third activation primitive layer, uses ReLU activation primitive；
Third pond layer, size are 2*2；
S223, building perceptron module, perceptron module includes the first full articulamentum and the second full articulamentum, and first connects entirely
Connecing layer includes 128 neurons, and the second full articulamentum includes 3 neurons, total to possess 384 full connections；
S23, initialization convolutional neural networks model, using the memory information image data collection in training set as input number
According to, training convolutional neural networks model, and after training using the memory information image data collection in test set as input data,
Convolutional neural networks model is tested for the accuracy of picture classification, if accuracy reaches desired value, by the convolution after training
Neural network model saves.
The number of above-mentioned selection convolution kernel is described as follows：The feature that the usual more meanings of convolution kernel can be extracted is also more
It is more, but the feature validity features that are not intended to much more, for the present embodiment, by testing continuous repetitive exercise, verifying
Trained recognition correct rate obtains a most effective numerical value, and there is no sole criterions for this parameter, need by convolution
The continuous repetitive exercise of neural network model, parameter are adjusted and just be can determine that, so, in the present embodiment, convolutional neural networks mould
32 or 64 convolution kernels are set in pattern type, validity feature can be extracted to greatest extent.
Using the picture C to be predicted in picture B, Fig. 5 to be predicted in picture A, Fig. 4 to be predicted in Fig. 3 as memory
The input of disaggregated model has obtained the classification results of Fig. 6.
Fig. 1 and Fig. 2 is please referred to, the embodiment of the present invention two is：
A kind of memory Warning Service end 1 based on convolutional neural networks, including memory 2, processor 3 and it is stored in storage
On device 2 and the computer program that can run on processor 3, processor 3 are realized in examples detailed above one when executing computer program
The step of.
In conclusion a kind of memory method for early warning and server-side based on convolutional neural networks provided by the invention, utilizes
Convolutional neural networks can effectively extract the pictorial information formed by memory information, to effectively predict the memory of acquisition
Whether information, which is overflowed or existed, overflows risk, provides superior technique for client, server-side, O&M and supports；Simultaneously as
Client only needs to be communicated with server-side, to enable the invention to cross-platform general；If memory information is replaced
The information such as network traffic data are changed to, the corresponding warning function of flow early warning energy also may be implemented, and can be according to client's
Demand carries out third party's customization, so that the present invention has good scalability and flexibility, convolution kernel is a through reasonable settings
Number, more effective feature can be learnt into picture by adding pond layer appropriate, be thus allowed for more and more smart
Thin picture classification, to reach significantly more efficient memory early warning；Meanwhile that is, the present invention provides it is a kind of it is cross-platform it is general, have
Good scalability and more authentic and valid memory spilling early warning scheme.
The above description is only an embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalents made by bright specification and accompanying drawing content are applied directly or indirectly in relevant technical field, similarly include
In scope of patent protection of the invention.
Claims (10)
1. a kind of memory method for early warning based on convolutional neural networks, which is characterized in that including step：
S1, server-side receive the memory information that client is sent, and generate pictorial information data to be predicted；
S2, it calls memory disaggregated model to classify the pictorial information data to be predicted, obtains the picture letter to be predicted
The corresponding memory of breath data as a result, and the memory result is back to client, the memory disaggregated model is to have trained
At convolutional neural networks model.
2. a kind of memory method for early warning based on convolutional neural networks according to claim 1, which is characterized in that the step
Generating pictorial information data to be predicted in rapid S1 is specially：
Pictorial information data to be predicted are generated according to memory information continuous in setting time.
3. a kind of memory method for early warning based on convolutional neural networks according to claim 2, which is characterized in that the step
Generating pictorial information data to be predicted according to memory information continuous in setting time in rapid S1 is specially：
First time, the second time are preset, judges whether acquisition time reached for the second time, if it is not, then at interval of primary
A memory information is just acquired at the first time, and accumulating operation is carried out to acquisition time；If so, acquisition time is zeroed, and
Acquisition time coordinate diagram corresponding with memory information is generated, pictorial information data to be predicted are obtained.
4. a kind of memory method for early warning based on convolutional neural networks according to claim 1, which is characterized in that the step
The training step of memory disaggregated model in rapid S2 is as follows：
S21, memory information image data collection is collected, the memory information image data collection is divided into test set and training set,
And according to memory, normal, memory overflows, memory risk establishes file to classify respectively in two set；
S22, building convolutional neural networks model structure, the convolutional neural networks model structure include that input module, feature mention
Modulus block, perceptron module；
S23, the initialization convolutional neural networks model, using the memory information image data collection in the training set as input
Data, the training convolutional neural networks model, and made after training with the memory information image data collection in the test set
For input data, the convolutional neural networks model is tested for the accuracy of picture classification, if accuracy reaches desired value,
The convolutional neural networks model after training is saved.
5. a kind of memory method for early warning based on convolutional neural networks according to claim 4, which is characterized in that the step
Suddenly S22 is specially：
S221, building input module, the input module includes input layer, and the input layer is provided with and the picture to be predicted
The one-to-one neuron of the picture pixels of information data；
S222, construction feature extraction module, the characteristic extracting module are followed successively by from top to bottom：First convolutional layer, the first activation
Function layer, the first pond layer, the second convolutional layer, the second activation primitive layer, the second pond layer, third convolutional layer, third activate letter
Several layers, third pond layer；
S223, building perceptron module, the perceptron module include the first full articulamentum and the second full articulamentum, described second
Full articulamentum includes the neuron with memory result same number.
6. a kind of memory Warning Service end based on convolutional neural networks, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, which is characterized in that the processor is realized when executing the computer program
Following steps：
S1, the memory information that client is sent is received, generates pictorial information data to be predicted；
S2, it calls memory disaggregated model to classify the pictorial information data to be predicted, obtains the picture letter to be predicted
The corresponding memory of breath data as a result, and the memory result is back to client, the memory disaggregated model is to have trained
At convolutional neural networks model.
7. a kind of memory Warning Service end based on convolutional neural networks according to claim 6, which is characterized in that described
Pictorial information data to be predicted are generated in step S1 is specially：
Pictorial information data to be predicted are generated according to memory information continuous in setting time.
8. a kind of memory Warning Service end based on convolutional neural networks according to claim 7, which is characterized in that described
Generating pictorial information data to be predicted according to memory information continuous in setting time in step S1 is specially：
First time, the second time are preset, judges whether acquisition time reached for the second time, if it is not, then at interval of primary
A memory information is just acquired at the first time, and accumulating operation is carried out to acquisition time；If so, acquisition time is zeroed, and
Acquisition time coordinate diagram corresponding with memory information is generated, pictorial information data to be predicted are obtained.
9. a kind of memory Warning Service end based on convolutional neural networks according to claim 6, which is characterized in that described
The training step of memory disaggregated model in step S2 is as follows：
S21, memory information image data collection is collected, the memory information image data collection is divided into test set and training set,
And according to memory, normal, memory overflows, memory risk establishes file to classify respectively in two set；
S22, building convolutional neural networks model structure, the convolutional neural networks model structure include that input module, feature mention
Modulus block, perceptron module；
S23, the initialization convolutional neural networks model, using the memory information image data collection in the training set as input
Data, the training convolutional neural networks model, and made after training with the memory information image data collection in the test set
For input data, the convolutional neural networks model is tested for the accuracy of picture classification, if accuracy reaches desired value,
The convolutional neural networks model after training is saved.
10. a kind of memory Warning Service end based on convolutional neural networks according to claim 9, which is characterized in that institute
Stating step S22 is specially：
S221, building input module, the input module includes input layer, and the input layer is provided with and the picture to be predicted
The one-to-one neuron of the picture pixels of information data；
S222, construction feature extraction module, the characteristic extracting module are followed successively by from top to bottom：First convolutional layer, the first activation
Function layer, the first pond layer, the second convolutional layer, the second activation primitive layer, the second pond layer, third convolutional layer, third activate letter
Several layers, third pond layer；
S223, building perceptron module, the perceptron module include the first full articulamentum and the second full articulamentum, described second
Full articulamentum includes the neuron with memory result same number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810512007.XA CN108897614A (en) | 2018-05-25 | 2018-05-25 | A kind of memory method for early warning and server-side based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810512007.XA CN108897614A (en) | 2018-05-25 | 2018-05-25 | A kind of memory method for early warning and server-side based on convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108897614A true CN108897614A (en) | 2018-11-27 |
Family
ID=64343080
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810512007.XA Withdrawn CN108897614A (en) | 2018-05-25 | 2018-05-25 | A kind of memory method for early warning and server-side based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108897614A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109976908A (en) * | 2019-03-15 | 2019-07-05 | 北京工业大学 | A kind of server cluster dynamic retractility method based on RNN time series forecasting |
| CN110631592A (en) * | 2019-10-28 | 2019-12-31 | 华南理工大学 | Indoor autonomous navigation AGV (automatic guided vehicle) motion track fusion method based on LSTM (least squares metric model) |
| CN110780646A (en) * | 2019-09-21 | 2020-02-11 | 苏州浪潮智能科技有限公司 | Memory quality early warning method based on MES system |
| CN111143076A (en) * | 2019-12-31 | 2020-05-12 | 上海联影医疗科技有限公司 | Magnetic resonance reconstruction memory estimation method and device, electronic equipment and medium |
| CN111488197A (en) * | 2020-04-14 | 2020-08-04 | 浙江新再灵科技股份有限公司 | Deep learning model deployment method and system based on cloud server |
| CN111767204A (en) * | 2019-04-02 | 2020-10-13 | 杭州海康威视数字技术股份有限公司 | Overflow risk detection method, device and equipment |
| WO2024139359A1 (en) * | 2022-12-30 | 2024-07-04 | 千寻位置网络有限公司 | Program memory overflow early warning method and system, electronic device, and computer storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763305A (en) * | 2009-12-29 | 2010-06-30 | 青岛海信宽带多媒体技术有限公司 | Method for detecting memory leak of embedded system |
| CN106019359A (en) * | 2016-05-17 | 2016-10-12 | 浪潮集团有限公司 | Earthquake prediction system based on neural network |
| CN107104978A (en) * | 2017-05-24 | 2017-08-29 | 赖洪昌 | A kind of network risks method for early warning based on deep learning |
| US20180012110A1 (en) * | 2016-07-06 | 2018-01-11 | Accenture Global Solutions Limited | Machine learning image processing |
-
2018
- 2018-05-25 CN CN201810512007.XA patent/CN108897614A/en not_active Withdrawn
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763305A (en) * | 2009-12-29 | 2010-06-30 | 青岛海信宽带多媒体技术有限公司 | Method for detecting memory leak of embedded system |
| CN106019359A (en) * | 2016-05-17 | 2016-10-12 | 浪潮集团有限公司 | Earthquake prediction system based on neural network |
| US20180012110A1 (en) * | 2016-07-06 | 2018-01-11 | Accenture Global Solutions Limited | Machine learning image processing |
| CN107104978A (en) * | 2017-05-24 | 2017-08-29 | 赖洪昌 | A kind of network risks method for early warning based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 王勇等: "基于深度卷积神经网络的网络流量分类方法", 《通信学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109976908A (en) * | 2019-03-15 | 2019-07-05 | 北京工业大学 | A kind of server cluster dynamic retractility method based on RNN time series forecasting |
| CN111767204A (en) * | 2019-04-02 | 2020-10-13 | 杭州海康威视数字技术股份有限公司 | Overflow risk detection method, device and equipment |
| CN111767204B (en) * | 2019-04-02 | 2024-05-28 | 杭州海康威视数字技术股份有限公司 | Spill risk detection method, device and equipment |
| CN110780646A (en) * | 2019-09-21 | 2020-02-11 | 苏州浪潮智能科技有限公司 | Memory quality early warning method based on MES system |
| CN110780646B (en) * | 2019-09-21 | 2021-11-26 | 苏州浪潮智能科技有限公司 | Memory quality early warning method based on MES system |
| CN110631592A (en) * | 2019-10-28 | 2019-12-31 | 华南理工大学 | Indoor autonomous navigation AGV (automatic guided vehicle) motion track fusion method based on LSTM (least squares metric model) |
| CN110631592B (en) * | 2019-10-28 | 2022-03-29 | 华南理工大学 | Indoor autonomous navigation AGV (automatic guided vehicle) motion track fusion method based on LSTM (least squares metric model) |
| CN111143076A (en) * | 2019-12-31 | 2020-05-12 | 上海联影医疗科技有限公司 | Magnetic resonance reconstruction memory estimation method and device, electronic equipment and medium |
| CN111488197A (en) * | 2020-04-14 | 2020-08-04 | 浙江新再灵科技股份有限公司 | Deep learning model deployment method and system based on cloud server |
| WO2024139359A1 (en) * | 2022-12-30 | 2024-07-04 | 千寻位置网络有限公司 | Program memory overflow early warning method and system, electronic device, and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108897614A (en) | A kind of memory method for early warning and server-side based on convolutional neural networks | |
| CN110210286B (en) | Fundus image-based abnormality identification method, fundus image-based abnormality identification device, fundus image-based abnormality identification equipment and fundus image-based abnormality identification storage medium | |
| CN110163234A (en) | A kind of model training method, device and storage medium | |
| CN108229341A (en) | Sorting technique and device, electronic equipment, computer storage media, program | |
| CN105975916A (en) | Age estimation method based on multi-output convolution neural network and ordered regression | |
| CN110363084A (en) | A kind of class state detection method, device, storage medium and electronics | |
| CN109740155A (en) | A kind of customer service system artificial intelligence quality inspection rule self concludes the method and system of model | |
| CN108984369A (en) | Caton prediction method and device and mobile terminal | |
| CN111476296A (en) | Sample generation method, classification model training method, identification method and corresponding devices | |
| CN109302541A (en) | Electronic device, distribution method of attending a banquet and computer readable storage medium | |
| CN110555830A (en) | Deep neural network skin detection method based on deep Labv3+ | |
| CN116523268B (en) | Person post matching analysis method and device based on big data portrait | |
| CN109086765B (en) | Licence plate recognition method, device, medium, server and automobile data recorder | |
| CN106663210A (en) | Perception based multimedia processing | |
| CN116385430A (en) | Machine vision flaw detection method, device, medium and equipment | |
| CN115081025A (en) | Sensitive data management method and device based on digital middlebox and electronic equipment | |
| CN116311539A (en) | Sleep motion capturing method, device, equipment and storage medium based on millimeter waves | |
| CN111680575A (en) | Human epithelial cell staining and classifying device, equipment and storage medium | |
| CN111414930B (en) | Deep learning model training method and device, electronic equipment and storage medium | |
| CN114511759A (en) | Method and system for identifying categories and determining characteristics of skin state images | |
| CN108038467B (en) | A kind of sparse face identification method of mirror image in conjunction with thickness level | |
| CN109033378A (en) | A kind of application method of Zero-shot Learning in intelligent customer service system | |
| CN109003259A (en) | Fingerprint quality evaluation method based on streakline Quality Expert's visual cognition machine learning | |
| CN112489628A (en) | Voice data selection method and device, electronic equipment and storage medium | |
| CN117085251A (en) | Transcranial direct current stimulation device for determining intervention stimulation mode based on scene |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181127 |