CN114461468A - Microprocessor application scene recognition method based on artificial neural network - Google Patents
Microprocessor application scene recognition method based on artificial neural network Download PDFInfo
- Publication number
- CN114461468A CN114461468A CN202210072862.XA CN202210072862A CN114461468A CN 114461468 A CN114461468 A CN 114461468A CN 202210072862 A CN202210072862 A CN 202210072862A CN 114461468 A CN114461468 A CN 114461468A
- Authority
- CN
- China
- Prior art keywords
- data
- time
- microprocessor
- network
- application scene
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 21
- 230000008569 process Effects 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims description 6
- 241000238876 Acari Species 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims 2
- 230000017525 heat dissipation Effects 0.000 abstract description 3
- 238000013461 design Methods 0.000 abstract description 2
- 238000005457 optimization Methods 0.000 abstract 1
- 210000002569 neuron Anatomy 0.000 description 6
- 230000000306 recurrent effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/22—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
- G06F11/2205—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing using arrangements specific to the hardware being tested
- G06F11/2236—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing using arrangements specific to the hardware being tested to test CPU or processors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/22—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
- G06F11/2263—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/22—Detection or location of defective computer hardware by testing during standby operation or during idle time, e.g. start-up testing
- G06F11/2273—Test methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the field of electronic design automation, and relates to a microprocessor application scene recognition method based on an artificial neural network. In the running process of the microprocessor, the Intel-pcm is used for extracting data, the voltage and frequency and the heat dissipation condition of the processor are firstly fixed in bios, and the setting of the system is determined not to influence the scene characteristics. And analyzing the relation between the output data, reducing the dimension in a space range, and keeping the key information in the space range. In the time dimension, the data is processed into time sequence data segments with fixed length by adopting a sliding window method, and the moving step length of each window is one or more statistical periods of the data. And marking type labels on the intercepted and processed data segments according to different application types, sending the data segments into an LSTM network for training, reasonably modifying the applied labels according to a training result, and repeatedly training to obtain a trained network. In subsequent use, the data is processed and then directly sent to a network, and the current application type can be identified in real time. The invention can run in real time and identify the running application scene of the microprocessor, thereby laying a cushion for the subsequent targeted optimization work.
Description
Technical Field
The invention belongs to the field of electronic design automation, and relates to a microprocessor application scene recognition method based on an artificial neural network.
Background
Time-series has been a hot problem in the field of data mining. The time series refers to sequence data recorded with attribute values according to time sequence, and the time series analysis is to analyze the data change process and future trend according to historical record values. With the advent of the big data era and the development of computer hardware technology, a great amount of multivariate time sequence data is generated in the fields of aviation, finance, medical treatment, industrial production and the like. The implicit information in the original time sequence is further explored, and rules and modes are found beneficially. The time series classification not only needs to consider the numerical value relationship among different attributes, but also needs to consider the precedence relationship of time sequence points in the sequence, so that the time series classification has greater challenge than the traditional classification problem.
In recent years, researchers at home and abroad try to introduce deep learning into a time series classification task, some researchers represent a hidden layer of a Recurrent Neural Network (RNN) in the deep learning as a feature, and then classify the state of the hidden layer by using a classifier, so that the reliability of a classification result is improved, but the Recurrent Neural Network is not stable enough due to the problems of gradient explosion and gradient disappearance, and the calculation complexity is increased along with the increase of input data. In order to solve the problems of gradient explosion and gradient disappearance of the cyclic neural Network, researchers propose Long Short Term neural networks (LSTM). In addition, researchers are gradually turning to feature extraction from the spatial index dimension, feature-based classification methods instead of the original time series, and partial subsequences instead.
The microprocessor can face different application scenes in the using process, and if the current application scene of the microprocessor is identified in the using process, the current application scene can be optimized in a targeted mode, so that the efficient use of the microprocessor is facilitated. In the past, someone judges the currently used application by analyzing a network data packet when the application is started, but the method can only be used during networking; some people also construct feature data sets for some applications by marking behavior features of users (such as features of touching a mobile phone screen) and then recognize the feature data sets through the data sets, but the process of constructing the data sets is subjective and complicated, and data acquisition for all applications is not practical.
In summary, in view of the above problems, it is very important for efficient use of a microprocessor to provide an identification method that can be used for identifying different application scenarios that the microprocessor may face during use.
Disclosure of Invention
In order to solve the problems in the background art, a microprocessor application scene recognition method based on an artificial neural network is provided. The invention aims at solving the problem of identifying the application scene of a microprocessor in the operation process, firstly, a data set is constructed, the acquired system data needs to be input in real time in the identification process, and according to the previous research, the invention discovers that the comprehensive data in the operation process of the microprocessor can be acquired by using an Intel-pcm (Intel Performance Counter monitor); then, Principal Component Analysis (PCA) is carried out according to the data, redundant data dimensionality is deleted, the data set is compressed from the perspective of space, and the complexity of calculation of a neural network algorithm is effectively reduced; and finally, inputting the processed data into a trained LSTM network, and identifying the application scene through a forgetting mechanism and information dependence of an LSTM model.
The technical scheme of the invention is as follows:
firstly, fixing the voltage and frequency of a processor in a bios (basic Input Output system), determining that a power strategy of a system cannot influence the use scene characteristics of the user, secondly, fixing the heat dissipation condition of the microprocessor, ensuring that the rotating speed of a fan of the system is fixed by the system so as to keep the external heat dissipation condition stable and unchanged, and reducing the statistical influence on performance data of the microprocessor.
And step two, because the data extracted from the Intel-pcm contains too many items, great computation time complexity is needed when the LSTM network is used for identification as a whole, the relation between the data is analyzed and extracted, and the key information in the data is reserved for subsequent identification steps.
And thirdly, counting data of a system in a period of time when the system processes different applications, wherein the system data extracted from the Intel-pcm is a long data sequence, the system data needs to be processed into a time sequence data segment with a fixed length when an LSTM network is used for training, a sliding window method is adopted for intercepting the data, the moving step length of each window is a counting period of the data, and the specific length of the window is determined by the user.
And fourthly, before training, marking application labels on the intercepted and processed data segments according to different applications, sending the marked data into an LSTM network for training, wherein the condition that the identification of a verification set is inaccurate can occur, the main reason is that some applications are similar to the calling of system resources, the applications cannot be divided into two types of applications simply because of different applications, and the applications are classified into a single type by modifying the applied labels.
And step five, processing the data which is not tested and verified in the step three and the step four, sending the data into the trained neural network, comparing the obtained output with the actual label, and verifying the accuracy of the method.
Drawings
The present invention will be described in detail below with reference to the accompanying drawings and embodiments.
FIG. 1 is a flow chart of a microprocessor application scenario identification method based on an artificial neural network;
FIG. 2 is a flow chart of collecting and processing data;
fig. 3 is a diagram of a repeating unit in an LSTM network.
Fig. 4 shows the structure of an LSTM network to which the present invention is applied.
Detailed Description
In order to better explain the technical implementation flow of the invention, the technical parts will be more clearly explained below by combining with the drawings in the invention. The examples described in this disclosure are only a few, and not all examples are applicable. All other examples, which can be obtained by a person skilled in the art without any inventive step based on the examples of the present invention, are within the scope of the present invention.
FIG. 1 is a flow chart of a microprocessor application scenario recognition method based on an artificial neural network.
In the invention, according to the characteristics of multivariate time sequence data, a microprocessor application scene recognition method based on an artificial neural network is provided, the whole frame comprises 6 levels of data preprocessing, an input layer, a hidden layer, an output layer, network training and network prediction classification, the preprocessing process is responsible for processing an original data set into a data format which can be used by a lower layer and reducing the dimension of redundant data, then the obtained standard format input data is transmitted to the input layer, and the input layer transmits the received data set according to the ratio of 7: 3, the hidden layer builds a recurrent neural network by using the LSTM cells shown in the figure 4, the parameters of the training network are stored, and the output layer obtains a prediction classification result. The training part of the network is completed, an artificial neural network which can be used for recognition is obtained, and the whole recognition process can be completed by using the network and the data processed in real time.
FIG. 2 is a flow chart for collecting and processing data.
The goal of this process is to reduce the dimensions of the data and process it into a data format for subsequent use, the experiment is performed on a computer with an Intel9750h, and after the processor frequency and fan speed are fixed in the BIOS, different applications are run to collect the original experimental data. The collected original data count to 214 items (including items such as System Core C-States, System Pack C-States, Core0(Socket0) to Core11(Socket 0)), redundant data of spatial dimensions are processed, and then 23 items of data such as EXEC, IPC FREQ, AFREQ, L3MISS, L2MISS, L3HIT, L3MPI, L2MPI, READ, WRITE, INST, ACYC, TIME (ticks), phyipc, INSTnom, C0res, C1res, C3res, C7res, and Proc Energy (Joules) are retained for identification. The data is fetched by adopting a sliding window frame in the time dimension, and a section of data of every 50 time steps is used as single type identification data to be input, namely the window size is 50, and each time step is determined to be 10ms, namely the data of 500ms is input in a single time.
Fig. 3 is a diagram of a repeating unit in an LSTM network.
The first step of the LSTM unit is to decide what information should be forgotten by the neuron. This is made up of a Sigmod layer called the "forget gate layer". A "1" in the Sigmod layer means "this is completely reserved", and a "0" means "this is completely forgotten". It inputs Ht-1And Xt, then at Ct-1Each neuron state of (a) outputs a number between 0 and 1. The next step is to decide what information we want to keep in the neuronal cells, which consists of two parts. First, a Sigmod layer, called the "input gate layer," determines the values we want to update. Then, a tanh layer generates a new candidate value gt. It will be added to the neuronal state. In the next step we combine these two steps to generate an updated state value. Finally, we decide what to output. This output is based on our neuron state, but with a filter. First, we use the Sigmod layer to decide which part of the neuron states need to be output; then we let the neuron state go through the tanh (let the output value become between-1 ~ 1) layer and multiply the output of the Sigmod threshold, just output what we want to output.
Fig. 4 shows the structure of an LSTM network.
The invention uses a double-layer LSTM network, wherein the first layer is an input layer, the second layer is an LSTM layer, the third layer is a dropout layer, the fourth layer is an LSTM layer, the fifth layer is a full connection layer, the sixth layer is an activation function layer, the seventh layer is an output layer, and an application identification result value is output to update network parameters after data passes through the network.
The invention relates to a microprocessor application scene recognition method based on an artificial neural network, and examples of the method for recognizing the application scene of the microprocessor based on the artificial neural network, wherein the method for recognizing the application scene of the microprocessor based on the artificial neural network and the examples of the method for recognizing the application scene of the microprocessor based on the artificial neural network are used for explaining each processing flow of the microprocessor in detail, but the method is not limited to the method, and the technical scheme can be further optimized afterwards, so that the essence of the corresponding technical scheme does not depart from the spirit and the scope of each example technical scheme of the invention.
Claims (4)
1. A microprocessor application scene recognition method based on an artificial neural network is characterized in that: real-time identification is carried out based on performance counter data in the running process of the microprocessor; simultaneously, the dimension of the data is reduced in a time dimension and a space dimension, the network size is reduced, and the identification speed is increased; and constructing an LSTM network for application scene recognition, and performing real-time recognition after training.
2. The microprocessor-based real-time identification of performance counter data during operation of the microprocessor as recited in claim 1, wherein: the real-TIME identification based on the performance counter data in the microprocessor operation process means that a large amount of performance counter data is generated in the microprocessor operation process, identification is performed differently from a software layer (for example, application white list construction and network data packet analysis are performed), data is extracted from the performance counter layer, and the data is specifically but not limited to 23 items of data, namely EXEC, IPC, FREQ, AFREQ, L3MISS, L2MISS, L3HIT, L3MPI, L2MPI, READ, WRITE, INST, ACYC, TIME (ticks), phyipc, INSTnom, C0res, C1res, C3res, C7res and Proc Energy, and application scene identification is performed by using the performance counter data.
3. The simultaneous dimensionality reduction of data in both the time and space dimensions according to claim 1, characterized by: the simultaneous dimensionality reduction of the data in the time dimension and the space dimension means that the extracted performance counter data contains too many items, and the overall identification by using the LSTM network needs to calculate time and space complexity greatly. And analyzing and extracting the relation between the data in the spatial dimension, and keeping the key information in the relation. In the time dimension, data of a period of time system in processing different applications are counted and processed into a time sequence data segment with a fixed length, a sliding window method is adopted for intercepting the data, the moving step length of each window is one or more counting periods of the data, and the size of the window can be randomly selected in a reasonable range.
4. The method of claim 1 for constructing an LSTM network for identification classification, wherein: the construction of the LSTM network for application scene recognition, and the real-time recognition after the training is finished means that the artificially constructed LSTM network is used for training the performance counter data of the divided data sets, the network parameters are stored after the training process is finished, and then the network can be used for real-time application scene recognition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210072862.XA CN114461468A (en) | 2022-01-21 | 2022-01-21 | Microprocessor application scene recognition method based on artificial neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210072862.XA CN114461468A (en) | 2022-01-21 | 2022-01-21 | Microprocessor application scene recognition method based on artificial neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114461468A true CN114461468A (en) | 2022-05-10 |
Family
ID=81409404
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210072862.XA Pending CN114461468A (en) | 2022-01-21 | 2022-01-21 | Microprocessor application scene recognition method based on artificial neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114461468A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190005237A1 (en) * | 2017-06-30 | 2019-01-03 | Paul J. Long | Method and apparatus for identifying, predicting, preventing network malicious attacks |
| CN110147711A (en) * | 2019-02-27 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Video scene recognition methods, device, storage medium and electronic device |
| US20190273509A1 (en) * | 2018-03-01 | 2019-09-05 | Crowdstrike, Inc. | Classification of source data by neural network processing |
| WO2021184468A1 (en) * | 2020-03-18 | 2021-09-23 | 中国科学院深圳先进技术研究院 | Action recognition method and apparatus, device, and medium |
| CN113904961A (en) * | 2021-09-09 | 2022-01-07 | 中山大学 | User behavior identification method, system, equipment and storage medium |
| WO2022111506A1 (en) * | 2020-11-26 | 2022-06-02 | 北京灵汐科技有限公司 | Video action recognition method and apparatus, electronic device and storage medium |
-
2022
- 2022-01-21 CN CN202210072862.XA patent/CN114461468A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190005237A1 (en) * | 2017-06-30 | 2019-01-03 | Paul J. Long | Method and apparatus for identifying, predicting, preventing network malicious attacks |
| US20190273509A1 (en) * | 2018-03-01 | 2019-09-05 | Crowdstrike, Inc. | Classification of source data by neural network processing |
| CN110147711A (en) * | 2019-02-27 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Video scene recognition methods, device, storage medium and electronic device |
| WO2021184468A1 (en) * | 2020-03-18 | 2021-09-23 | 中国科学院深圳先进技术研究院 | Action recognition method and apparatus, device, and medium |
| WO2022111506A1 (en) * | 2020-11-26 | 2022-06-02 | 北京灵汐科技有限公司 | Video action recognition method and apparatus, electronic device and storage medium |
| CN113904961A (en) * | 2021-09-09 | 2022-01-07 | 中山大学 | User behavior identification method, system, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 王珂林等: "基于随机森林的抗混淆 Ａｎｄｒｏｉｄ 恶意应用检测" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | Temporal dynamic appearance modeling for online multi-person tracking | |
| CN109447099B (en) | PCA (principal component analysis) dimension reduction-based multi-classifier fusion method | |
| CN112118141A (en) | Communication network-oriented alarm event correlation compression method and device | |
| CN109871780A (en) | A kind of face quality decision method, system and face identification method, system | |
| CN106940805A (en) | A kind of group behavior analysis method based on mobile phone sensor | |
| CN109784015A (en) | A kind of authentication identifying method and device | |
| CN113420294A (en) | Malicious code detection method based on multi-scale convolutional neural network | |
| CN109886334A (en) | A kind of shared nearest neighbor density peak clustering method of secret protection | |
| CN112491891B (en) | Network attack detection method based on hybrid deep learning in Internet of things environment | |
| CN109271912A (en) | Video classification methods, device, electronic equipment and storage medium | |
| CN115964258A (en) | Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis | |
| KR102005952B1 (en) | Apparatus and Method for refining data of removing noise data in Machine learning modeling | |
| Bashar et al. | Algan: Time series anomaly detection with adjusted-lstm gan | |
| CN111898694B (en) | Non-invasive load identification method and device based on random tree classification | |
| CN115954019A (en) | Environmental noise identification method and system integrating self-attention and convolution operation | |
| CN114461468A (en) | Microprocessor application scene recognition method based on artificial neural network | |
| Fumera et al. | Multi‐stage ranking approach for fast person re‐identification | |
| CN109614854B (en) | Video data processing method and device, computer device and readable storage medium | |
| Li et al. | Recognition algorithm of athletes' partially occluded face based on a deep learning algorithm | |
| CN112073582A (en) | Smart phone use situation identification method based on touch behavior sequence | |
| Lee et al. | Transformer Based Early Classification for Real-Time Human Activity Recognition in Smart Homes | |
| CN118332360B (en) | Cross-block chain transaction address tracking method, system, equipment and medium | |
| CN117692261B (en) | Malicious Bot recognition method based on behavior subgraph characterization | |
| CN113420628B (en) | Group behavior identification method and device, computer equipment and storage medium | |
| CN115240647B (en) | Sound event detection method, device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20220510 |