CN109376849A - A kind of control method and device of convolutional neural networks system - Google Patents
A kind of control method and device of convolutional neural networks system Download PDFInfo
- Publication number
- CN109376849A CN109376849A CN201811125333.1A CN201811125333A CN109376849A CN 109376849 A CN109376849 A CN 109376849A CN 201811125333 A CN201811125333 A CN 201811125333A CN 109376849 A CN109376849 A CN 109376849A
- Authority
- CN
- China
- Prior art keywords
- network
- arithmetic element
- parameter
- convolutional neural
- neural networks
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a kind of control methods of convolutional neural networks system, the described method includes: 1), for target application scene convolutional neural networks system each layer of network, obtain the parameter of the network, wherein, the parameter of the network includes: the size for the operation that the network is related to, variable；2) parameter of the network, is parsed, and arithmetic element corresponding with the parameter of the network is obtained from the set of preset arithmetic element according to the parameter of the network, wherein, the set of the arithmetic element includes: arithmetic element used in the convolutional neural networks system under several different application scenes；3), using the arithmetic element as the arithmetic element of the network, and the operation of the network is carried out using the arithmetic element.The invention discloses a kind of control devices of convolutional neural networks system.Using the embodiment of the present invention, the adaptability of convolutional neural networks chip is improved.
Description
Technical field
The present invention relates to a kind of control method and device of neural network, are more particularly to a kind of convolutional neural networks system
Control method and device.
Background technique
CNN (Convolutional Neural Network, convolutional neural networks), is a kind of feedforward neural network.Mesh
Before, CNN has become one of the research hotspot of numerous scientific domains, especially in pattern classification field, since the network avoids
The complicated early period of image is pre-processed, can directly input original image, thus has obtained more being widely applied.Due to
The feature detection layer of CNN is learnt by training data, so explicit feature extraction is avoided when using CNN, and it is hidden
Learnt from training data likes；Furthermore since the neuron weight on same Feature Mapping face is identical, so network can
With collateral learning, this is also convolutional network is connected with each other a big advantage of network relative to neuron.Convolutional neural networks are with it
The shared special construction of local weight has a unique superiority in terms of speech recognition and image procossing, layout closer to
Actual biological neural network, the shared complexity for reducing network of weight, the especially image of multidimensional input vector can be straight
Connect the complexity that input this feature of network avoids data reconstruction in feature extraction and assorting process.
The essential characteristic of convolutional neural networks is the non-linear or even unusual letter for removing to approach various natures with linear model
Exponential model.Cost is that calculator is bigger than normal, and benefit is that same model, same structure are only needed by changing weight coefficient, part ginseng
Number, so that it may be applied to a variety of different scenes.These weight coefficients, network structure are that model is established by deep learning, instruction
It gets.CNN most basic arithmetic element includes: matrix convolution, matrix pool, and matrix connects entirely.
But what existing convolutional neural networks model was developed both for specific application scenarios, therefore, convolutional Neural
The arithmetic element for including in network chip can be applied only to specific application scenarios, and then lead to convolutional neural networks chip
Adaptability it is bad.
Summary of the invention
Technical problem to be solved by the present invention lies in provide the control method and dress of a kind of convolutional neural networks system
It sets, the bad technical problem of the adaptability to solve convolutional neural networks chip existing in the prior art.
The present invention is to solve above-mentioned technical problem by the following technical programs:
The embodiment of the invention provides a kind of control methods of convolutional neural networks system, which comprises
1), for each layer of network of the convolutional neural networks system of target application scene, the ginseng of the network is obtained
Number, wherein the parameter of the network includes: the size for the operation that the network is related to, variable；
2) parameter of the network, is parsed, and is obtained from the set of preset arithmetic element according to the parameter of the network
Take arithmetic element corresponding with the parameter of the network, wherein the set of the arithmetic element includes: several different application fields
Arithmetic element used in convolutional neural networks system under scape；
3), using the arithmetic element as the arithmetic element of the network, and the net is carried out using the arithmetic element
The operation of network.
Optionally, the step 1), comprising:
The network is ranked up according to the sequencing of each layer of network of convolutional neural networks system, from sequence
The most preceding network of order starts, and using the network as current network, obtains the parameter of the network；
The step 3), comprising:
Using the arithmetic element as the arithmetic element of the current network, and the net is carried out using the arithmetic element
The operation of network；Using next network of the current network as current network, and returns and execute the ginseng for obtaining the network
Several step, until each layer network of the convolutional neural networks system has all had arithmetic element.
Optionally, the parameter according to the network obtains and the network from the set of preset arithmetic element
The corresponding arithmetic element of parameter, comprising:
According to the parameter of network, the size of arithmetic element needed for obtaining the network, operation variable；
Matched arithmetic element is obtained according to the size of arithmetic element, operation variable.
The embodiment of the invention also provides a kind of control device of convolutional neural networks system, described device includes:
Module is obtained, for each layer of network of the convolutional neural networks system for target application scene, obtains institute
State the parameter of network, wherein the parameter of the network includes: the size for the operation that the network is related to, variable；
Parsing module, for parsing the parameter of the network, and according to the parameter of the network from preset arithmetic element
Set in obtain corresponding with the parameter of network arithmetic element, wherein the set of the arithmetic element includes: several
Arithmetic element used in convolutional neural networks system under different application scene；
Computing module for using the arithmetic element as the arithmetic element of the network, and uses the arithmetic element
Carry out the operation of the network.
Optionally, the acquisition module, is used for:
The network is ranked up according to the sequencing of each layer of network of convolutional neural networks system, from sequence
The most preceding network of order starts, and using the network as current network, obtains the parameter of the network；
The arithmetic element, is used for:
Using the arithmetic element as the arithmetic element of the current network, and the net is carried out using the arithmetic element
The operation of network；Using next network of the current network as current network, and returns and execute the ginseng for obtaining the network
Several step, until each layer network of the convolutional neural networks system has all had arithmetic element.
Optionally, the parsing module, is used for:
According to the parameter of network, the size of arithmetic element needed for obtaining the network, operation variable；
Matched arithmetic element is obtained according to the size of arithmetic element, operation variable.
The present invention has the advantage that compared with prior art
Using the embodiment of the present invention, a variety of arithmetic elements that plurality of application scenes is used all are integrated in the same operation list
In the set of member, when carrying out operation using convolutional neural networks, according to the parameter selection of each layer network of convolutional neural networks
Corresponding computing unit, and then carry out the operation of convolutional neural networks, compared with the existing technology in convolutional neural networks chips
Applied to specific application scene, the convolutional neural networks chip based on the embodiment of the present invention can be applied to more arithmetic element
Kind application scenarios, improve the adaptability of convolutional neural networks chip.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of the control method of convolutional neural networks system provided in an embodiment of the present invention；
Fig. 2 is a kind of schematic illustration of the control method of convolutional neural networks system provided in an embodiment of the present invention；
Fig. 3 is a kind of structural schematic diagram of the control device of convolutional neural networks system provided in an embodiment of the present invention.
Specific embodiment
It elaborates below to the embodiment of the present invention, the present embodiment carries out under the premise of the technical scheme of the present invention
Implement, the detailed implementation method and specific operation process are given, but protection scope of the present invention is not limited to following implementation
Example.
The embodiment of the invention provides a kind of control method and device of convolutional neural networks system, first below with regard to this hair
A kind of control method for convolutional neural networks system that bright embodiment provides is introduced.
Fig. 1 is a kind of flow diagram of the control method of convolutional neural networks system provided in an embodiment of the present invention；Fig. 2
For a kind of schematic illustration of the control method of convolutional neural networks system provided in an embodiment of the present invention, such as Fig. 1 and Fig. 2 institute
Show, which comprises
S101: for each layer of network of the convolutional neural networks system of target application scene, the network is obtained
Parameter, wherein the parameter of the network includes: the size for the operation that the network is related to, variable.
Specifically, convolutional neural networks systematic difference scene can there are many kinds of, e.g., image recognition, refers to recognition of face
Line identification etc..Arithmetic element used in each layer network under each application scenarios in corresponding convolutional neural networks can phase
Together can be different, for example, arithmetic element used in convolutional neural networks can be arithmetic element a, operation in image recognition
Unit b and arithmetic element c；And in recognition of face, arithmetic element used in convolutional neural networks can be arithmetic element b
And arithmetic element c, arithmetic element d；In fingerprint recognition, arithmetic element used in convolutional neural networks can be operation list
First e, arithmetic element f and arithmetic element g.
The parameter of each layer network can be the same or different in convolutional neural networks, and the parameter of network determines the network
Using any arithmetic element, therefore, arithmetic element corresponding with the parameter can be found according to the parameter of network, for example, can
Suitable arithmetic element is found with the size of the arithmetic element according to required for network, is then sought in suitable arithmetic element again
The arithmetic element with variable needed for network is looked for, is then further found further according to other parameters, or is terminated
It finds.
In practical applications, arithmetic element includes but are not limited to: matrix convolution, matrix pool, and matrix connects entirely.
S102: the parameter of the network is parsed, and according to the parameter of the network from the set of preset arithmetic element
Obtain arithmetic element corresponding with the parameter of the network, wherein the set of the arithmetic element includes: several different applications
Arithmetic element used in convolutional neural networks system under scene.
Specifically, in the set of the arithmetic element in the convolutional neural networks chip based on embodiment illustrated in fig. 1 of the present invention
Arithmetic element used in convolutional neural networks including several application scenarios such as image recognition, recognition of face, fingerprint recognitions.
Then required arithmetic element is selected from the set of preset arithmetic element according to required convolutional neural networks system.
S103: using the arithmetic element as the arithmetic element of the network, and using described in arithmetic element progress
The operation of network.
According to the arithmetic element of selection, operation is carried out to each layer network.
In practical applications, user can be configured by software, then be realized shown in Fig. 1 of the present invention in hardware view
Embodiment, it is to be understood that S101 step memory in 2 hardware of corresponding diagram and reads argument section, i.e., reads from memory
For the parameter of convolutional neural networks.Parameter analysis of electrochemical and network-control in 2 hardware of S102 step corresponding diagram, wherein network
Control refers to is obtaining fortune corresponding with the parameter of the network according to corresponding parameter from the set of preset arithmetic element
Calculate unit.Computing module, address control in 2 hardware of S103 step corresponding diagram, wherein computing module refers to according to S102 step
The arithmetic element of middle acquisition obtains the address of arithmetic element.Address control refers to, according to change of address realization pair in calculating process
The scheduling of arithmetic element.
Using embodiment illustrated in fig. 1 of the present invention, a variety of arithmetic elements that plurality of application scenes is used all are integrated in same
In the set of a arithmetic element, when carrying out operation using convolutional neural networks, according to each layer network of convolutional neural networks
The corresponding computing unit of parameter selection, and then carry out the operation of convolutional neural networks, compared with the existing technology in convolutional Neural net
For the arithmetic element of network chip applied to specific application scene, the convolutional neural networks chip based on the embodiment of the present invention can be with
Applied to plurality of application scenes, the adaptability of convolutional neural networks chip is improved.
In addition, the convolutional neural networks chip based on embodiment illustrated in fig. 1 of the present invention, due to having several convolution mind
The arithmetic element used through network can switch between multiple convolutional neural networks according to configuration, in extension convolution mind
While through network chip adaptability, the flexibility of convolutional neural networks chip application can also be improved.
Furthermore when the prior art needs to realize plurality of application scenes, multiple convolutional neural networks chips are needed to configure, are applied
Multiple convolutional neural networks are integrated into the same chip by embodiment illustrated in fig. 1 of the present invention, it is possible to reduce number of chips, drop
Low energy consumption can also reduce flow cost during chip production.
In a kind of specific embodiment of the embodiment of the present invention, the step 1), comprising:
According to the convolutional neural networks system of target application scene each layer of network sequencing to the network
It is ranked up, since the most preceding network of sequential order, using the network as current network, obtains the parameter of the network；
The step 3), comprising:
Using the arithmetic element as the arithmetic element of the current network, and the net is carried out using the arithmetic element
The operation of network；Using next network of the current network as current network, and returns and execute the ginseng for obtaining the network
Several step, until each layer network of the convolutional neural networks system has all had arithmetic element.
Using the above embodiment of the present invention, sequencing that can in sequence to each layer network in convolutional neural networks
Obtain corresponding arithmetic element.
In a kind of specific embodiment of the embodiment of the present invention, the parameter according to the network is from preset operation
Arithmetic element corresponding with the parameter of the network is obtained in the set of unit, comprising:
According to the parameter of network, the size of arithmetic element needed for obtaining the network, operation variable；
Matched arithmetic element is obtained according to the size of arithmetic element, operation variable.
Using the above embodiment of the present invention, available corresponding arithmetic element.
Corresponding with embodiment illustrated in fig. 1 of the present invention, the embodiment of the invention also provides a kind of convolutional neural networks systems
Control device.
Fig. 3 is a kind of structural schematic diagram of the control device of convolutional neural networks system provided in an embodiment of the present invention, such as
Shown in Fig. 3, described device includes:
Module 301 is obtained, for each layer of network of the convolutional neural networks system for target application scene, is obtained
The parameter of the network, wherein the parameter of the network includes: the size for the operation that the network is related to, variable；
Parsing module 302, for parsing the parameter of the network, and according to the parameter of the network from preset operation list
Arithmetic element corresponding with the parameter of the network is obtained in the set of member, wherein the set of the arithmetic element includes: several
Arithmetic element used in convolutional neural networks system under a different application scene；
Computing module 303 for using the arithmetic element as the arithmetic element of the network, and uses the operation list
Member carries out the operation of the network.
Using embodiment illustrated in fig. 3 of the present invention, a variety of arithmetic elements that plurality of application scenes is used all are integrated in same
In the set of a arithmetic element, when carrying out operation using convolutional neural networks, according to each layer network of convolutional neural networks
The corresponding computing unit of parameter selection, and then carry out the operation of convolutional neural networks, compared with the existing technology in convolutional Neural net
For the arithmetic element of network chip applied to specific application scene, the convolutional neural networks chip based on the embodiment of the present invention can be with
Applied to plurality of application scenes, the adaptability of convolutional neural networks chip is improved.
In a kind of specific embodiment of the embodiment of the present invention, the acquisition module 301 is used for:
According to the convolutional neural networks system of target application scene each layer of network sequencing to the network
It is ranked up, since the most preceding network of sequential order, using the network as current network, obtains the parameter of the network；
The computing module 303, is used for:
Using the arithmetic element as the arithmetic element of the current network, and the net is carried out using the arithmetic element
The operation of network；Using next network of the current network as current network, and returns and execute the ginseng for obtaining the network
Several step, until each layer network of the convolutional neural networks system has all had arithmetic element.
Using the above embodiment of the present invention, sequencing that can in sequence to each layer network in convolutional neural networks
Obtain corresponding arithmetic element.
In a kind of specific embodiment of the embodiment of the present invention, the parsing module 302 is used for:
According to the parameter of network, the size of arithmetic element needed for obtaining the network, operation variable；
Matched arithmetic element is obtained according to the size of arithmetic element, operation variable.
Using the above embodiment of the present invention, available corresponding arithmetic element.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (6)
1. a kind of control method of convolutional neural networks system, which is characterized in that the described method includes:
1), for each layer of network of the convolutional neural networks system of target application scene, the parameter of the network is obtained,
In, the parameter of the network includes: the size for the operation that the network is related to, variable；
2), parse the parameter of the network, and is obtained from the set of preset arithmetic element according to the parameter of the network and
The corresponding arithmetic element of the parameter of the network, wherein the set of the arithmetic element includes: under several different application scenes
Convolutional neural networks system used in arithmetic element；
3), using the arithmetic element as the arithmetic element of the network, and the network is carried out using the arithmetic element
Operation.
2. a kind of control method of convolutional neural networks system according to claim 1, which is characterized in that the step
1), comprising:
The network is carried out according to the sequencing of each layer of network of the convolutional neural networks system of target application scene
Sequence, since the most preceding network of sequential order, using the network as current network, obtains the parameter of the network；
The step 3), comprising:
Using the arithmetic element as the arithmetic element of the current network, and the network is carried out using the arithmetic element
Operation；Using next network of the current network as current network, and returns and execute the parameter for obtaining the network
Step, until each layer network of the convolutional neural networks system has all had arithmetic element.
3. a kind of control method of convolutional neural networks system according to claim 1, which is characterized in that described according to institute
The parameter for stating network obtains arithmetic element corresponding with the parameter of the network from the set of preset arithmetic element, comprising:
According to the parameter of network, the size of arithmetic element needed for obtaining the network, operation variable；
Matched arithmetic element is obtained according to the size of arithmetic element, operation variable.
4. a kind of control device of convolutional neural networks system, which is characterized in that described device includes:
Module is obtained, for each layer of network of the convolutional neural networks system for target application scene, obtains the net
The parameter of network, wherein the parameter of the network includes: the size for the operation that the network is related to, variable；
Parsing module, for parsing the parameter of the network, and according to the parameter of the network from the collection of preset arithmetic element
Arithmetic element corresponding with the parameter of the network is obtained in conjunction, wherein the set of the arithmetic element includes: several differences
Arithmetic element used in convolutional neural networks system under application scenarios；
Computing module for using the arithmetic element as the arithmetic element of the network, and is carried out using the arithmetic element
The operation of the network.
5. a kind of control device of convolutional neural networks system according to claim 4, which is characterized in that the acquisition mould
Block is used for:
The network is ranked up according to the sequencing of each layer of network of convolutional neural networks system, from sequential order
Most preceding network starts, and using the network as current network, obtains the parameter of the network；
The arithmetic element, is used for:
Using the arithmetic element as the arithmetic element of the current network, and the network is carried out using the arithmetic element
Operation；Using next network of the current network as current network, and returns and execute the parameter for obtaining the network
Step, until each layer network of the convolutional neural networks system has all had arithmetic element.
6. a kind of control device of convolutional neural networks system according to claim 4, which is characterized in that the parsing mould
Block is used for:
According to the parameter of network, the size of arithmetic element needed for obtaining the network, operation variable；
Matched arithmetic element is obtained according to the size of arithmetic element, operation variable.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811125333.1A CN109376849A (en) | 2018-09-26 | 2018-09-26 | A kind of control method and device of convolutional neural networks system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811125333.1A CN109376849A (en) | 2018-09-26 | 2018-09-26 | A kind of control method and device of convolutional neural networks system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109376849A true CN109376849A (en) | 2019-02-22 |
Family
ID=65402537
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811125333.1A Pending CN109376849A (en) | 2018-09-26 | 2018-09-26 | A kind of control method and device of convolutional neural networks system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109376849A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109948789A (en) * | 2019-03-21 | 2019-06-28 | 百度在线网络技术（北京）有限公司 | Data load method and device for convolutional neural networks |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106940815A (en) * | 2017-02-13 | 2017-07-11 | 西安交通大学 | A kind of programmable convolutional neural networks Crypto Coprocessor IP Core |
| CN107783840A (en) * | 2017-10-27 | 2018-03-09 | 福州瑞芯微电子股份有限公司 | A kind of Distributed-tier deep learning resource allocation methods and device |
| CN108416438A (en) * | 2018-05-30 | 2018-08-17 | 济南浪潮高新科技投资发展有限公司 | A kind of convolutional neural networks hardware module dispositions method |
| CN108549934A (en) * | 2018-04-25 | 2018-09-18 | 福州瑞芯微电子股份有限公司 | A kind of operation method and device based on automated cluster neural network chip group |
-
2018
- 2018-09-26 CN CN201811125333.1A patent/CN109376849A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106940815A (en) * | 2017-02-13 | 2017-07-11 | 西安交通大学 | A kind of programmable convolutional neural networks Crypto Coprocessor IP Core |
| CN107783840A (en) * | 2017-10-27 | 2018-03-09 | 福州瑞芯微电子股份有限公司 | A kind of Distributed-tier deep learning resource allocation methods and device |
| CN108549934A (en) * | 2018-04-25 | 2018-09-18 | 福州瑞芯微电子股份有限公司 | A kind of operation method and device based on automated cluster neural network chip group |
| CN108416438A (en) * | 2018-05-30 | 2018-08-17 | 济南浪潮高新科技投资发展有限公司 | A kind of convolutional neural networks hardware module dispositions method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109948789A (en) * | 2019-03-21 | 2019-06-28 | 百度在线网络技术（北京）有限公司 | Data load method and device for convolutional neural networks |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Song et al. | Pyramid dilated deeper convlstm for video salient object detection | |
| US20210065058A1 (en) | Method, apparatus, device and readable medium for transfer learning in machine learning | |
| WO2021190127A1 (en) | Data processing method and data processing device | |
| WO2017206936A1 (en) | Machine learning based network model construction method and apparatus | |
| CN108764133A (en) | Image-recognizing method, apparatus and system | |
| CN109871828A (en) | Video frequency identifying method and identification device, storage medium | |
| CN110210513A (en) | Data classification method, device and terminal device | |
| CN109801269B (en) | Tongue fur physique classification method based on competitive extrusion and excitation neural network | |
| CN109614975A (en) | A kind of figure embedding grammar, device and storage medium | |
| US11468296B2 (en) | Relative position encoding based networks for action recognition | |
| CN104391879B (en) | The method and device of hierarchical clustering | |
| CN110321892B (en) | Picture screening method and device and electronic equipment | |
| CN110765960A (en) | Pedestrian re-identification method for adaptive multi-task deep learning | |
| CN109978074A (en) | Image aesthetic feeling and emotion joint classification method and system based on depth multi-task learning | |
| CN111353591A (en) | Computing device and related product | |
| CN109657784A (en) | Neural network creation method and device, image processing method and electronic equipment | |
| Kim et al. | Fine-grained neural architecture search | |
| CN109376849A (en) | A kind of control method and device of convolutional neural networks system | |
| Dvornik et al. | Flow graph to video grounding for weakly-supervised multi-step localization | |
| US20220067495A1 (en) | Intelligent processor, data processing method and storage medium | |
| CN114492581A (en) | Method for classifying small sample pictures based on transfer learning and attention mechanism element learning application | |
| WO2021159633A1 (en) | Method and system for training image recognition model, and image recognition method | |
| CN112434785B (en) | Distributed parallel deep neural network performance evaluation method for supercomputer | |
| CN113609337A (en) | Pre-training method, device, equipment and medium of graph neural network | |
| Jordans et al. | Instruction-set architecture exploration strategies for deeply clustered vliw asips |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190222 |