CN112949781A - Image classification method, device, equipment and storage medium based on improved CNN model - Google Patents
Image classification method, device, equipment and storage medium based on improved CNN model Download PDFInfo
- Publication number
- CN112949781A CN112949781A CN202110429693.6A CN202110429693A CN112949781A CN 112949781 A CN112949781 A CN 112949781A CN 202110429693 A CN202110429693 A CN 202110429693A CN 112949781 A CN112949781 A CN 112949781A
- Authority
- CN
- China
- Prior art keywords
- layer
- input data
- intermediate calculation
- execution
- redundancy
- 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 64
- 238000004364 calculation method Methods 0.000 claims abstract description 108
- 238000004590 computer program Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 abstract description 6
- 238000013527 convolutional neural network Methods 0.000 description 55
- 238000010586 diagram Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
Abstract
The application discloses an image classification method based on an improved CNN model, in the method, a redundancy execution layer can perform redundancy calculation on input data to obtain two intermediate calculation results, and if and only if the two intermediate calculation results are the same, the two intermediate calculation results are transmitted to a next layer, so that soft errors are avoided, and the reliability of classification results is improved. On one hand, in order to avoid serious computational overhead, the method only selects partial layers of the CNN model as redundancy execution layers. On the other hand, since the influence of errors generated by the layers closer to the front on the final result is larger, in the method, the redundant execution layer is arranged before the normal execution layer, and the redundant calculation is performed by the previous layers, so that the classification accuracy is ensured to the maximum extent. In addition, the application also provides an image classification device, equipment and a readable storage medium based on the improved CNN model, and the technical effect of the image classification device, the equipment and the readable storage medium corresponds to the technical effect of the method.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to an image classification method, apparatus, device, and readable storage medium based on an improved CNN model.
Background
As a basis of modern artificial intelligence, CNNs (Convolutional Neural Networks) are widely attracting attention in industry and academia, for example, showing excellent effects in image classification and natural language processing.
Soft errors are a common error factor in modern computer systems, and are usually caused by energetic particles striking memory or logic circuits, which may occur in data or instructions. Unlike permanent hardware errors, soft errors are transient errors that recover themselves over time. Nevertheless, soft errors can still have catastrophic consequences, for example, an unmanned vehicle may not recognize a pedestrian ahead due to the presence of the soft error, resulting in serious consequences.
In summary, how to reduce the influence caused by soft errors and ensure the reliability of the image classification result when using CNN to realize image classification is a problem to be solved by those skilled in the art.
Disclosure of Invention
The application aims to provide an image classification method, device, equipment and readable storage medium based on an improved CNN model, which are used for solving the problem that due to the existence of soft errors, when the CNN model is used for image classification, the reliability of a classification result is low. The specific scheme is as follows:
in a first aspect, the present application provides an image classification method based on an improved CNN model, where the improved CNN model includes a redundant execution layer and a normal execution layer located after the redundant execution layer, and the method includes:
reading an original image as input data of a first one of the redundancy execution layers;
for each redundancy execution layer, respectively processing the input data of two current layers by using the redundancy execution layer to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer;
for each normal execution layer, processing the input data of the current layer by using the normal execution layer to obtain an intermediate calculation result and output the intermediate calculation result as the input data of the next layer until the normal execution layer is the last layer of the improved CNN model;
and outputting the classification result of the original image.
Preferably, after the comparing the two intermediate calculation results, the method further includes:
if the input data is inconsistent with the input data, the redundancy execution layer is used again to process the input data of the current layer, and an intermediate calculation result is obtained and output to be used as the input data of the next layer.
Preferably, the processing operation performed on the input data of the two current layers by using the redundancy execution layer to obtain two intermediate calculation results includes:
processing the input data of the current layer by using the redundancy execution layer to obtain a first intermediate calculation result;
and processing the input data of the current layer by using the redundancy execution layer again to obtain a second intermediate calculation result.
Preferably, the number of the redundant execution layers is 3, and the number of the normal execution layers is 2.
In a second aspect, the present application provides an image classification apparatus based on an improved CNN model, the improved CNN model including a redundant execution layer and a normal execution layer located after the redundant execution layer, the apparatus comprising:
an original image reading module for reading an original image as input data of a first one of the redundancy execution layers;
the redundancy execution module is used for respectively processing the input data of the two current layers by utilizing the redundancy execution layer to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer;
a normal execution module, configured to, for each normal execution layer, perform processing operation on input data of a current layer by using the normal execution layer to obtain an intermediate calculation result and output the intermediate calculation result as input data of a next layer until the normal execution layer is a last layer of the improved CNN model;
and the classification result output module is used for outputting the classification result of the original image.
Preferably, the redundancy execution module is further configured to:
if the input data is inconsistent with the input data, the redundancy execution layer is used again to process the input data of the current layer, and an intermediate calculation result is obtained and output to be used as the input data of the next layer.
Preferably, the redundancy execution module is configured to:
processing the input data of the current layer by using the redundancy execution layer to obtain a first intermediate calculation result;
and processing the input data of the current layer by using the redundancy execution layer again to obtain a second intermediate calculation result.
Preferably, the number of the redundant execution layers is 3, and the number of the normal execution layers is 2.
In a third aspect, the present application provides an image classification device based on an improved CNN model, including:
a memory for storing a computer program;
a processor for executing the computer program to implement the image classification method based on the improved CNN model as described above.
In a fourth aspect, the present application provides a readable storage medium having stored thereon a computer program for implementing the image classification method based on an improved CNN model as described above when being executed by a processor.
The application provides an image classification method based on an improved CNN model, wherein the improved CNN model comprises a redundancy execution layer and a normal execution layer positioned behind the redundancy execution layer, and in the classification process, an original image is read firstly to serve as input data of a first redundancy execution layer; for each redundancy execution layer, respectively processing the input data of the two current layers by using the redundancy execution layer to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer; for each normal execution layer, processing the input data of the current layer by using the normal execution layer to obtain an intermediate calculation result and output the intermediate calculation result as the input data of the next layer until the normal execution layer is the last layer of the improved CNN model; and outputting the classification result of the original image.
Therefore, in the method, the redundancy execution layer performs redundancy calculation on the input data to obtain two intermediate calculation results, and if and only if the two intermediate calculation results are the same, the two intermediate calculation results are transmitted to the next layer, so that soft errors are avoided, and the reliability of the classification result is improved. It can be understood that the more layers of redundancy calculation, the higher the reliability of the classification result, but this also brings about a serious overhead, so this method only selects part of the layers of the CNN model as the redundancy execution layers. In addition, experiments show that the influence of errors generated by the previous layers on the final result is larger, so that the redundant execution layer is arranged before the normal execution layer in the method, and the classification accuracy is ensured to the maximum extent by performing redundant calculation through the previous layers.
In addition, the application also provides an image classification device, equipment and a readable storage medium based on the improved CNN model, and the technical effect of the image classification device, the equipment and the readable storage medium corresponds to the technical effect of the method, and the details are not repeated here.
Drawings
For a clearer explanation of the embodiments or technical solutions of the prior art of the present application, the drawings needed for the description of the embodiments or prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of a first embodiment of an image classification method based on an improved CNN model provided in the present application;
fig. 2 is a schematic structural diagram of a five-layer CNN model in a second embodiment of the image classification method based on an improved CNN model provided in the present application;
fig. 3 is a flowchart of a second embodiment of an image classification method based on an improved CNN model provided in the present application;
fig. 4 is a functional block diagram of an embodiment of an image classification apparatus based on an improved CNN model provided in the present application;
fig. 5 is a schematic structural diagram of an embodiment of an image classification device based on an improved CNN model provided in the present application.
Detailed Description
The core of the application is to provide an image classification method, device, equipment and readable storage medium based on an improved CNN model, and redundancy calculation is performed through partial layers in front of the CNN model, so that the influence caused by soft errors is reduced, excessive calculation overhead is avoided as far as possible, and the reliability of the image classification process is improved.
In order that those skilled in the art will better understand the disclosure, the following detailed description will be given with reference to the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The following describes a first embodiment of an image classification method based on an improved CNN model provided in the present application. As shown in fig. 1, the method comprises the steps of:
s11, reading the original image as the input data of the first redundancy execution layer;
s12, for each redundancy execution layer, respectively processing the input data of the two current layers by using the redundancy execution layer to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer;
s13, for each normal execution layer, processing the input data of the current layer by using the normal execution layer to obtain an intermediate calculation result and output the intermediate calculation result as the input data of the next layer until the normal execution layer is the last layer of the improved CNN model;
and S14, outputting the classification result of the original image.
In this embodiment, the improved CNN model includes a redundant execution layer and a normal execution layer, and the redundant execution layer is located before the normal execution layer. The normal execution layer is the same as the execution layer of the conventional CNN model, and is used for processing the input data of the current layer to obtain an intermediate calculation result and outputting the intermediate calculation result to the next layer. Different from the normal execution layer, the redundant execution layer is used for respectively carrying out the same processing operation on the input data of the two current layers to obtain two intermediate calculation results, and if and only if the two intermediate calculation results are the same, any one of the intermediate calculation results is transmitted to the next layer. By the method, the redundant execution layer can avoid the influence caused by soft errors, and the reliability of the image classification process is improved.
Theoretically, the more layers of redundant computation are performed in the CNN model, the higher the reliability of the system, but this also causes a serious computational overhead. Therefore, the present embodiment selects only a part of the layers for redundancy calculation, thereby improving reliability. It is found through experiments that the more advanced layers, the more wrong data participate in the calculation, and the more influence on the final result is. Therefore, in the embodiment, the first layers of the CNN model are used for performing redundancy calculation, so as to reduce the calculation amount caused by the redundancy calculation and ensure the calculation efficiency. The specific number of the redundancy execution layers may be set or adjusted according to actual requirements, which is not limited in this embodiment.
It can be understood that, in the redundancy execution layer, if the two intermediate calculation results are not consistent, the processing operation on the input data may be repeated to obtain the intermediate calculation result and perform the comparison operation until the comparison result is consistent, and the intermediate calculation result is output to the next layer. Or directly processing the input data of the current layer to obtain an intermediate calculation result and outputting the intermediate calculation result to the next layer.
Specifically, the redundancy execution layer may process two input data one by one, or may process two data simultaneously. When the redundancy execution layer processes two input data one by one, the process is as follows: processing the input data of the current layer by using the redundancy execution layer to obtain a first intermediate calculation result; processing the input data of the current layer by using the redundancy execution layer again to obtain a second intermediate calculation result; and comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the result to the next layer.
In the method, a redundancy execution layer performs redundancy calculation on input data to obtain two intermediate calculation results, and if and only if the two intermediate calculation results are the same, the two intermediate calculation results are transmitted to a next layer, so that a soft error is avoided and the reliability of a classification result is improved. It can be understood that the more layers of redundancy calculation, the higher the reliability of the classification result, but this also brings about a serious overhead, so this method only selects part of the layers of the CNN model as the redundancy execution layers. In addition, experiments show that the influence of errors generated by the previous layers on the final result is larger, so that the redundant execution layer is arranged before the normal execution layer in the method, and the classification accuracy is ensured to the maximum extent by performing redundant calculation through the previous layers.
An embodiment two of the image classification method based on the improved CNN model provided by the present application is described in detail below, and the embodiment two is implemented based on the embodiment one, and the image classification process is explained in detail by taking practical application as an example.
Taking a five-layer CNN network as an example, the model structure and operation process of the scheme are described below.
First, model structure
The purpose of this process is to select the layers that have the greatest impact on the final result for redundant computation. As described above, the influence of the error on the final result is greater for the earlier layers, so the first three layers of the CNN model are selected as the redundant execution layers in this embodiment, and the final improved CNN model structure is as shown in fig. 2, where the number of the redundant execution layers is 3 and the number of the normal execution layers is 2.
Second, the operation process
When an original image is input into the improved CNN model, a first redundancy execution layer processes original data to obtain an intermediate calculation result and stores the intermediate calculation result, then, redundancy calculation is started, and the original data is input into the first redundancy execution layer again to be processed to obtain another intermediate calculation result. Comparing the two intermediate calculation results, if the two intermediate calculation results are the same, indicating that no soft error occurs in the layer, and transmitting the layer output to the lower layer as the lower layer input. If the difference is not the same, the layer is indicated to have errors, the calculation of the layer is executed again, and the calculation result is taken as the final output of the layer.
All the redundancy execution layers are executed according to the operation, and the layers which are not protected by redundancy in the network, namely the normal execution layers, are executed according to the normal execution sequence. The overall operation is shown in fig. 3.
It can be understood that the redundant computation in the CNN model brings a large amount of computation overhead, and the more layers that are protected by redundancy, the larger the computation overhead. For example, when the first layer is protected, the average error coverage rate reaches 25%, the accuracy rate is close to 99%, but the execution time is about 1.1 times that of the unprotected layer; when the first three layers of redundancy protection are carried out, the average error coverage rate reaches 54 percent, the accuracy rate reaches 99 percent, but the execution time is expanded to 1.54 times that of the unprotected time. The sensitivity to soft errors is often different for different network structures, and the more layers of the network, the greater the impact of soft errors, and therefore, more layers of redundancy protection are generally required to improve reliability.
Overall, for the five-layer CNN model, the first three layers of the network are protected by redundant execution, which achieves a better effect. The final experiment shows that when the current three layers are protected by redundancy, the error coverage rate can reach 54%, and the accuracy of the result reaches 99%.
In the following, embodiments of the image classification device based on the improved CNN model provided in the present application are introduced, and the image classification device based on the improved CNN model described below and the image classification method based on the improved CNN model described above may be referred to correspondingly.
In the image classification apparatus based on the improved CNN model of this embodiment, the improved CNN model includes a redundancy execution layer and a normal execution layer located after the redundancy execution layer, as shown in fig. 4, the apparatus includes:
an original image reading module 41 for reading an original image as input data of a first one of the redundancy execution layers;
the redundancy execution module 42 is configured to perform processing operations on the input data of the two current layers by using the redundancy execution layer, respectively, so as to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer;
a normal execution module 43, configured to, for each normal execution layer, perform processing operation on input data of a current layer by using the normal execution layer, obtain an intermediate calculation result, and output the intermediate calculation result as input data of a next layer until the normal execution layer is a last layer of the improved CNN model;
and a classification result output module 44, configured to output a classification result of the original image.
In some specific embodiments, the redundancy execution module is further configured to:
if the input data is inconsistent with the input data, the redundancy execution layer is used again to process the input data of the current layer, and an intermediate calculation result is obtained and output to be used as the input data of the next layer.
In some specific embodiments, the redundant execution module is configured to:
processing the input data of the current layer by using the redundancy execution layer to obtain a first intermediate calculation result;
and processing the input data of the current layer by using the redundancy execution layer again to obtain a second intermediate calculation result.
In some specific embodiments, the number of the redundant execution layers is 3, and the number of the normal execution layers is 2.
The image classification device based on the improved CNN model of this embodiment is used to implement the aforementioned image classification method based on the improved CNN model, so the specific implementation of this device can be found in the foregoing embodiment of the image classification method based on the improved CNN model, and its function corresponds to that of the above method, and is not described here again.
In addition, the present application also provides an image classification apparatus based on the improved CNN model, as shown in fig. 5, including:
a memory 100 for storing a computer program;
a processor 200 for executing the computer program for implementing the image classification method based on the improved CNN model as described above.
Finally, the present application provides a readable storage medium having stored thereon a computer program for implementing the improved CNN model based image classification method as described above when executed by a processor.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The above detailed descriptions of the solutions provided in the present application, and the specific examples applied herein are set forth to explain the principles and implementations of the present application, and the above descriptions of the examples are only used to help understand the method and its core ideas of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (10)
1. An image classification method based on an improved CNN model, wherein the improved CNN model comprises a redundant execution layer and a normal execution layer positioned after the redundant execution layer, the method comprising:
reading an original image as input data of a first one of the redundancy execution layers;
for each redundancy execution layer, respectively processing the input data of two current layers by using the redundancy execution layer to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer;
for each normal execution layer, processing the input data of the current layer by using the normal execution layer to obtain an intermediate calculation result and output the intermediate calculation result as the input data of the next layer until the normal execution layer is the last layer of the improved CNN model;
and outputting the classification result of the original image.
2. The method of claim 1, wherein after said comparing said two intermediate calculation results, further comprising:
if the input data is inconsistent with the input data, the redundancy execution layer is used again to process the input data of the current layer, and an intermediate calculation result is obtained and output to be used as the input data of the next layer.
3. The method of claim 1, wherein the utilizing the redundant execution layer to perform respective processing operations on the input data of the two current layers to obtain two intermediate calculation results comprises:
processing the input data of the current layer by using the redundancy execution layer to obtain a first intermediate calculation result;
and processing the input data of the current layer by using the redundancy execution layer again to obtain a second intermediate calculation result.
4. The method of any of claims 1 to 3, wherein the number of redundant execution layers is 3 and the number of normal execution layers is 2.
5. An image classification apparatus based on an improved CNN model, wherein the improved CNN model includes a redundant execution layer and a normal execution layer located after the redundant execution layer, the apparatus comprising:
an original image reading module for reading an original image as input data of a first one of the redundancy execution layers;
the redundancy execution module is used for respectively processing the input data of the two current layers by utilizing the redundancy execution layer to obtain two intermediate calculation results; comparing the two intermediate calculation results, and if the two intermediate calculation results are consistent, outputting the intermediate calculation result as input data of the next layer;
a normal execution module, configured to, for each normal execution layer, perform processing operation on input data of a current layer by using the normal execution layer to obtain an intermediate calculation result and output the intermediate calculation result as input data of a next layer until the normal execution layer is a last layer of the improved CNN model;
and the classification result output module is used for outputting the classification result of the original image.
6. The apparatus of claim 5, wherein the redundancy execution module is further to:
if the input data is inconsistent with the input data, the redundancy execution layer is used again to process the input data of the current layer, and an intermediate calculation result is obtained and output to be used as the input data of the next layer.
7. The apparatus of claim 5, wherein the redundancy execution module is to:
processing the input data of the current layer by using the redundancy execution layer to obtain a first intermediate calculation result;
and processing the input data of the current layer by using the redundancy execution layer again to obtain a second intermediate calculation result.
8. The apparatus of any of claims 5 to 7, wherein the number of redundant execution layers is 3 and the number of normal execution layers is 2.
9. An image classification device based on an improved CNN model, comprising:
a memory for storing a computer program;
a processor for executing the computer program to implement the method for image classification based on an improved CNN model according to any one of claims 1 to 4.
10. A readable storage medium, having stored thereon a computer program for implementing the method of image classification based on an improved CNN model according to any one of claims 1 to 4 when being executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110429693.6A CN112949781A (en) | 2021-04-21 | 2021-04-21 | Image classification method, device, equipment and storage medium based on improved CNN model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110429693.6A CN112949781A (en) | 2021-04-21 | 2021-04-21 | Image classification method, device, equipment and storage medium based on improved CNN model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112949781A true CN112949781A (en) | 2021-06-11 |
Family
ID=76233142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110429693.6A Pending CN112949781A (en) | 2021-04-21 | 2021-04-21 | Image classification method, device, equipment and storage medium based on improved CNN model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112949781A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150100839A1 (en) * | 2013-10-07 | 2015-04-09 | King Fahd University Of Petroleum And Minerals | Generalized modular redundancy fault tolerance method for combinational circuits |
| US20190205649A1 (en) * | 2017-12-28 | 2019-07-04 | Microsoft Technology Licensing, Llc | Querying video data with reduced latency and cost |
| US20190318822A1 (en) * | 2018-04-13 | 2019-10-17 | International Business Machines Corporation | Deep image classification of medical images |
-
2021
- 2021-04-21 CN CN202110429693.6A patent/CN112949781A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150100839A1 (en) * | 2013-10-07 | 2015-04-09 | King Fahd University Of Petroleum And Minerals | Generalized modular redundancy fault tolerance method for combinational circuits |
| US20190205649A1 (en) * | 2017-12-28 | 2019-07-04 | Microsoft Technology Licensing, Llc | Querying video data with reduced latency and cost |
| US20190318822A1 (en) * | 2018-04-13 | 2019-10-17 | International Business Machines Corporation | Deep image classification of medical images |
Non-Patent Citations (1)
| Title |
|---|
| 平丽琪等: "SERN Modeling and Analyzing the Soft Error Reliability of Convolutional Neural Networks", 《2020 GREAT LAKES SYMPOSIUM》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10721326B2 (en) | Method, apparatus, and electronic device for processing consensus requests in a blockchain consensus network | |
| US20230230367A1 (en) | Safety monitor for image misclassification | |
| CN111259078B (en) | Block chain consensus method, apparatus, computer device and storage medium | |
| CN109783014B (en) | Data storage method and device | |
| EP2413506A1 (en) | Method for adding redundancy data to a distributed data storage system and corresponding device | |
| CN112000512A (en) | Data restoration method and related device | |
| CN114792132A (en) | Pulse neural network accelerated computing system, method, equipment and medium | |
| US11210199B2 (en) | Safety monitor for invalid image transform | |
| CN109844774B (en) | Parallel deconvolution computing method, single-engine computing method and related products | |
| CN109034384B (en) | Data processing method and device | |
| CN112949781A (en) | Image classification method, device, equipment and storage medium based on improved CNN model | |
| CN111176881A (en) | Fault-tolerant method of parallel linear processing system based on linear coding | |
| CN103873068A (en) | Low-density-parity-check decoding method and electronic device | |
| CN111309265B (en) | Node storage method, system, server and storage medium based on neural network | |
| CN112015325B (en) | Method for generating decoding matrix, decoding method and corresponding device | |
| CN111863107B (en) | Flash memory error correction method and device | |
| CN108733628B (en) | Parallel matrix multiplication algorithm reinforcing method | |
| CN101957823B (en) | Transaction control method for database | |
| CN110806938B (en) | Multi-core processor-oriented self-adaptive fault-tolerant data parallel computing implementation method | |
| CN114064331B (en) | Fault-tolerant computing method, fault-tolerant computing device, storage medium, and computer apparatus | |
| CN114461703B (en) | Multichannel data processing method and device, electronic equipment and storage medium | |
| US11971803B2 (en) | Safety monitor for invalid image transform | |
| CN115482422B (en) | Training method of deep learning model, image processing method and device | |
| CN111488969B (en) | Execution optimization method and device based on neural network accelerator | |
| CN113360314B (en) | Data processing method, system, equipment and computer readable 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210611 |
|
| RJ01 | Rejection of invention patent application after publication |