CN112766462A - Data processing method, device and computer readable storage medium - Google Patents
Data processing method, device and computer readable storage medium Download PDFInfo
- Publication number
- CN112766462A CN112766462A CN202110062458.XA CN202110062458A CN112766462A CN 112766462 A CN112766462 A CN 112766462A CN 202110062458 A CN202110062458 A CN 202110062458A CN 112766462 A CN112766462 A CN 112766462A
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- convolutional neural
- incremental
- original
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 30
- 238000012545 processing Methods 0.000 claims abstract description 212
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 168
- 230000008859 change Effects 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013526 transfer learning Methods 0.000 claims description 4
- 238000000034 method Methods 0.000 abstract description 36
- 238000004364 calculation method Methods 0.000 description 19
- 238000013138 pruning Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 230000006870 function Effects 0.000 description 7
- 238000011176 pooling Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000013139 quantization Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
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
-
- 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)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a data processing method, a data processing device and a computer readable storage medium. The method comprises the steps of training a data processing model in advance, wherein the data processing model comprises an input layer, an original convolutional neural network, an incremental convolutional neural network and an output layer, and the incremental convolutional neural network is generated based on the original convolutional neural network and is smaller than the original convolutional neural network in scale; if the computer data to be processed is the change data of the target historical data, processing the computer data to be processed by using the incremental convolutional neural network, and determining a final data processing result of the computer data to be processed according to an original processing result of the target historical data by the original convolutional neural network and an output result of the incremental convolutional neural network; and if the computer data to be processed is not the change data of the target historical data, processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result. The method and the device can improve data processing efficiency and effectively improve the computing performance of hardware computing equipment.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a data processing method and apparatus, and a computer-readable storage medium.
Background
CNN (Convolutional Neural Network) is a kind of Neural Network, and can learn the relationship between input and output by a large amount of data input by imitating the principle of brain vision, and is widely applied to the fields of deep learning, artificial intelligence, big data processing, and the like. At present, CNNs are widely deployed on various computing components such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field-Programmable Gate Array), and the like, where the FPGA is widely applied to various embedded systems, terminals, server accelerator cards, and the like due to its good real-time performance, lower energy consumption, and better portability.
The convolutional neural network generally comprises several convolutional layers conv, pooling layers pool and full-connected layers FC, fig. 1 is a schematic structural diagram of a typical convolutional neural network VGG-16, in which the convolution principle is shown in fig. 2, i.e. convolution is a dot product of two vectors, one of which is the region of the input image 3x3, and the other of which is a fixed coefficient of 3x3, and the fixed coefficient or weight is also called convolution kernel. The pooling principle is shown in fig. 3, and pooling is down-sampling, in which the largest data in the 2 × 2 region is sampled. A fully-connected layer can be viewed as a special convolutional layer with a convolutional kernel size of 1X1, and unlike convolutional layers, the weight coefficients of the full-connection also vary within one graph. SOFTMAX is a probability classification function output by the convolutional neural network; assuming that there is an array A, and Ai represents the ith element in A, the SOFTMAX value of this element is: e.g. of the typeAi/(sum(eAi) E) is an exponential function and sum is the sum of the whole number group.
It can be understood that the larger the calculation amount of the convolutional neural network is, the lower the efficiency of the task executed by the convolutional neural network is, and the lower the performance of the hardware computing device is. In order to reduce the amount of computation of the convolutional neural network, the existing method includes computation acceleration, SVD matrix decomposition; network pruning; compressing the model; network optimization and the like, and methods such as BNN and AdderNet have also appeared in recent years. Taking a pruning method as an example, as shown in fig. 4, the pruning process includes constructing a convolutional neural network model to be pruned, adding a pruning strategy sparseness training model to obtain sparse parameters, defining a pruning strategy and a pruning standard according to the parameters, cutting the model by using the pruning standard, deleting a network channel with an excessively small parameter value, reintegrating the network parameter, constructing a new model, sparsely training the new model again, generating new sparse parameters, and performing a new round of pruning operation; repeating the sparse training and the pruning, continuously updating network parameters until the training precision of the generated latest model is obviously reduced, and stopping the pruning operation; and obtaining a high-efficiency pruning model for hardware operation equipment.
However, the related art performs clipping or quantization mathematically, and although a part of the amount of calculation can be reduced, the amount of calculation is not reduced much when clipping is less or more quantization bits are reserved, and performance is degraded more when clipping is more or less quantization bits are reserved, requiring retraining.
Disclosure of Invention
The application provides a data processing method, a data processing device and a computer readable storage medium, which can improve the data processing efficiency and effectively improve the computing performance of hardware computing equipment.
In order to solve the above technical problems, embodiments of the present invention provide the following technical solutions:
an embodiment of the present invention provides a data processing method, including:
pre-training a data processing model, the data processing model comprising an input layer, an original convolutional neural network, an incremental convolutional neural network, and an output layer, the incremental convolutional neural network being generated based on the original convolutional neural network but smaller in scale than the original convolutional neural network;
if the computer data to be processed is the change data of the target historical data, processing the computer data to be processed by using the incremental convolutional neural network, and determining a data processing result according to the original processing result of the original convolutional neural network on the target historical data and the output result of the incremental convolutional neural network;
and if the computer data to be processed is not the change data of the target historical data, processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result.
Optionally, the processing the computer data to be processed by using the incremental convolutional neural network, and determining a data processing result according to an original processing result of the original convolutional neural network on the target historical data and an output result of the incremental convolutional neural network includes:
acquiring an original processing result of the original convolutional neural network on the target historical data;
inputting the incremental data of the computer data to be processed and the target historical data into the incremental convolution neural network through the input layer to obtain an incremental data result;
and superposing the original processing result and the incremental data result and inputting the superposed results into the output layer to obtain the data processing result.
Optionally, after the original processing result and the incremental data result are superimposed and input to the output layer to obtain the data processing result, the method further includes:
judging whether the difference value between the original processing result and the data processing result is within a preset error range;
if the difference value between the original processing result and the data processing result is not within a preset error range;
and updating the original processing result by utilizing the processing result of the original convolution neural network on the computer data to be processed, and taking the original processing result as the data processing result of the computer data to be processed.
Optionally, the processing the computer data to be processed by using the incremental convolutional neural network, and determining a data processing result according to an original processing result of the original convolutional neural network on the target historical data and an output result of the incremental convolutional neural network includes:
inputting the computer data to be processed into the incremental convolutional neural network to obtain a new data processing result;
and if the difference value between the new data processing result and the original processing result is within a preset error range, outputting the new data processing result as the data processing result.
Optionally, the generating of the incremental convolutional neural network based on the original convolutional neural network but smaller than the size thereof includes:
the model scale of the incremental convolutional neural network is 1/10-1/5 of the model scale of the original convolutional neural network;
and training an incremental convolutional neural network framework by using transfer learning based on the trained original convolutional neural network to obtain network model parameters of the incremental convolutional neural network.
Optionally, the data processing result is a plurality of selected policy probabilities of the monte carlo search tree.
Another aspect of an embodiment of the present invention provides a data processing apparatus, including:
a model pre-training module for pre-training a data processing model, the data processing model comprising an input layer, an original convolutional neural network, an incremental convolutional neural network, and an output layer, the incremental convolutional neural network being generated based on the original convolutional neural network but smaller in scale than the original convolutional neural network;
the data type judging module is used for judging whether the computer data to be processed is the change data of the target historical data;
the data processing module is used for processing the computer data to be processed by utilizing the incremental convolutional neural network if the computer data to be processed is the change data of the target historical data, and determining a data processing result according to the original processing result of the original convolutional neural network on the target historical data and the output result of the incremental convolutional neural network; and if the computer data to be processed is not the change data of the target historical data, processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result.
Optionally, the data processing module is further configured to:
acquiring an original processing result of the original convolutional neural network on the target historical data; inputting the incremental data of the computer data to be processed and the target historical data into the incremental convolution neural network through the input layer to obtain an incremental data result; and superposing the original processing result and the incremental data result and inputting the superposed results into the output layer to obtain the data processing result.
An embodiment of the present invention further provides a data processing apparatus, which includes a processor, and the processor is configured to implement the steps of the data processing method according to any one of the foregoing when executing the computer program stored in the memory.
Finally, an embodiment of the present invention provides a computer-readable storage medium, where a data processing program is stored on the computer-readable storage medium, and when the data processing program is executed by a processor, the data processing method implements the steps of the data processing method according to any one of the foregoing items.
The method has the advantages that a small-scale incremental convolutional neural network is obtained based on the original convolutional neural network, the original convolutional neural network only carries out data processing on original data once, and as long as the current data to be processed is changed data which is processed before, namely the original data, the processing result of the incremental convolutional neural network and the basic convolutional result are used for determining the final data processing result. Because the scale of the incremental convolutional neural network is smaller than that of the original convolutional neural network, the calculated amount of the incremental convolutional neural network is small, and the data processing efficiency of the data to be processed can be improved by greatly reducing the calculated amount of the convolutional neural network; overall, the calculation quantity of the original convolutional neural network can be far lower than that of the incremental convolutional neural network, so that the network calculation quantity can be effectively reduced, the data calculation efficiency is improved, and the calculation performance of hardware operation equipment is effectively improved.
In addition, the embodiment of the invention also provides a corresponding implementation device and a computer readable storage medium for the data processing method, so that the method has higher practicability, and the device and the computer readable storage medium have corresponding advantages.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the related art, the drawings required to be used in the description of the embodiments or the related art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a block diagram of a VGG-16 convolutional neural network architecture of an exemplary application scenario provided by an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating a convolution principle of a VGG-16 convolutional neural network according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating the pooling principle of the VGG-16 convolutional neural network provided by the embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating a pruning principle of a convolutional neural network in the prior art according to an embodiment of the present invention;
fig. 5 is a schematic flowchart of a data processing method according to an embodiment of the present invention;
FIG. 6 is a flow chart illustrating another data processing method according to an embodiment of the present invention;
FIG. 7 is a flowchart illustrating another data processing method according to an embodiment of the present invention;
FIG. 8 is a flow chart illustrating a data processing method according to an illustrative example provided by an embodiment of the present invention;
FIG. 9 is a block diagram of an embodiment of a data processing apparatus according to the present invention;
fig. 10 is a block diagram of another embodiment of a data processing apparatus according to an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the disclosure, the invention will be described in further detail with reference to the accompanying drawings and specific embodiments. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. 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 invention.
The terms "first," "second," "third," "fourth," and the like in the description and claims of this application and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements but may include other steps or elements not expressly listed.
Having described the technical solutions of the embodiments of the present invention, various non-limiting embodiments of the present application are described in detail below.
Referring to fig. 5, fig. 5 is a schematic flow chart of a data processing method according to an embodiment of the present invention, where the embodiment of the present invention includes the following contents:
s501: the data processing model is trained in advance.
The data processing model is pre-trained and used for processing received computer data. The data processing module can comprise an input layer, an original convolutional neural network, an incremental convolutional neural network and an output layer, computer data to be processed is input into the original convolutional neural network or the incremental convolutional neural network through the input layer, processing results of the original convolutional neural network and the incremental convolutional neural network are directly input into the output layer, and output results of the data processing model are processed through the output layer and then output. The original convolutional neural network is a standard convolutional neural network, which can be any convolutional neural network model, and the incremental convolutional neural network is generated based on the original convolutional neural network but is smaller than the original convolutional neural network. That is, the incremental convolutional neural network is identical to the original convolutional neural network except that the number of connections per neuron is different, and other structures such as convolution kernels are identical. As an alternative embodiment, the model size of the incremental convolutional neural network may be preset to 1/10-1/5 of the model size of the original convolutional neural network, for example, it may be 1/8 of the model size of the original convolutional neural network, and the calculation amount of the convolutional neural network is greatly reduced by the small-scale incremental convolutional neural network, which is expected to be increased by more than 10 times compared with the existing design. The network model parameters of the incremental convolutional neural network can be obtained by training the incremental convolutional neural network framework by using transfer learning based on the trained original convolutional neural network, and after the network model parameters are determined, the structure of the incremental convolutional neural network is determined.
S502: and judging whether the computer data to be processed is the change data of the target historical data, if so, executing S503, and if not, executing S504.
In this embodiment, the to-be-processed computer processing includes, but is not limited to, image data, text data, web page data, or a data structure, and the to-be-processed computer data is change data of the target history data, which may be obtained by adding a part of data to the target history data to obtain to-be-processed computer data, modifying a part of data of the target history data to obtain to-be-processed computer data, or deleting a part of data of the target history data to obtain to-be-processed computer data. That is, as long as the to-be-processed computer data is adjusted on the basis of the target history data, the adjustment includes, but is not limited to, addition, deletion, and modification, the to-be-processed computer data is the change data of the target history data, and conversely, the to-be-processed computer data is not the change data of the target history data.
S503: and processing the computer data to be processed by utilizing the incremental convolutional neural network, and determining a data processing result according to an original processing result of the original convolutional neural network on the target historical data and an output result of the incremental convolutional neural network.
In the step, the final data processing result is determined based on the original processing result, because the data input to the incremental convolutional neural network can be changed data or the whole to-be-processed computer data, if the changed data is the data, the final processing result of the to-be-processed computer data needs to be determined together according to the output result of the incremental convolutional neural network and the original processing result, and if the changed data is the whole to-be-processed computer data, the processing result of the to-be-processed computer data is the output result of the incremental convolutional neural network, and the original processing result can be used as a reference for measuring whether the output result is accurate.
It should be noted that, in order to ensure the processing accuracy of the data processing result, the incremental data of the computer data to be processed, i.e. the changed data, should not be too large compared with the incremental data of the target historical data, and may be applied to, for example, a neural network computation of reinforcement learning type, which generally includes a large number of slightly changed neural network computation branches and is insensitive to small computation errors.
S504: and processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result.
In this step, the computer data to be processed can be directly input into the original convolutional neural network through the input layer, the output result of the original convolutional neural network is the final result of the computer data to be processed, and the data result can be stored into the system to be used as the original processing result of the subsequent change data of the computer data to be processed.
In the technical scheme provided by the embodiment of the invention, a small-scale incremental convolutional neural network is obtained based on an original convolutional neural network, the original convolutional neural network only carries out data processing on original data once, and as long as the current data to be processed is the processed data, namely the changed data of the original data, the processing result of the incremental convolutional neural network and the basic convolutional result are utilized to determine the final data processing result. Because the scale of the incremental convolutional neural network is smaller than that of the original convolutional neural network, the calculated amount of the incremental convolutional neural network is small, and the data processing efficiency of the data to be processed can be improved by greatly reducing the calculated amount of the convolutional neural network; overall, the calculation quantity of the original convolutional neural network can be far lower than that of the incremental convolutional neural network, so that the network calculation quantity can be effectively reduced, the data calculation efficiency is improved, and the calculation performance of hardware operation equipment is effectively improved.
In the above embodiment, how to execute step S504 is not limited, but an implementation manner is provided in this embodiment, a tiny increment may be first sent to an increment convolutional neural network to obtain an increment result; the convolution result is obtained by adding the result to the basic convolution result, as shown in fig. 6 and 7, and the method may include the following steps:
acquiring an original processing result of the original convolutional neural network on target historical data;
inputting incremental data of computer data to be processed and target historical data into an incremental convolution neural network through an input layer to obtain an incremental data result;
and superposing the original processing result and the incremental data result and inputting the superposed result into an output layer to obtain a data processing result.
As a further guarantee of data processing accuracy, the convolutional neural network error can also be tracked, the original convolutional neural network result can be periodically compared with the new data processing result, for example, every 10 times, and if the error is large, the output result of the original convolutional neural network result is readjusted, that is, the original processing result is readjusted, which may include the following:
judging whether the difference value of the original processing result and the data processing result is within a preset error range;
if the difference value between the original processing result and the data processing result is not within the preset error range;
and updating the original processing result by using the processing result of the original convolutional neural network to the computer data to be processed, and taking the original processing result as the data processing result of the computer data to be processed.
The present application further provides another data processing result output method, which is parallel to the foregoing method, and therefore, as another optional implementation, the method may include:
inputting computer data to be processed into an incremental convolutional neural network to obtain a new data processing result;
and if the difference value between the new data processing result and the original processing result is within the preset error range, outputting the new data processing result as a data processing result.
The preset error range may be determined according to an actual application scenario, which is not limited in this application.
According to the embodiment, the convolution neural network result is periodically tracked, and the old result is timely replaced by the new convolution neural network reference result before the error is increased, so that the data processing precision is guaranteed.
In an alternative embodiment, referring to fig. 8, in a reinforcement learning application, a convolutional neural network is typically used to construct a monte carlo search tree, and the data processing result may be a plurality of selected strategy probabilities of the monte carlo search tree. In this embodiment, the fast convolution calculation is a process of processing by using an incremental convolution neural network, the standard convolution calculation is a process of processing by using an original convolution neural network, an error function may be constructed by using the probability, the error function may be obtained by, for example, mean square error calculation, when an error between an output result of the incremental convolution neural network and an original processing result is greater than a given threshold, generation operation of the original processing result is re-executed, the amount of calculation is reduced in a complicated calculation process by using an incremental calculation method, and the calculation efficiency of the device is improved.
It should be noted that, in the present application, there is no strict sequential execution order among the steps, and as long as a logical order is met, the steps may be executed simultaneously or according to a certain preset order, and fig. 5 to 8 are only schematic manners, and do not represent only such an execution order.
The embodiment of the invention also provides a corresponding device for the data processing method, so that the method has higher practicability. Wherein the means can be described separately from the functional module point of view and the hardware point of view. In the following, the data processing apparatus provided by the embodiment of the present invention is introduced, and the data processing apparatus described below and the data processing method described above may be referred to correspondingly.
Based on the angle of the functional module, referring to fig. 9, fig. 9 is a structural diagram of a data processing apparatus according to an embodiment of the present invention, in a specific implementation, the apparatus may include:
a model pre-training module 901, configured to pre-train a data processing model, where the data processing model includes an input layer, an original convolutional neural network, an incremental convolutional neural network, and an output layer, and the incremental convolutional neural network is generated based on the original convolutional neural network but has a smaller scale than the original convolutional neural network;
a data type determining module 902, configured to determine whether the computer data to be processed is changed data of the target historical data;
the data processing module 903 is configured to, if the computer data to be processed is changed data of the target historical data, process the computer data to be processed using the incremental convolutional neural network, and determine a data processing result according to an original processing result of the original convolutional neural network on the target historical data and an output result of the incremental convolutional neural network; and if the computer data to be processed is not the change data of the target historical data, processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result.
Optionally, in some embodiments of this embodiment, the data processing module 903 may be further configured to: acquiring an original processing result of the original convolutional neural network on target historical data; inputting incremental data of computer data to be processed and target historical data into an incremental convolution neural network through an input layer to obtain an incremental data result; and superposing the original processing result and the incremental data result and inputting the superposed result into an output layer to obtain a data processing result.
As an optional implementation manner of this embodiment, the data processing module 903 may further include a result tracking sub-module, for example, where the result tracking sub-module is configured to determine whether a difference between an original processing result and a data processing result is within a preset error range; if the difference value between the original processing result and the data processing result is not within the preset error range; and updating the original processing result by using the processing result of the original convolutional neural network to the computer data to be processed, and taking the original processing result as the data processing result of the computer data to be processed.
As a parallel implementation manner to the above embodiment, the data processing module 903 may be further configured to: inputting computer data to be processed into an incremental convolutional neural network to obtain a new data processing result; and if the difference value between the new data processing result and the original processing result is within the preset error range, outputting the new data processing result as a data processing result.
Optionally, in other embodiments of this embodiment, the model pre-training module 901 may be further configured to: the model scale of the incremental convolutional neural network is 1/10-1/5 of the model scale of the original convolutional neural network; and training an incremental convolutional neural network framework by using transfer learning based on the trained original convolutional neural network to obtain network model parameters of the incremental convolutional neural network.
The functions of the functional modules of the data processing apparatus according to the embodiment of the present invention may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the description related to the foregoing method embodiment, which is not described herein again.
Therefore, the embodiment of the invention can improve the data processing efficiency and effectively improve the computing performance of the hardware computing equipment.
The data processing device mentioned above is described from the perspective of functional modules, and further, the present application also provides a data processing device described from the perspective of hardware. Fig. 10 is a block diagram of another data processing apparatus according to an embodiment of the present application. As shown in fig. 10, the apparatus includes a memory 100 for storing a computer program;
a processor 101 for implementing the steps of the data processing method as mentioned in any of the above embodiments when executing the computer program.
The processor 101 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 101 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 101 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 101 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, the processor 101 may further include an AI (Artificial Intelligence) processor for processing computing operations related to machine learning.
Memory 100 may include one or more computer-readable storage media, which may be non-transitory. Memory 100 may also include high speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In this embodiment, the memory 100 is at least used for storing a computer program 1001, wherein after being loaded and executed by the processor 101, the computer program can implement the relevant steps of the data processing method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 100 may also include an operating system 1002, data 1003, and the like, and the storage manner may be a transient storage manner or a permanent storage manner. The operating system 1002 may include Windows, Unix, Linux, etc. The data 1003 may include, but is not limited to, data corresponding to a data processing result, and the like.
In some embodiments, the data processing apparatus may further include a display screen 102, an input/output interface 103, a communication interface 104, a power supply 105, and a communication bus 106.
Those skilled in the art will appreciate that the configuration shown in fig. 10 does not constitute a limitation of the data processing apparatus and may include more or fewer components than those shown, for example, sensor 107.
The functions of the functional modules of the data processing apparatus according to the embodiment of the present invention may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the description related to the foregoing method embodiment, which is not described herein again.
Therefore, the embodiment of the invention can improve the data processing efficiency and effectively improve the computing performance of the hardware computing equipment.
It is to be understood that, if the data processing method in the above-described embodiments is implemented in the form of software functional units and sold or used as a stand-alone product, it may be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application may be substantially or partially implemented in the form of a software product, which is stored in a storage medium and executes all or part of the steps of the methods of the embodiments of the present application, or all or part of the technical solutions. And the aforementioned storage medium includes: a U disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), an electrically erasable programmable ROM, a register, a hard disk, a removable magnetic disk, a CD-ROM, a magnetic or optical disk, and other various media capable of storing program codes.
Based on this, the embodiment of the present invention further provides a computer-readable storage medium, which stores a data processing program, and the data processing program is executed by a processor, and the steps of the data processing method according to any one of the above embodiments are provided.
The functions of the functional modules of the computer-readable storage medium according to the embodiment of the present invention may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the related description of the foregoing method embodiment, which is not described herein again.
Therefore, the embodiment of the invention can improve the data processing efficiency and effectively improve the computing performance of the hardware computing equipment.
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.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
A data processing method, an apparatus and a computer-readable storage medium provided by the present application are described in detail above. The principles and embodiments of the present invention are explained herein using specific examples, which are presented only to assist in understanding the method and its core concepts. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present application.
Claims (10)
1. A data processing method, comprising:
pre-training a data processing model, the data processing model comprising an input layer, an original convolutional neural network, an incremental convolutional neural network, and an output layer, the incremental convolutional neural network being generated based on the original convolutional neural network but smaller in scale than the original convolutional neural network;
if the computer data to be processed is the change data of the target historical data, processing the computer data to be processed by using the incremental convolutional neural network, and determining a data processing result according to the original processing result of the original convolutional neural network on the target historical data and the output result of the incremental convolutional neural network;
and if the computer data to be processed is not the change data of the target historical data, processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result.
2. The data processing method of claim 1, wherein the processing the computer data to be processed using the incremental convolutional neural network and determining a data processing result according to a raw processing result of the raw convolutional neural network on the target historical data and the incremental convolutional neural network output result comprises:
acquiring an original processing result of the original convolutional neural network on the target historical data;
inputting the incremental data of the computer data to be processed and the target historical data into the incremental convolution neural network through the input layer to obtain an incremental data result;
and superposing the original processing result and the incremental data result and inputting the superposed results into the output layer to obtain the data processing result.
3. The data processing method of claim 2, wherein after the superimposing the raw processing result and the incremental data result and inputting the superimposed raw processing result and incremental data result into the output layer to obtain the data processing result, further comprising:
judging whether the difference value between the original processing result and the data processing result is within a preset error range;
if the difference value between the original processing result and the data processing result is not within a preset error range;
and updating the original processing result by utilizing the processing result of the original convolution neural network on the computer data to be processed, and taking the original processing result as the data processing result of the computer data to be processed.
4. The data processing method of claim 1, wherein the processing the computer data to be processed using the incremental convolutional neural network and determining a data processing result according to a raw processing result of the raw convolutional neural network on the target historical data and the incremental convolutional neural network output result comprises:
inputting the computer data to be processed into the incremental convolutional neural network to obtain a new data processing result;
and if the difference value between the new data processing result and the original processing result is within a preset error range, outputting the new data processing result as the data processing result.
5. The data processing method of any of claims 1-4, wherein the generating of the incremental convolutional neural network based on the original convolutional neural network but smaller in scale comprises:
the model scale of the incremental convolutional neural network is 1/10-1/5 of the model scale of the original convolutional neural network;
and training an incremental convolutional neural network framework by using transfer learning based on the trained original convolutional neural network to obtain network model parameters of the incremental convolutional neural network.
6. The data processing method of claim 5, wherein the data processing results are a plurality of selected policy probabilities of a Monte Carlo search tree.
7. A data processing apparatus, comprising:
a model pre-training module for pre-training a data processing model, the data processing model comprising an input layer, an original convolutional neural network, an incremental convolutional neural network, and an output layer, the incremental convolutional neural network being generated based on the original convolutional neural network but smaller in scale than the original convolutional neural network;
the data type judging module is used for judging whether the computer data to be processed is the change data of the target historical data;
the data processing module is used for processing the computer data to be processed by utilizing the incremental convolutional neural network if the computer data to be processed is the change data of the target historical data, and determining a data processing result according to the original processing result of the original convolutional neural network on the target historical data and the output result of the incremental convolutional neural network; and if the computer data to be processed is not the change data of the target historical data, processing the computer data to be processed by using the original convolutional neural network to obtain a data processing result.
8. The data processing apparatus of claim 7, wherein the data processing module is further configured to:
acquiring an original processing result of the original convolutional neural network on the target historical data; inputting the incremental data of the computer data to be processed and the target historical data into the incremental convolution neural network through the input layer to obtain an incremental data result; and superposing the original processing result and the incremental data result and inputting the superposed results into the output layer to obtain the data processing result.
9. A data processing apparatus comprising a processor for implementing the steps of the data processing method of any one of claims 1 to 6 when executing a computer program stored in a memory.
10. A computer-readable storage medium, on which a data processing program is stored, which when executed by a processor implements the steps of the data processing method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110062458.XA CN112766462A (en) | 2021-01-18 | 2021-01-18 | Data processing method, device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110062458.XA CN112766462A (en) | 2021-01-18 | 2021-01-18 | Data processing method, device and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112766462A true CN112766462A (en) | 2021-05-07 |
Family
ID=75702744
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110062458.XA Pending CN112766462A (en) | 2021-01-18 | 2021-01-18 | Data processing method, device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112766462A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494661A (en) * | 2022-03-31 | 2022-05-13 | 清华大学 | Three-dimensional point cloud data processing method and equipment of incremental sparse three-dimensional convolution |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109146060A (en) * | 2018-08-09 | 2019-01-04 | 郑州云海信息技术有限公司 | A kind of method and device based on convolutional neural networks processing data |
| CN110135510A (en) * | 2019-05-22 | 2019-08-16 | 电子科技大学中山学院 | Dynamic domain self-adaptive method, equipment and computer readable storage medium |
| CN110309837A (en) * | 2019-07-05 | 2019-10-08 | 北京迈格威科技有限公司 | Data processing method and image processing method based on convolutional neural networks characteristic pattern |
| CN110874636A (en) * | 2018-09-04 | 2020-03-10 | 杭州海康威视数字技术股份有限公司 | Neural network model compression method and device and computer equipment |
-
2021
- 2021-01-18 CN CN202110062458.XA patent/CN112766462A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109146060A (en) * | 2018-08-09 | 2019-01-04 | 郑州云海信息技术有限公司 | A kind of method and device based on convolutional neural networks processing data |
| CN110874636A (en) * | 2018-09-04 | 2020-03-10 | 杭州海康威视数字技术股份有限公司 | Neural network model compression method and device and computer equipment |
| CN110135510A (en) * | 2019-05-22 | 2019-08-16 | 电子科技大学中山学院 | Dynamic domain self-adaptive method, equipment and computer readable storage medium |
| CN110309837A (en) * | 2019-07-05 | 2019-10-08 | 北京迈格威科技有限公司 | Data processing method and image processing method based on convolutional neural networks characteristic pattern |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494661A (en) * | 2022-03-31 | 2022-05-13 | 清华大学 | Three-dimensional point cloud data processing method and equipment of incremental sparse three-dimensional convolution |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11875268B2 (en) | Object recognition with reduced neural network weight precision | |
| US11593658B2 (en) | Processing method and device | |
| WO2019018375A1 (en) | Neural architecture search for convolutional neural networks | |
| CN112001498A (en) | Data identification method and device based on quantum computer and readable storage medium | |
| Gural et al. | Memory-Optimal Direct Convolutions for Maximizing Classification Accuracy in Embedded Applications. | |
| US11568212B2 (en) | Techniques for understanding how trained neural networks operate | |
| CN111731326B (en) | Obstacle avoidance strategy determination method and device and storage medium | |
| CN111105017A (en) | Neural network quantization method and device and electronic equipment | |
| US20230092453A1 (en) | Parameter updating method and apparatus and storage medium | |
| EP3607496A1 (en) | Conditional graph execution based on prior simplified graph execution | |
| CN113348472A (en) | Convolutional neural network with soft kernel selection | |
| CN114781650A (en) | Data processing method, device, equipment and storage medium | |
| CN112766462A (en) | Data processing method, device and computer readable storage medium | |
| US20200074277A1 (en) | Fuzzy input for autoencoders | |
| CN116721420B (en) | Semantic segmentation model construction method and system for ultraviolet image of electrical equipment | |
| WO2022018548A1 (en) | Online training of neural networks | |
| CN114792097B (en) | Method and device for determining prompt vector of pre-training model and electronic equipment | |
| CN112580787B (en) | Data processing method, device and equipment of neural network accelerator and storage medium | |
| CN110852361B (en) | Image classification method and device based on improved deep neural network and electronic equipment | |
| CN114169923A (en) | Photovoltaic power station power prediction method, system, terminal and storage medium | |
| US11574196B2 (en) | Dynamic management of weight update bit length | |
| CN114267422A (en) | Method and system for predicting surface water quality parameters, computer equipment and storage medium | |
| CN115238873B (en) | Neural network model deployment method and device, and computer equipment | |
| CN117762642B (en) | Convolutional neural network model loading method, device and storage medium | |
| JP7000586B2 (en) | Data processing system and data processing method |
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: 20210507 |