Data error correction method, apparatus, device, and readable storage medium

1. A data error correction method, comprising:

acquiring target data to be error-corrected;

performing error correction on the target data using an error-correcting code (ECC) to obtain first ECC-corrected data;

responsive to determining that the first ECC-corrected data have one or more first errors, correcting the target data using a target neural network to obtain network-corrected data, wherein the network-corrected data comprise a version of the target data corrected by the target neural network;

responsive to determining that the network-corrected data have one or more second errors,

updating the target data to be the network-corrected data; and

performing the error correction on the updated target data using the ECC to obtain second ECC-corrected data; and

responsive to the performing of the error correction on the updated target data being successful, determining the second ECC-corrected data as error-corrected target data which are error free.

2. The data error correction method according to claim 1 , wherein correcting the target data using the target neural network to obtain the network-corrected data comprises:

separating a data part and a check part of the target data; and

correcting the data part using the target neural network to generate a processing result; and

combining the processing result and the check part to obtain the network-corrected data.

3. The data error correction method according to claim 2 , wherein correcting the data part using the target neural network to generate the processing result comprises:

inputting the data part to the target neural network in a reverse order to generate the processing result.

4. The data error correction method according to claim 1 , wherein correcting the target data using the target neural network to obtain the network-corrected data comprises:

inputting the target data to the target neural network in a reverse order to obtain the network-corrected data.

5. The data error correction method according to claim 1 , further comprising:

constructing a neural network comprising an encoding subnetwork and a decoding subnetwork;

inputting training data comprising different types of data to train the neural network; and

responsive to an error loss function of the neural network converging to a steady state during the training, determining the neural network as the target neural network.

6. The data error correction method according to claim 5 , wherein after determining the neural network as the target neural network, the method further comprises:

dividing the training data into multiple data sets, different data in each data set having an association relationship; and

training the target neural network using each data set separately to obtain multiple trained target neural networks corresponding to the multiple data sets, respectively.

7. The data error correction method according to claim 6 , wherein correcting the target data using the target neural network to obtain the network-corrected data comprises:

determining data features of the target data; and

selecting, from the multiple trained target neural networks, a trained target neural network having the highest matching degree with the data features; and

correcting the target data using the trained target neural network to obtain the network-corrected data.

8. A data error correction device, comprising:

a memory configured to store a computer program; and

a processor configured to execute the computer program to perform operations comprising:

acquiring target data to be error-corrected;

performing error correction on the target data using an error-correcting code (ECC) to obtain first ECC-corrected data;

responsive to determining that the first ECC-corrected data have one or more first errors, correcting the target data using a target neural network to obtain network-corrected data, wherein the network-corrected data comprise a version of the target data corrected by the target neural network;

responsive to determining that the network-corrected data have one or more second errors,

updating the target data to be the network-corrected data; and

performing the error correction on the updated target data using the ECC to obtain second ECC-corrected data; and

responsive to the performing of the error correction on the updated target data being successful, determining the second ECC-corrected data as error-corrected target data which are error free.

9. The data error correction device according to claim 8 , wherein to correct the target data using the target neural network to obtain the network-corrected data, the processor is configured to:

separate a data part and a check part of the target data;

correct the data part using the target neural network to generate a processing result; and

combine the processing result with the check part to obtain the network-corrected data.

10. The data error correction device according to claim 9 , wherein to correct the data part using the target neural network to generate the processing result, the processor is configured to:

input the data part to the target neural network in a reverse order to generate the processing result.

11. The data error correction device according to claim 8 , wherein to correct the target data using the target neural network to obtain the network-corrected data, the processor is configured to:

input the target data to the target neural network in a reverse order to obtain the network-corrected data.

12. The data error correction device according to claim 8 , wherein the processor is further configured to:

construct a neural network comprising an encoding subnetwork and a decoding subnetwork;

input training data comprising different types of data to train the neural network; and

responsive to an error loss function of the neural network converging to a steady state during the training, determine the neural network as the target neural network.

13. The data error correction device according to claim 12 , wherein after determining the neural network as the target neural network, the processor is further configured to:

divide the training data into multiple data sets, different data in each data set having an association relationship; and

train the target neural network using each data set separately to obtain multiple trained target neural networks corresponding to the multiple data sets, respectively.

14. The data error correction device according to claim 13 , wherein to correct the target data using the target neural network to obtain the network-corrected data, the processor is further configured to:

determine data features of the target data; and

select, from the multiple trained target neural networks, a trained target neural network having the highest matching degree with the data features; and

correct the target data using the trained target neural network to obtain the network-corrected data.

15. A non-transitory computer-readable storage medium configured to store a computer program, wherein the computer program, when executed by a processor, implements a data error correction method comprising:

acquiring target data to be error-corrected;

performing error correction on the target data using an error-correcting code (ECC) to obtain first ECC-corrected data;

responsive to determining that the first ECC-corrected data have one or more first errors, correcting the target data using a target neural network to obtain network-corrected data, wherein the network-corrected data comprise a version of the target data corrected by the target neural network;

responsive to determining that the network-corrected data have one or more second errors,

updating the target data to be the network-corrected data; and

performing the error correction on the updated target data using the ECC to obtain second ECC-corrected data; and

responsive to the performing of the error correction on the updated target data being successful, determining the second ECC-corrected data as error-corrected target data which are error free.

16. The non-transitory computer-readable storage medium of claim 15 , wherein correcting the target data using the target neural network to obtain the network-corrected data comprises:

separating a data part and a check part of the target data;

correcting the data part using the target neural network to generate a processing result; and

combining the processing result with the check part to obtain the network-corrected data.

17. The non-transitory computer-readable storage medium of claim 16 , wherein correcting the data part using the target neural network to generate the processing result comprises:

inputting the data part to the target neural network in a reverse order to obtain the processing result.

18. The non-transitory computer-readable storage medium of claim 15 , wherein the data error correction method further comprises:

constructing a neural network comprising an encoding subnetwork and a decoding subnetwork;

inputting training data comprising different types of data to train the neural network; and

responsive to an error loss function of the neural network converging to a steady state during the training, determining the neural network as the target neural network.

19. The non-transitory computer-readable storage medium of claim 18 , wherein after determining the neural network as the target neural network, the data error correction method further comprises:

dividing the training data into multiple data sets, different data in each data set having an association relationship; and

training the target neural network using each data set separately to obtain multiple trained target neural networks corresponding to the multiple data sets, respectively.

20. The non-transitory computer-readable storage medium of claim 19 , wherein correcting the target data using the target neural network to obtain the network- corrected data comprises:

determining data features of the target data; and

selecting, from the multiple trained target neural networks, a trained target neural network having the highest matching degree with the data features; and

correcting the target data using the trained target neural network to obtain the network-corrected data.