Classification-based error recovery with reinforcement learning
A processing device in a memory sub-system identifies a set of parameters associated with one or more errors detected with respect to a memory device of a memory sub-system. A vector representing the set of parameters is generated. Based on the vector, a classification value corresponding to the one or more errors is generated. Based on the classification value, a set of error recovery operations is selected from a plurality of sets of error recovery operations, and the set of error recovery operations is executed.
1. A system comprising:
a memory device; and
a processing device, operatively coupled with the memory device, to perform operations comprising:
identifying a set of parameters associated with a set of a plurality of errors detected with respect to the memory device of a memory sub-system;
generating a vector representing the set of parameters associated with the set of the plurality of errors;
mapping the vector to a classification value corresponding to the set of the plurality of errors;
selecting, based on the classification value, a first set of a plurality of error recovery operations from a plurality of sets of error recovery operations;
executing the first set of the plurality of error recovery operations; and
generating reinforcement information based at least in part on the classification value and the first set of the plurality of error recovery operations.
2. The system of claim 1 , the operations further comprising determining an error recovery result based on the executing of the first set of the plurality of error recovery operations.
3. The system of claim 1 , wherein the classification value represents an error type corresponding to the set of the plurality of errors.
4. The system of claim 3 , wherein the first set of the plurality of error recovery operations comprises a first ordered combination of operations to be performed to recover from set of the plurality of errors having the error type corresponding to the classification value.
5. The system of claim 1 , the operations further comprising:
adjusting, based on the reinforcement information, one or more weights associated with a classifier function used to generate the classification value.
6. The system of claim 5 , wherein the adjusting comprises one of:
rewarding the one or more weights in response to a passing error recovery result associated with the executing of the first set of the plurality of error recovery operations; or
penalizing the one or more weights in response to a failing error recovery result associated with the executing of the first set of the plurality of error recovery operations.
7. A method comprising:
identifying, by a processing device, a set of parameters associated with a set of a plurality of errors detected with respect to a memory device of a memory sub-system;
generating a vector representing the set of parameters associated with the set of the plurality of errors;
mapping the vector to a classification value corresponding to the set of the plurality of errors;
selecting, based on the classification value, a first set of a plurality of error recovery operations from a plurality of sets of error recovery operations;
executing the first set of the plurality of error recovery operations; and
generating reinforcement information based at least in part on the classification value and the first set of the plurality of error recovery operations.
8. The method of claim 7 , further comprising determining an error recovery result based on the executing of the first set of the plurality of error recovery operations.
9. The method of claim 7 , wherein the classification value represents an error type corresponding to the set of the plurality of errors.
10. The method of claim 9 , wherein the first set of the plurality of error recovery operations comprises an ordered combination of operations configured to recover from the set of the plurality of errors having the error type corresponding to the classification value.
11. The method of claim 7 , further comprising adjusting, based on the reinforcement information, one or more weights associated with at least a portion of a classifier function used to generate the classification value.
12. The method of claim 11 , wherein the adjusting comprises one of:
rewarding the one or more weights in response to a passing error recovery result associated with the executing of the first set of the plurality of error recovery operations; or
penalizing the one or more weights in response to a failing error recovery result associated with the executing of the first set of the plurality of error recovery operations.
13. A non-transitory computer-readable storage medium comprising instructions that, when executed by a processing device, cause the processing device to perform operations comprising:
identifying a set of parameters associated with a set of a plurality of errors detected with respect to a memory device of a memory sub-system;
generating a vector representing the set of parameters associated with the set of the plurality of errors;
mapping the vector to a classification value corresponding to the set of the plurality of errors;
selecting, based on the classification value, a first set of a plurality of error recovery operations from a plurality of sets of error recovery operations;
executing the first set of the plurality of error recovery operations; and
generating reinforcement information based at least in part on the classification value and the first set of the plurality of error recovery operations.
14. The non-transitory computer-readable storage medium of claim 13 , the operations further comprising determining an error recovery result based on the executing of the first set of the plurality of error recovery operations.
15. The non-transitory computer-readable storage medium of claim 13 , wherein the classification value represents an error type corresponding to the set of the plurality of errors.
16. The non-transitory computer-readable storage medium of claim 15 , wherein the first set of the plurality of error recovery operations comprises an ordered combination of operations configured to recover from the set of the plurality of errors having the error type corresponding to the classification value.
17. The non-transitory computer-readable storage medium of claim 13 , the operations further comprising adjusting, based on the reinforcement information, one or more weights associated with at least a portion of a classifier function used to generate the classification value, wherein the adjusting comprises one of:
rewarding the one or more weights in response to a passing error recovery result associated with the executing of the first set of the plurality of error recovery operations; or
penalizing the one or more weights in response to a failing error recovery result associated with the executing of the first set of the plurality of error recovery operations.