Deep learning based regression framework for read thresholds in a nand flash memory

A computer-implemented method for reading client data from a memory, the computer-implemented method comprising:

• inputting, to a regression model, data related to one or more input conditions associated with the memory, wherein the regression model comprises a neural network, and wherein the neural network comprises an output node corresponding to a voltage read threshold for the memory;
• receiving the voltage read threshold based on the data;
• and
• reading, from the memory storing the client data as a codeword that comprises bits based on an error correction code (ECC) encoding procedure, the client data based on the voltage read threshold, wherein reading the client data comprises decoding the codeword based on an ECC decoding procedure that uses values for the bits based on the voltage read threshold.

A non-transitory computer-readable storage medium storing instructions that, upon execution on a computer system that includes a memory storing client data, cause the computer system to perform operations comprising:

• inputting, to a regression model, data related to one or more input conditions associated with the memory, wherein the regression model comprises a neural network, and wherein the neural network comprises an output node corresponding to a voltage read threshold for the memory;
• receiving the voltage read threshold based on the data;
• and
• reading, from the memory storing the client data as a codeword that comprises bits based on an error correction code (ECC) encoding procedure, the client data based on the voltage read threshold, wherein reading the client data comprises decoding the codeword based on an ECC decoding procedure that uses values for the bits based on the voltage read threshold.