NAND De-noising with Multiple Threshold-Expert ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
NAND flash memory devices face challenges in maintaining data integrity due to noise susceptibility when storing multiple bits per memory cell, leading to increased computational power requirements that hinder performance, especially in mobile devices.
Innovation Solution
Implementing a method that uses multiple shallow threshold-expert machine learning models to classify data bits and decode using error correction codes, selecting a threshold network based on read information to generate improved memory channel outputs, thereby reducing noise and improving ECC operations with lower latency and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple bits are stored per memory cell to improve manufacturing costs and performance, then storage density increases, but noise susceptibility increases and data integrity deteriorates
Solution Approach 1:
The patent divides the memory cell array into multiple sub-arrays and processes data from different sub-arrays through separate machine learning models. This segmentation allows independent optimization of each sub-array's noise characteristics while maintaining overall high storage density, thereby improving data integrity without sacrificing storage capacity
Solution Approach 2:
The patent introduces machine learning models as intermediary components between the memory cells and the data output. These models act as mediators that detect and correct noise patterns in the read data, enabling reliable data retrieval from high-density multi-bit memory cells while maintaining the benefits of increased storage capacity
2Reliability
If computational power is increased to compensate for noise in multi-bit memory cells, then data integrity improves, but performance in mobile devices deteriorates due to power consumption
Solution Approach 1:
The patent applies machine learning models selectively to only those memory sub-arrays or data portions that exhibit noise characteristics requiring correction. This partial application approach maintains data integrity for problematic regions while avoiding unnecessary computational overhead and power consumption in clean regions, optimizing the trade-off between reliability and energy usage
Solution Approach 2:
The patent dynamically adjusts the complexity and computational parameters of the machine learning models based on the actual noise levels detected in the memory data. When noise is low, simpler models with lower power consumption are used; when noise is high, more sophisticated models are deployed. This adaptive parameter adjustment maintains data integrity while minimizing power consumption across varying operational conditions
Data Source
AI summary
Systems and methods of the present disclosure may be used to improve equalization module architectures for NAND cell read information. For example, embodiments of the present disclosure may provide for de-noising of NAND cell read information using a Multiple Shallow Threshold-Expert Machine Learning Models (MTM) equalizer. An MTM equalizer may include multiple shallow machine learning models, where each machine learning model is trained to specifically solve a classification task (e.g., a binary classification task) corresponding to a weak decision range between two possible read information values for a given NAND cell read operation. Accordingly, during inference, each read sample with a read value within a weak decision range is passed through a corresponding shallow machine learning model (e.g., a corresponding threshold expert) that is associated with (e.g., trained for) the particular weak decision range.