Quantizing for Noisy Flash Memory Channels

2025-06-21

AI Summary

In high-density multi-level cell (MLC) NAND flash-based processing-in-memory (PIM), reliability degradation caused by device-level noise severely compromises computational accuracy—particularly metrics such as mean squared error (MSE) and peak signal-to-noise ratio (PSNR)—yet existing verification-level optimization methods fail to jointly satisfy stringent accuracy requirements. Method: This work proposes, for the first time, a joint quantization and verification-level optimization framework tailored to flash PIM systems. Formulated to minimize MSE, it co-designs quantization strategies and verification levels to simultaneously suppress quantization error and channel-induced noise. An iterative optimization algorithm enables efficient joint solution. Results: Evaluated on image storage and SwinIR model parameter deployment, the framework significantly reduces storage error, achieving up to a 2.1 dB PSNR improvement. It enhances both computational accuracy and overall system reliability under realistic flash noise conditions.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Flash memory-based processing-in-memory (flash-based PIM) offers high storage capacity and computational efficiency but faces significant reliability challenges due to noise in high-density multi-level cell (MLC) flash memories. Existing verify level optimization methods are designed for general storage scenarios and fail to address the unique requirements of flash-based PIM systems, where metrics such as mean squared error (MSE) and peak signal-to-noise ratio (PSNR) are critical. This paper introduces an integrated framework that jointly optimizes quantization and verify levels to minimize the MSE, considering both quantization and flash memory channel errors. We develop an iterative algorithm to solve the joint optimization problem. Experimental results on quantized images and SwinIR model parameters stored in flash memory show that the proposed method significantly improves the reliability of flash-based PIM systems.

Problem

• Optimizing quantization for noisy flash memory channels
• Addressing reliability challenges in flash-based PIM systems
• Minimizing MSE by joint quantization and verify level optimization

Innovation

• Joint optimization of quantization and verify levels
• Iterative algorithm for minimizing mean squared error
• Enhanced reliability in flash-based PIM systems