# MemSE: Fast MSE Prediction for Noisy Memristor-Based DNN Accelerators

arXiv: https://arxiv.org/abs/2205.01707 · HTML: https://mlsanity.com/paper/2205.01707/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 3 of its 14 references that are in the map.
- First posted on arXiv 3 May 2022.

## Builds on (most cited first, 3 of 3)

- [Dark Memory and Accelerator-Rich System Optimization in the Dark Silicon Era](https://mlsanity.com/paper/1602.04183/) (2016)
- [Training Modern Deep Neural Networks for Memory-Fault Robustness](https://mlsanity.com/paper/1911.10287/) (2019)
- [Layerwise Noise Maximisation to Train Low-Energy Deep Neural Networks](https://mlsanity.com/paper/1912.10764/) (2019)

## Built on by (most cited first, 2 of 2)

- [Lookbehind-SAM: k steps back, 1 step forward](https://mlsanity.com/paper/2307.16704/) (2023)
- [SAMSON: Sharpness-Aware Minimization Scaled by Outlier Normalization for Improving DNN Generalization and Robustness](https://mlsanity.com/paper/2211.11561/) (2022)

## Abstract (arXiv)

Memristors enable the computation of matrix-vector multiplications (MVM) in memory and, therefore, show great potential in highly increasing the energy efficiency of deep neural network (DNN) inference accelerators. However, computations in memristors suffer from hardware non-idealities and are subject to different sources of noise that may negatively impact system performance. In this work, we theoretically analyze the mean squared error of DNNs that use memristor crossbars to compute MVM. We take into account both the quantization noise, due to the necessity of reducing the DNN model size, and the programming noise, stemming from the variability during the programming of the memristance value. Simulations on pre-trained DNN models showcase the accuracy of the analytical prediction. Furthermore the proposed method is almost two order of magnitude faster than Monte-Carlo simulation, thus making it possible to optimize the implementation parameters to achieve minimal error for a given power constraint.