# Improving Noise Tolerance of Mixed-Signal Neural Networks

arXiv: https://arxiv.org/abs/1904.01705 · HTML: https://mlsanity.com/paper/1904.01705/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 Apr 2019.

## Built on by (most cited first, 9 of 9)

- [High-performance deep spiking neural networks with 0.3 spikes per neuron](https://mlsanity.com/paper/2306.08744/) (2023)
- [On the Accuracy of Analog Neural Network Inference Accelerators](https://mlsanity.com/paper/2109.01262/) (2021)
- [Quantum-limited stochastic optical neural networks operating at a few quanta per activation](https://mlsanity.com/paper/2307.15712/) (2023)
- [Random and Adversarial Bit Error Robustness: Energy-Efficient and Secure DNN Accelerators](https://mlsanity.com/paper/2104.08323/) (2021)
- [In-memory Training on Analog Devices with Limited Conductance States via Multi-tile Residual Learning](https://mlsanity.com/paper/2510.02516/) (2025)
- [Effect of Batch Normalization on Noise Resistant Property of Deep Learning Models](https://mlsanity.com/paper/2205.07372/) (2022)
- [Impact of L1 Batch Normalization on Analog Noise Resistant Property of Deep Learning Models](https://mlsanity.com/paper/2205.04886/) (2022)
- [Provable Accuracy Bounds for Hybrid Dynamical Optimization and Sampling](https://mlsanity.com/paper/2410.06397/) (2024)
- [CRIMP: Compact & Reliable DNN Inference on In-Memory Processing via Crossbar-Aligned Compression and Non-ideality Adaptation](https://mlsanity.com/paper/2607.08015/) (2026)

## Abstract (arXiv)

Mixed-signal hardware accelerators for deep learning achieve orders of magnitude better power efficiency than their digital counterparts. In the ultra-low power consumption regime, limited signal precision inherent to analog computation becomes a challenge. We perform a case study of a 6-layer convolutional neural network running on a mixed-signal accelerator and evaluate its sensitivity to hardware specific noise. We apply various methods to improve noise robustness of the network and demonstrate an effective way to optimize useful signal ranges through adaptive signal clipping. The resulting model is robust enough to achieve 80.2% classification accuracy on CIFAR-10 dataset with just 1.4 mW power budget, while 6 mW budget allows us to achieve 87.1% accuracy, which is within 1% of the software baseline. For comparison, the unoptimized version of the same model achieves only 67.7% accuracy at 1.4 mW and 78.6% at 6 mW.