# Improving Realistic Worst-Case Performance of NVCiM DNN Accelerators through Training with Right-Censored Gaussian Noise

arXiv: https://arxiv.org/abs/2307.15853 · HTML: https://mlsanity.com/paper/2307.15853/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 13 of its 31 references that are in the map.
- First posted on arXiv 29 Jul 2023.

## Builds on (most cited first, 13 of 13)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [Zero-Shot Text-to-Image Generation](https://mlsanity.com/paper/2102.12092/) (2021)
- [BackPACK: Packing more into backprop](https://mlsanity.com/paper/1912.10985/) (2019)
- [Device-Circuit-Architecture Co-Exploration for Computing-in-Memory Neural Accelerators](https://mlsanity.com/paper/1911.00139/) (2019)
- [Formalizing Generalization and Robustness of Neural Networks to Weight Perturbations](https://mlsanity.com/paper/2103.02200/) (2021)
- [Uncertainty Modeling of Emerging Device-based Computing-in-Memory Neural Accelerators with Application to Neural Architecture Search](https://mlsanity.com/paper/2107.06871/) (2021)
- [SWIM: Selective Write-Verify for Computing-in-Memory Neural Accelerators](https://mlsanity.com/paper/2202.08395/) (2022)
- [When Single Event Upset Meets Deep Neural Networks: Observations, Explorations, and Remedies](https://mlsanity.com/paper/1909.04697/) (2019)
- [Computing-In-Memory Neural Network Accelerators for Safety-Critical Systems: Can Small Device Variations Be Disastrous?](https://mlsanity.com/paper/2207.07626/) (2022)
- [CorrectNet: Robustness Enhancement of Analog In-Memory Computing for Neural Networks by Error Suppression and Compensation](https://mlsanity.com/paper/2211.14917/) (2022)
- [RADARS: Memory Efficient Reinforcement Learning Aided Differentiable Neural Architecture Search](https://mlsanity.com/paper/2109.05691/) (2021)

## Built on by (most cited first, 8 of 8)

- [MaxK-GNN: Extremely Fast GPU Kernel Design for Accelerating Graph Neural Networks Training](https://mlsanity.com/paper/2312.08656/) (2023)
- [Robust Implementation of Retrieval-Augmented Generation on Edge-based Computing-in-Memory Architectures](https://mlsanity.com/paper/2405.04700/) (2024)
- [Revisiting Outage for Edge Inference Systems](https://mlsanity.com/paper/2504.03686/) (2025)
- [Compute-in-Memory based Neural Network Accelerators for Safety-Critical Systems: Worst-Case Scenarios and Protections](https://mlsanity.com/paper/2312.06137/) (2023)
- [Zero-Space Cost Fault Tolerance for Transformer-based Language Models on ReRAM](https://mlsanity.com/paper/2401.11664/) (2024)
- [When Small Variations Become Big Failures: Reliability Challenges in Compute-in-Memory Neural Accelerators](https://mlsanity.com/paper/2603.03491/) (2026)
- [U-SWIM: Universal Selective Write-Verify for Computing-in-Memory Neural Accelerators](https://mlsanity.com/paper/2401.05357/) (2023)
- [Interface-Aware KV Cache Quantization for Dense On-Chip NVM in Long-Context LLM Decoding](https://mlsanity.com/paper/2609.05764/) (2026)

## Abstract (arXiv)

Compute-in-Memory (CiM), built upon non-volatile memory (NVM) devices, is promising for accelerating deep neural networks (DNNs) owing to its in-situ data processing capability and superior energy efficiency. Unfortunately, the well-trained model parameters, after being mapped to NVM devices, can often exhibit large deviations from their intended values due to device variations, resulting in notable performance degradation in these CiM-based DNN accelerators. There exists a long list of solutions to address this issue. However, they mainly focus on improving the mean performance of CiM DNN accelerators. How to guarantee the worst-case performance under the impact of device variations, which is crucial for many safety-critical applications such as self-driving cars, has been far less explored. In this work, we propose to use the k-th percentile performance (KPP) to capture the realistic worst-case performance of DNN models executing on CiM accelerators. Through a formal analysis of the properties of KPP and the noise injection-based DNN training, we demonstrate that injecting a novel right-censored Gaussian noise, as opposed to the conventional Gaussian noise, significantly improves the KPP of DNNs. We further propose an automated method to determine the optimal hyperparameters for injecting this right-censored Gaussian noise during the training process. Our method achieves up to a 26% improvement in KPP compared to the state-of-the-art methods employed to enhance DNN robustness under the impact of device variations.