# R2F: A Remote Retraining Framework for AIoT Processors with Computing Errors

arXiv: https://arxiv.org/abs/2107.03096 · HTML: https://mlsanity.com/paper/2107.03096/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 13 of its 53 references that are in the map.
- First posted on arXiv 7 Jul 2021.

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

- [TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems](https://mlsanity.com/paper/1603.04467/) (2016)
- [Efficient Processing of Deep Neural Networks: A Tutorial and Survey](https://mlsanity.com/paper/1703.09039/) (2017)
- [Caffe: Convolutional Architecture for Fast Feature Embedding](https://mlsanity.com/paper/1408.5093/) (2014)
- [Robust Machine Learning Systems: Challenges, Current Trends, Perspectives, and the Road Ahead](https://mlsanity.com/paper/2101.02559/) (2021)
- [FT-ClipAct: Resilience Analysis of Deep Neural Networks and Improving their Fault Tolerance using Clipped Activation](https://mlsanity.com/paper/1912.00941/) (2019)
- [Analyzing and Mitigating the Impact of Permanent Faults on a Systolic Array Based Neural Network Accelerator](https://mlsanity.com/paper/1802.04657/) (2018)
- [Deep neural networks are robust to weight binarization and other non-linear distortions](https://mlsanity.com/paper/1606.01981/) (2016)
- [Making Convolutions Resilient via Algorithm-Based Error Detection Techniques](https://mlsanity.com/paper/2006.04984/) (2020)
- [On the Resilience of RTL NN Accelerators: Fault Characterization and Mitigation](https://mlsanity.com/paper/1806.09679/) (2018)
- [AxTrain: Hardware-Oriented Neural Network Training for Approximate Inference](https://mlsanity.com/paper/1805.08309/) (2018)
- [FTT-NAS: Discovering Fault-Tolerant Convolutional Neural Architecture](https://mlsanity.com/paper/2003.10375/) (2020)
- [YodaNN: An Architecture for Ultra-Low Power Binary-Weight CNN Acceleration](https://mlsanity.com/paper/1606.05487/) (2016)
- [Efficient Error-Tolerant Quantized Neural Network Accelerators](https://mlsanity.com/paper/1912.07394/) (2019)

## Built on by (most cited first, 6 of 6)

- [Winograd Convolution: A Perspective from Fault Tolerance](https://mlsanity.com/paper/2202.08675/) (2022)
- [MRFI: An Open Source Multi-Resolution Fault Injection Framework for Neural Network Processing](https://mlsanity.com/paper/2306.11758/) (2023)
- [Statistical Modeling of Soft Error Influence on Neural Networks](https://mlsanity.com/paper/2210.05876/) (2022)
- [Exploring Winograd Convolution for Cost-effective Neural Network Fault Tolerance](https://mlsanity.com/paper/2308.08230/) (2023)
- [Cross-Layer Optimization for Fault-Tolerant Deep Learning](https://mlsanity.com/paper/2312.13754/) (2023)
- [Adaptive Soft Error Protection for Neural Network Processing](https://mlsanity.com/paper/2407.19664/) (2024)

## Abstract (arXiv)

AIoT processors fabricated with newer technology nodes suffer rising soft errors due to the shrinking transistor sizes and lower power supply. Soft errors on the AIoT processors particularly the deep learning accelerators (DLAs) with massive computing may cause substantial computing errors. These computing errors are difficult to be captured by the conventional training on general purposed processors like CPUs and GPUs in a server. Applying the offline trained neural network models to the edge accelerators with errors directly may lead to considerable prediction accuracy loss. To address the problem, we propose a remote retraining framework (R2F) for remote AIoT processors with computing errors. It takes the remote AIoT processor with soft errors in the training loop such that the on-site computing errors can be learned with the application data on the server and the retrained models can be resilient to the soft errors. Meanwhile, we propose an optimized partial TMR strategy to enhance the retraining. According to our experiments, R2F enables elastic design trade-offs between the model accuracy and the performance penalty. The top-5 model accuracy can be improved by 1.93%-13.73% with 0%-200% performance penalty at high fault error rate. In addition, we notice that the retraining requires massive data transmission and even dominates the training time, and propose a sparse increment compression approach for the data transmission optimization, which reduces the retraining time by 38%-88% on average with negligible accuracy loss over a straightforward remote retraining.