Research map / Architectures and efficiency / CNN design
In-Place Zero-Space Memory Protection for CNN
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace standard SEC-DED with in-place ECC | VGG16 · ImageNet, fault rate 1e-03 | accuracy drop (%): 0.96 ± 0.14 -> 0.93 ± 0.23 | No clear effect |
| Replace standard SEC-DED with in-place ECC | ResNet18 · ImageNet, fault rate 1e-03 | accuracy drop (%): 2.8 ± 0.31 -> 2.96 ± 0.81 | No clear effect |
| Replace standard SEC-DED with in-place ECC | SqueezeNet · ImageNet, fault rate 1e-03 | accuracy drop (%): 5.37 ± 0.66 -> 5.19 ± 1.08 | No clear effect |
| Use in-place ECC instead of no memory protection | VGG16 · ImageNet, fault rate 1e-03 | accuracy drop (%): 21.93 ± 5.7 -> 0.93 ± 0.23 | Helped |
| Use in-place ECC instead of no memory protection | ResNet18 · ImageNet, fault rate 1e-03 | accuracy drop (%): 72.96 ± 1.48 -> 2.96 ± 0.81 | Helped |
| Use in-place ECC instead of no memory protection | SqueezeNet · ImageNet, fault rate 1e-03 | accuracy drop (%): 64.83 ± 0.5 -> 5.19 ± 1.08 | Helped |
| Use in-place ECC instead of Parity Zero | VGG16 · ImageNet, fault rate 1e-03 | accuracy drop (%): 1.04 ± 0.31 -> 0.93 ± 0.23 | Helped |
| Use in-place ECC instead of Parity Zero | ResNet18 · ImageNet, fault rate 1e-03 | accuracy drop (%): 4.35 ± 1.21 -> 2.96 ± 0.81 | Helped |
| Use in-place ECC instead of Parity Zero | SqueezeNet · ImageNet, fault rate 1e-03 | accuracy drop (%): 8.16 ± 2.4 -> 5.19 ± 1.08 | Helped |
| Apply QAT with throttling rather than ADMM-based training (ablation) | WOT · ImageNet CNN models | accuracy after weight throttling | Helped |
| Apply QAT with throttling to SqueezeNet weights (ablation) | SqueezeNet · ImageNet | top-1 accuracy (%): 11.54% -> 57.11% | Helped |
Closest papers by meaning
- High Performance Zero-Memory Overhead Direct Convolutions (ICML 2018 · similarity 0.84)
- Zero Stability Well Predicts Performance of Convolutional Neural Networks (AAAI 2022 · similarity 0.82)
- ParZC: Parametric Zero-Cost Proxies for Efficient NAS (AAAI 2025 · similarity 0.80)
- MEET: Towards Memory-Efficient Temporal Sparse Deep Neural Networks (CVPR 2025 · similarity 0.80)
- NeuralFuse: Learning to Recover the Accuracy of Access-Limited Neural Network Inference in Low-Voltage Regimes (NeurIPS 2024 · similarity 0.80)
- EfficientViM: Efficient Vision Mamba with Hidden State Mixer based State Space Duality (CVPR 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.