Research map / Adversarial robustness and security / Certified and empirical robustness
Being Friends Instead of Adversaries: Deep Networks Learn from Data Simplified by Other Networks
Simone Marullo, Matteo Tiezzi, Marco Gori, Stefano Melacci
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use a shared auxiliary neural network to simplify inputs, then drop it after training | CNN-B · MNIST-rot-back-image | test error rate: 22.81 ± 0.36 -> 22.22 ± 0.62 | Improved |
| Use a shared auxiliary neural network to simplify inputs, then drop it after training | FC-B · IMDB | test error rate: 13.27 ± 0.19 -> 11.93 ± 0.09 | Improved |
| Use a shared auxiliary neural network to simplify inputs, then drop it after training | FC-B · Wines | test error rate: 17.07 ± 0.11 -> 17.15 ± 0.12 | Hurt |
| Use a shared auxiliary neural network to simplify inputs, then drop it after training | CNN-B · CIFAR-10 | test error rate: 29.75 ± 0.37 -> 29.00 ± 0.36 | Improved |
| Use a shared auxiliary neural network to simplify inputs, then drop it after training | ResNet18 · CIFAR-10-n10 | test error rate: 8.92 ± 0.23 -> 8.10 ± 0.19 | Improved |
| Increase the maximum perturbation penalty eta_max (ablation) | NFT with CNN-A · MNIST-back-image | test error rate | Mixed |
| Use a shorter simplification schedule (ablation) | NFT with CNN-A · MNIST-back-image | test error rate | Hurt |
| Vary the number of auxiliary-network features (ablation) | NFT with CNN-A · MNIST-back-image | test error rate | No clear effect |
| Increase the auxiliary-network learning rate beta (ablation) | NFT with CNN-A · MNIST-back-image | test error rate | Improved |
Closest papers by meaning
- Attribute-Guided Adversarial Training for Robustness to Natural Perturbations (AAAI 2021 · similarity 0.85)
- Conserve-Update-Revise to Cure Generalization and Robustness Trade-off in Adversarial Training (ICLR 2024 · similarity 0.84)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.84)
- VNN: Verification-Friendly Neural Networks with Hard Robustness Guarantees (ICML 2024 · similarity 0.84)
- Towards Trustworthy Deep Learning (AAAI 2024 · similarity 0.84)
- On the Hardness of Training Deep Neural Networks Discretely (AAAI 2025 · similarity 0.84)
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.
