Research map / Adversarial robustness and security / Certified and empirical robustness
Proximal Mapping for Deep Regularization
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 |
|---|---|---|---|
| Use proximal mapping for multiview representations instead of RRM | Sketchy image recognition · Sketchy, 20 classes | test error (%): 15.2 ± 0.6 -> 13.7 ± 0.3 | Helped |
| Use proximal mapping for multiview representations instead of RRM | Sketchy image recognition · Sketchy, 125 classes | test error (%): 28.1 ± 0.4 -> 22.0 ± 0.4 | Helped |
| Use proximal mapping for multiview representations instead of RRM | XRMB speech recognition · XRMB, 0% noise, acoustic | PER: 16.1 ± 0.5 -> 12.9 ± 0.4 | Helped |
| Use proximal mapping for multiview representations instead of RRM | XRMB speech recognition · XRMB, 80% noise, logit-avg | PER: 23.9 ± 0.3 -> 20.1 ± 0.5 | Helped |
| Use proximal mapping for multiview embeddings | Crosslingual word embedding · Multilingual WS-REL, EN | Spearman's correlation: 66.6 -> 71.0 | Helped |
| Use proximal mapping for multiview embeddings | Crosslingual word embedding · Multilingual WS-SIM, EN | Spearman's correlation: 78.7 -> 78.3 | Hurt |
| Apply proximal robust regularization to LSTM state updates | ProxLSTM · Japanese vowels (JV) | test accuracy: 94.96 ± 0.44 -> 95.52 ± 0.63 | Helped |
| Apply proximal robust regularization to LSTM state updates | ProxLSTM · Human Activity Recognition (HAR) | test accuracy: 92.01 ± 0.18 -> 92.08 ± 0.23 | No clear effect |
| Apply proximal robust regularization to LSTM state updates | ProxLSTM · Arabic Digits (AD) | test accuracy: 97.45 ± 0.38 -> 97.99 ± 0.29 | Helped |
| Apply proximal robust regularization to LSTM state updates | ProxLSTM · IMDB | test accuracy: 93.65 ± 0.03 -> 94.16 ± 0.11 | Helped |
| Increase the number of Sketchy classes while using ProxNet | ProxNet · Sketchy, 20 to 125 classes | test error (%): 13.7 ± 0.3 -> 22.0 ± 0.4 | Mixed |
Closest papers by meaning
- Deconstructing the Regularization of BatchNorm (ICLR 2021 · similarity 0.87)
- A Kernel Perspective for Regularizing Deep Neural Networks (ICML 2019 · similarity 0.87)
- What's in a Prior? Learned Proximal Networks for Inverse Problems (ICLR 2024 · similarity 0.86)
- Efficient Proximal Mapping of the 1-path-norm of Shallow Networks (ICML 2020 · similarity 0.86)
- Combining Explicit and Implicit Regularization for Efficient Learning in Deep Networks (NeurIPS 2022 · similarity 0.86)
- Adaptive Proximal Gradient Methods for Structured Neural Networks (NeurIPS 2021 · similarity 0.86)
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.