Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised Representation Learning via Neighborhood-Relational Encoding
Mohammad Sabokrou, Mohammad Khalooei, Ehsan Adeli
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add neighborhood-relational loss to an auto-encoder | auto-encoder, linear SVM · MNIST | accuracy: 0.969 -> 0.978 | Improved |
| add neighborhood-relational loss to an auto-encoder | auto-encoder, RBF SVM · MNIST | accuracy: 0.972 -> 0.984 | Improved |
| use NRE-pretrained AlexNet instead of AE pretraining | AlexNet self-supervised · PASCAL VOC 2007 classification (all layers) | mAP: 53.8 -> 74.4 | Improved |
| use NRE-pretrained AlexNet instead of DeepCluster | AlexNet self-supervised · PASCAL VOC 2007 classification (all layers) | mAP: 73.7 -> 74.4 | Improved |
| use NRE-pretrained AlexNet instead of DeepCluster | AlexNet self-supervised · PASCAL VOC 2007 detection | mAP: 55.4 -> 54.7 | No clear effect |
| use NRE-pretrained AlexNet instead of DeepCluster | AlexNet self-supervised · PASCAL VOC 2012 segmentation | mIoU: 45.1 -> 51.1 | Improved |
| use NRE-pretrained AlexNet instead of supervised AlexNet | AlexNet · PASCAL VOC 2012 segmentation | mIoU: 48.0 -> 51.1 | Improved |
| use NRE-pretrained AlexNet instead of supervised AlexNet | AlexNet · PASCAL VOC 2007 classification (all layers) | mAP: 78.3 -> 74.4 | Improved |
| replace MagNet reconstruction with NRE reconstruction as FGSM defense | MNIST classifier · MNIST FGSM epsilon=0.3 | accuracy: 0.4242 -> 0.7166 | Improved |
| replace MagNet with NRE as FGSM defense | MNIST classifier · MNIST FGSM epsilon=0.01 | accuracy: 0.9238 -> 0.9802 | Improved |
| drop the lambda3 far-sample term (0.6,0.4,0) versus (0.6,0.2,0.2) (ablation) | NRE · UCSD Ped2 | EER (%): 17.5% -> 14% | Improved |
| use NRE single auto-encoder instead of Deep-cascade | anomaly detector · UCSD Ped2 | EER (%): 9% -> 14% | Hurt |
Closest papers by meaning
- Revisiting Self-Supervised Visual Representation Learning (CVPR 2019 · similarity 0.87)
- Memory-Based Neighbourhood Embedding for Visual Recognition (ICCV 2019 · similarity 0.86)
- SERE: Exploring Feature Self-Relation for Self-Supervised Transformer (TPAMI 2023 · similarity 0.86)
- Improving Generalization via Scalable Neighborhood Component Analysis (ECCV 2018 · similarity 0.85)
- Self-Organizing Visual Prototypes for Non-Parametric Representation Learning (ICML 2025 · similarity 0.85)
- A Theoretical Analysis of Self-Supervised Learning for Vision Transformers (ICLR 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.
