Representation Collapse
انهيار التمثيلات
حالة فشل في التعلم ذاتي الإشراف حيث يتقلّص المُرمِّز ليُخرج المتجه ذاته (أو متجهات شبه متطابقة) لكل المدخلات. يفقد التمثيل قدرته التمييزية ويصبح عديم الفائدة.
A failure mode in self-supervised learning where the encoder collapses to output the same (or nearly identical) vector for all inputs. The representation loses its discriminative power and becomes useless.
Also translated asالانهيار التمثيلي، طيّ الفضاء التمثيلي
First appears in this corpus in: Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning (2020)
Appears in these papers
- Barlow Twins: Self-Supervised Learning via Redundancy Reduction2021in the sky ✦
- Barlow Twins: Self-Supervised Learning via Redundancy Reduction2021in the sky ✦
- Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning2020in the sky ✦
- Emerging Properties in Self-Supervised Vision Transformers2021in the sky ✦
- Emerging Properties in Self-Supervised Vision Transformers2021in the sky ✦
- Exploring Simple Siamese Representation Learning2021in the sky ✦
- Exploring Simple Siamese Representation Learning2021in the sky ✦
- VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning2021in the sky ✦