Contrastive Loss
الخسارة التبايُنية
دالة خسارة تُقرّب متجهات الأزواج المتشابهة وتُبعد متجهات الأزواج المختلفة بهامش أدنى محدد. صُمّمت خصيصاً لتدريب الشبكات السيامية وهي حجر الأساس في التعلم المتري.
A loss function that pulls feature vectors of similar pairs closer together and pushes those of dissimilar pairs apart by a defined margin. Designed specifically for training Siamese networks and is a cornerstone of metric learning.
Also translated asدالة الخسارة التقابلية
First appears in this corpus in: Signature Verification Using a "Siamese" Time Delay Neural Network (1993)
Appears in these papers
- ALIGN: Scaling Up Visual and Vision-Language Representation Learning with Noisy Text Supervision2021in the sky ✦
- ALIGN: Scaling Up Visual and Vision-Language Representation Learning with Noisy Text Supervision2021in the sky ✦
- Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection2023in the sky ✦
- Signature Verification Using a "Siamese" Time Delay Neural Network1993in the sky ✦
- wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations2020in the sky ✦
- wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations2020in the sky ✦