Backward Pass
التمرير الخلفي
مرحلة حساب الاشتقاقات (المتجهات التفاضلية) بالاتجاه العكسي عبر الطبقات لتحديث أوزان النموذج أثناء التدريب.
The phase of computing gradients in reverse through the layers to update model weights during training.
Also translated asالانتشار الخلفي، التمرير العكسي، المرور الخلفي
First appears in this corpus in: Méthode Générale pour la Résolution des Systèmes d'Équations Simultanées (1847)
Appears in these papers
- Learning Representations by Back-Propagating Errors1986in the sky ✦
- Learning Representations by Back-Propagating Errors1986in the sky ✦
- DeepSeek-V3 Technical Report2024in the sky ✦
- FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning2023in the sky ✦
- FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning2023in the sky ✦
- FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness2022in the sky ✦
- Méthode Générale pour la Résolution des Systèmes d'Équations Simultanées1847in the sky ✦
- Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification2015in the sky ✦
- Optimizing Neural Networks with Kronecker-Factored Approximate Curvature2015in the sky ✦
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism2019in the sky ✦
- Shampoo: Preconditioned Stochastic Tensor Optimization2018in the sky ✦
- Sharpness-Aware Minimization for Efficiently Improving Generalization2021in the sky ✦
- Neural Discrete Representation Learning2017in the sky ✦
- ZeRO: Memory Optimizations Toward Training Trillion Parameter Models2019in the sky ✦