End-to-End Learning

التعلُّم من طرف إلى طرف

تدريب النظام بأكمله — من المدخل الخام إلى المخرج النهائي — كوحدة واحدة قابلة للاشتقاق، بحيث يُحسَّن كل مكوّن بشكل مشترك بدل تحسين كل مرحلة على حدة.

Training the entire system — from raw input to final output — as a single differentiable unit, so every component is jointly optimized rather than optimized in isolation.

Also translated asالتدريب الشامل

First appears in this corpus in: End-to-End Training of Deep Visuomotor Policies (2016)

Appears in these papers

Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware2023 in the sky ✦
REALM: Retrieval-Augmented Language Model Pre-Training2020 in the sky ✦
End-to-End Training of Deep Visuomotor Policies2016 in the sky ✦