雙向遞迴網絡bidirectional recurrent networkBRN)係一種進階嘅遞迴神經網絡。雙向遞迴網絡嘅特徵係有兩個彼此之間唔相連嘅隱藏層,分別叫向前狀態(forward states;)同向後狀態(backward states;)。一個雙向遞迴網絡每次會讀取 個輸入,foreach [1][2]

最先嗰個時間點嘅 以及最後嗰個時間點嘅 可以會當做一啲事先設定好嘅常數。一個雙向遞迴網絡(右)嘅結構圖解如下[1]

一個雙向遞迴網絡嘅輸出層會由過去同未來嗰度攞訊息-例如想像一個做機械翻譯嘅遞迴神經網絡,佢會攞一連串嘅英文字母做輸入,然後輸出就係一段相應嘅粵文字。喺每個時間點 ,佢會攞 10 個字,然後 foreach 字,佢會有一個輸出,個輸出會取決於打前嘅字同打後嘅字。雙向遞迴網絡最大嘅特徵係能夠埋考慮「未來」嘅訊息,而因為未來嘅訊息好多時都對做預測有用(尤其係喺語言處理上),所以雙向遞迴網絡能夠做到一啲普通嘅遞迴神經網絡做唔到嘅預測[1][3]

睇埋

編輯
  1. 1.0 1.1 1.2 Schuster, Mike, and Kuldip K. Paliwal. "Bidirectional recurrent neural networks." Signal Processing, IEEE Transactions on 45.11 (1997): 2673-2681.2. Awni Hannun, Carl Case, Jared Casper, Bryan Catanzaro, Greg Diamos, Erich Elsen, Ryan
  2. Understanding Bidirectional RNN in PyTorch. Towards Data Science.
  3. T. Robinson, M. Hochberg, and S. Renals, "The use of recurrent neural networks in continuous speech recognition," in Automatic Speech Recognition: Advanced Topics, C. H. Lee, F. K. Soong, and K. K. Paliwal, Eds. Boston, MA: Kluwer, 1996, pp. 233–258.