zoukankan      html  css  js  c++  java
  • Coursera, Deep Learning 5, Sequence Models, week3, Sequence models & Attention mechanism

    Sequence to Sequence models

    basic sequence-to-sequence model:

      

     basic image-to-sequence or called image captioning model:

      

    but there are some differences between how you write a model like this to generate a sequence, compared to how you were synthesizing novel text  using a language model. One of the key differences is,you don't want a randomly chosen translation,you maybe want the most likely translation,or you don't want a randomly chosen caption, maybe not,but you might want the best caption and most likely caption.So let's see in the next video how you go about generating that.

     

    Picking the most likely sentence

      

    找出最大可能性的P(y|x),最常用的算法是beam search.

      

    在介绍 beam search 之前,先了解一下 greedy search 已经为什么不用 greedy search?

    greedy search 的意思是,在已知一个值word的情况下,求下一个值word的最可能的情况,以此类推。。。 下图是一个很好的例子说明 greedy search 不适用的情况, 就不如求核能的 y^ 的组合的概率 p(y^1, y^2, ...|x) 然后找出最大概率,当然这样也有问题,就是比如说 10 个word 的输出,在一个 10,000 大的corpus 里就有 10,000 10 种组合情况,需要诉诸于更好的算法,且继续往下看

      

    转载请注明出处 http://www.cnblogs.com/mashuai-191/
  • 相关阅读:
    二分搜索树的深度优先遍历和广度优先遍历
    数据结构与算法之非比较排序【Java】
    数据结构与算法之比较排序【Java】
    像素 转换 px dp
    Toast
    MySQL丶auto_increment
    MariaDB · 性能优化 · Extended Keys
    加唯一索引怎么会导致丢数据
    语句执行错误一· Count(Distinct) ERROR
    innodb参数 · innodb_flush_log_at_trx_commit
  • 原文地址:https://www.cnblogs.com/mashuai-191/p/9005599.html
Copyright © 2011-2022 走看看