59 lines
2.4 KiB
Plaintext
59 lines
2.4 KiB
Plaintext
{
|
||
"cells": [
|
||
{
|
||
"cell_type": "markdown",
|
||
"id": "729b9613",
|
||
"metadata": {
|
||
"origin_pos": 0
|
||
},
|
||
"source": [
|
||
"# 现代循环神经网络\n",
|
||
":label:`chap_modern_rnn`\n",
|
||
"\n",
|
||
"前一章中我们介绍了循环神经网络的基础知识，\n",
|
||
"这种网络可以更好地处理序列数据。\n",
|
||
"我们在文本数据上实现了基于循环神经网络的语言模型，\n",
|
||
"但是对于当今各种各样的序列学习问题，这些技术可能并不够用。\n",
|
||
"\n",
|
||
"例如，循环神经网络在实践中一个常见问题是数值不稳定性。\n",
|
||
"尽管我们已经应用了梯度裁剪等技巧来缓解这个问题，\n",
|
||
"但是仍需要通过设计更复杂的序列模型来进一步处理它。\n",
|
||
"具体来说，我们将引入两个广泛使用的网络，\n",
|
||
"即*门控循环单元*（gated recurrent units，GRU）和\n",
|
||
"*长短期记忆网络*（long short-term memory，LSTM）。\n",
|
||
"然后，我们将基于一个单向隐藏层来扩展循环神经网络架构。\n",
|
||
"我们将描述具有多个隐藏层的深层架构，\n",
|
||
"并讨论基于前向和后向循环计算的双向设计。\n",
|
||
"现代循环网络经常采用这种扩展。\n",
|
||
"在解释这些循环神经网络的变体时，\n",
|
||
"我们将继续考虑 :numref:`chap_rnn`中的语言建模问题。\n",
|
||
"\n",
|
||
"事实上，语言建模只揭示了序列学习能力的冰山一角。\n",
|
||
"在各种序列学习问题中，如自动语音识别、文本到语音转换和机器翻译，\n",
|
||
"输入和输出都是任意长度的序列。\n",
|
||
"为了阐述如何拟合这种类型的数据，\n",
|
||
"我们将以机器翻译为例介绍基于循环神经网络的\n",
|
||
"“编码器－解码器”架构和束搜索，并用它们来生成序列。\n",
|
||
"\n",
|
||
":begin_tab:toc\n",
|
||
" - [gru](gru.ipynb)\n",
|
||
" - [lstm](lstm.ipynb)\n",
|
||
" - [deep-rnn](deep-rnn.ipynb)\n",
|
||
" - [bi-rnn](bi-rnn.ipynb)\n",
|
||
" - [machine-translation-and-dataset](machine-translation-and-dataset.ipynb)\n",
|
||
" - [encoder-decoder](encoder-decoder.ipynb)\n",
|
||
" - [seq2seq](seq2seq.ipynb)\n",
|
||
" - [beam-search](beam-search.ipynb)\n",
|
||
":end_tab:\n"
|
||
]
|
||
}
|
||
],
|
||
"metadata": {
|
||
"language_info": {
|
||
"name": "python"
|
||
},
|
||
"required_libs": []
|
||
},
|
||
"nbformat": 4,
|
||
"nbformat_minor": 5
|
||
} |