CNN / RNN
CNN / RNN
CNN
人均985的知乎上说：降维，融合一切的降维。
结构
- 各种预处理
- n层处理
- n层处理
- CONV：卷积计算层，线性乘积求和。
- RELU：一种对图像识别比较友好的激活函数。
- POOL：池化层，简言之，即取区域平均或最大或者其他方法。
- n层处理
- FC：全连接层，用于分类回归。
RNN
经典结构 - Elman
计算图
图中递归结构w为前段时间隐藏层的输出。注意，计算图展开后权值共享，更新参数时需要同步更新所有值。实际运用过程中还分为
![fig1][fig1]
训练方式：BPTT
按照时间线展开特定次数，然后按照BPTT算法逐步更新，每次更新权值时要使之前的所有层次的权值进行相同更新。
NARX
结构
不局限于当前时间至上一个时刻的相同权值递归，而是采用一定范围内时刻数据进行不同权值递归。不局限于隐藏层自递归，而是将输出结果递归作为输入。训练方式：还是按时序展开，BPTT。
变体 - NARX Series-Parallel
使用并行结构，即训练时输入直接采用标准输出训练，而不是采用生成的预期输出。训练方式：直接BP就可以，训练比原始模型要快速很多。
双向RNN
结构
思想：同时考虑过去和未来数据对当前的影响。不仅有向后的权值，还存在向前的权值。训练方式也为BPTT。
变体 - 深层双向RNN
设计初衷：一次记不下来啊，需要多记几次。
LSTM
结构
解决了RNN梯度消失问题，即难以更新较前位置的权值。
经典模型只有输入门与输出门，最新的模型添加了遗忘门。设计核心是递归输出权值设定为1，此节点简称为细胞。细胞前后存在输入门与输出门加权，递归输入存在遗忘门加权。三个门和总体加权输入包括原始输入、总体输出、细胞输出三个部分。
训练算法
可以参考未来的大佬Alex Graves写的一本书《Supervised Sequence Labelling with Recurrent Neural Networks》，也可以参考LSTM开山之作《LONG SHORT-TERM MEMORY》，就不再抄一遍了。
GRU
结构
LSTM的平民版，使用了同一个门控就同时可以进行遗忘和选择记忆，更容易进行训练，能够很大程度上提高训练效率。
ps：
- RNN还要考虑时间尺度问题，衍生出MTRNN
- 如果时间间隔不同的话，需要特殊处理，比如将时间间隔作为参数控制遗忘门（未经过实验的想法）。
- 一般性结论：只要大佬您能把权值更新算法计算出来，😅结构开心就好~
©️ Copyrights.RSMX.GUILIN.2020-05-17
[fig1]:data:image/jpg;base64,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
