介绍 import numpy as np
class Rnn ():
def 才能;__init__ (input_size,自我,还以为,hidden_size, num_layers,,双向=False):
,,,self.input_size =input_size
,,,self.hidden_size =hidden_size
,,,self.num_layers =num_layers
,,,self.bidirectional =,双向的
def 才能;饲料(自我,,x):
,,,& # 39;& # 39;& # 39;
,,,:param x:, (seq, batch_size,,嵌入)
,,,:返回:,,隐藏
,,,& # 39;& # 39;& # 39;
,,,#,x.shape [9月,,批处理,,特性)
,,,#,hidden.shape (hidden_size,批处理)
,,,#,Whh0.shape [hidden_size, hidden_size], Wih0.shape (hidden_size,,特性)
,,,#,Whh2.shape [hidden_size, hidden_size], Wih2.size [hidden_size, hidden_size]
,,,out =, []
,,,x,, hidden =, np.array (x), [np.zeros ((self.hidden_size, x.shape [1])), for 小姐:拷贝范围(self.num_layers)]
,,,Wih =, (np.random.random ((self.hidden_size, self.hidden_size)), for 小姐:拷贝范围(1,self.num_layers))
,,,Wih.insert (0,, np.random.random ((self.hidden_size, x.shape [2])))
,,,Whh =, (np.random.random ((self.hidden_size, self.hidden_size)), for 小姐:拷贝范围(self.num_layers)]
,,,time =, x.shape [0]
,,,for 小姐:拷贝范围(时间):
,,,,,隐藏[0],=,np.tanh (np.dot(看法。[0],np.transpose (x(我,,…),(1,0))),+
,,,,,,,,,,,,,,np.dot (Whh[0],,隐藏[0])
,,,,,,,,,,,,,,)
,,,,,for 小姐:拷贝范围(1,self.num_layers):
,,,,,,,隐藏[我],=,np.tanh (np.dot(看法。[我],隐藏(张)),+
,,,,,,,,,,,,,,,,,,np.dot (Whh[我],[我])隐藏
,,,,,,,,,,,,,,,,,,)
,,,,,out.append(隐藏[self.num_layers-1])
,,,return np.array(出),np.array(隐藏)
def 乙状结肠(x):
return 才能;1.0/(1.0,+,1.0/np.exp (x))
if __name__ ==, & # 39; __main__ # 39;:
rnn 才能=,Rnn (5 1,,,, 4)
时间=input 才能;np.random.random ((6,, 2,, 1))
,,,h =, rnn.feed(输入)
打印才能(f # 39; seq is {input.shape [0]},, batch_size is {input.shape [1]}, & # 39;,, & # 39; out.shape & # 39;,, out.shape,, & # 39;, h.shape & # 39;,, h.shape)
#,才能打印(乙状结肠(np.random.random ((2, 3))))
,#
#,才能element-wise 乘法
#,才能打印(np.array ([1, 2]) * np.array ((2, 1)))
这期内容当中小编将会给大家带来有关利用numpy怎么实现一个RNN功能,文章内容丰富且以专业的角度为大家分析和叙述,阅读完这篇文章希望大家可以有所收获。
上述就是小编为大家分享的利用numpy怎么实现一个RNN功能了,如果刚好有类似的疑惑,不妨参照上述分析进行理解。如果想知道更多相关知识,欢迎关注行业资讯频道。
