Top7：循环神经网络 RNN LSTM 实现字母分割

代码目的: hihell -> ihello

完整代码

"""
@Time:  2020/8/28/028   14:08
@Author:  Carl
@File:  rnn-hello
@Software:  PyCharm Professional Edition
"""
import tensorflow as tf
import numpy as np
from tensorflow.python.ops.rnn import dynamic_rnn
import tensorflow.contrib as con
# 设置随机种子
tf.set_random_seed(777)

# 建立字典
idx2char = ['h','i','e','l','o']

# 构造数据集
x_data = [0,1,0,2,3,3,] # hihell
"""
x_data = [0, 0, 0, 0, 0, 0]   # hihell
x_one_hot = [[[1, 0, 0, 0, 0],   # h 0
              [0, 1, 0, 0, 0],   # i 1
              [1, 0, 0, 0, 0],   # h 0
              [0, 0, 1, 0, 0],   # e 2
              [0, 0, 0, 1, 0],   # l 3
              [0, 0, 0, 1, 0]]]  # l 3
"""
x_one_hot = np.eye(5)[x_data].reshape(1,-1,5) # 独热编码

print(x_one_hot.shape)

y_data = [[1,0,2,3,3,4]] # ihello  不用独热编码

# 设置参数
batch_size = 1 # 批大小one sentence
sequence_length = 6 # 序列长度  ihello == 6
input_dim = 5 # 独热编码长度 one-hot size

hidden_size = 8 # 隐藏层神经元数量 output from the LSTM. 5 to directly predict one-hot
num_classes = 5 # 类别总数
learning_rate = 0.1 # 学习率

# 定义占位符
X = tf.placeholder(tf.float32,[None,sequence_length,input_dim]) # X input_dim 独热one-hot
Y = tf.placeholder(tf.int64,[None,sequence_length]) # Y label

# 建立模型
# 定义LTSM单元
cell = tf.nn.rnn_cell.LSTMCell(num_units=hidden_size,state_is_tuple=True)
initial_state = cell.zero_state(batch_size,tf.float32) # 设置初始状态0
outputs,_staets = dynamic_rnn(cell,X,initial_state=initial_state,dtype=tf.float32)
print('rnn输出',outputs.shape) # (1,6,8)

# 全连接层
X_for_fc = tf.reshape(outputs,[-1,hidden_size])
outputs = con.layers.fully_connected(inputs=X_for_fc,num_outputs=num_classes,activation_fn=None)
print('全连接输出',outputs.shape) # 经过一层全连接 变为(?,5) (6,5) [batch_size*sequence_length,num_classes]


# 改变维度准备计算序列损失reshape out for sequence_loss
outputs = tf.reshape(outputs,[batch_size,sequence_length,num_classes])
print(outputs.shape) # (1,6,5)

# 计算序列损失
weights = tf.ones([batch_size,sequence_length]) # 所有的权重都是1
sequence_loss = con.seq2seq.sequence_loss(logits=outputs,targets=Y,weights=weights)
loss = tf.reduce_mean(sequence_loss)
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(loss)

# 预测值
prediction = tf.argmax(outputs,axis=2) # 最后的outputs是三维的 所以axis=2
accuracy = tf.reduce_mean(tf.cast(tf.equal(prediction,Y),tf.float32))

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())

    for i in range(50):
        loss_val,_,acc_val = sess.run([loss,optimizer,accuracy],feed_dict={X:x_one_hot,Y:y_data})
        result = sess.run(prediction,feed_dict={X:x_one_hot})
        print(i,'loss:',loss_val,'prediction:',result,'true Y:',y_data,'Accuracy:',acc_val)

        # 预测结果查字典后输出字符串
        result_str = [idx2char[c] for c in np.squeeze(result)]
        print('Prediction str:',result_str,''.join(result_str))

        if acc_val >= 1.0:
            break

    # 用新数据测试
    t_data = [0,2,3,3,0,1] # hellhi
    result = sess.run(prediction,feed_dict={X:np.eye(5)[t_data].reshape(1,-1,5)})
    print(result)

    result_str = [idx2char[c]for c in np.squeeze(result)]
    print('Prediction str:',result_str,''.join(result_str))

结果展示

'''
0 loss: 1.5785435 prediction:  [[3 3 3 3 3 3]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.5
	Prediction str:  ['l', 'l', 'l', 'l', 'l', 'l'] llllll
1 loss: 1.4213508 prediction:  [[3 3 3 3 3 3]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.33333334
	Prediction str:  ['l', 'l', 'l', 'l', 'l', 'l'] llllll
2 loss: 1.278454 prediction:  [[3 3 3 3 3 3]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.33333334
	Prediction str:  ['l', 'l', 'l', 'l', 'l', 'l'] llllll
3 loss: 1.1136963 prediction:  [[2 3 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.33333334
	Prediction str:  ['e', 'l', 'e', 'l', 'l', 'o'] elello
4 loss: 0.90376115 prediction:  [[2 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.6666667
	Prediction str:  ['e', 'h', 'e', 'l', 'l', 'o'] ehello
5 loss: 0.7068202 prediction:  [[2 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.8333333
	Prediction str:  ['e', 'h', 'e', 'l', 'l', 'o'] ehello
6 loss: 0.5666248 prediction:  [[1 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] 0.8333333
	Prediction str:  ['i', 'h', 'e', 'l', 'l', 'o'] ihello
7 loss: 0.4362999 prediction:  [[1 0 2 3 3 4]] true Y:  [[1, 0, 2, 3, 3, 4]] 1.0
	Prediction str:  ['i', 'h', 'e', 'l', 'l', 'o'] ihello
[[1 3 3 4 4 4]]
	Prediction str:  ['i', 'l', 'l', 'o', 'o', 'o'] illooo

'''