Python关于维卷积的理解

tf.layers.conv1d(
inputs,
filters,
kernel_size,
strides=1,
padding='valid',
data_fORMat='channels_last',
dilation_rate=1,
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=tf.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None
)

# coding: utf-8
import Tensorflow as tf
 
num_filters = 2
kernel_size = 2
batch_size = 1
seq_length = 4
embedding_dim = 5
 
embedding_inputs = tf.constant(-1.0, shape=[batch_size, seq_length, embedding_dim], dtype=tf.float32)
 
with tf.name_scope("cnn"):
    conv = tf.layers.conv1d(embedding_inputs, num_filters, kernel_size, name='conv')
 
session = tf.Session()
session.run(tf.global_variables_initializer())
 
print (session.run(conv).shape)

    import matplotlib.pyplot as plt
    import numpy as np

#阶跃信号
def up(x):
    for i in range(len(x)):
        if x[i] < 0:
            y[i] = 0
        else:
            y[i] = 1
    return y

x= np.arange(-10,10,0.1)
y = np.zeros(len(x))

y = up(x)

plt.plot(y)
plt.show()



#高斯滤波器
def gauss(x,s):
    
    g=1/(((2*np.pi)**0.5)*s)*np.exp(-x**2/2/(s**2))
    return g

sample = np.arange(-10,10,1)
g = gauss(sample,10/3)

plt.plot(g)
plt.show()

#卷积一
f_1 = np.zeros(len(y)+len(g)-1)

#翻转
g = list(g)
g.reverse()

for i in range(len(f_1)):
    #长数据卷积部分短数据（前半部分）
    if i < (len(g)-1):
        for j in range(i+1):
            f_1[i] = y[j]*g[i-j]+f_1[i]
    #长数据卷积整个短数据
    elif i < (len(y)-1):
        for j in range(i-len(g)+1,i+1):
            f_1[i] = y[j]*g[i-j]+f_1[i]
    #长数据卷积部分短数据（后半部分）
    else:
        for j in range(i-len(g)+1,len(y)):
            f[i] = y[j]*g[i-j]+f[i]
           
               

            
#显示
plt.plot(f_1)
plt.show()


#卷积二

#存放卷积后的结果
f_2 = np.zeros(len(y)+len(g)-1)

#翻转
g = list(g)
g.reverse()

for i in range(len(f_2)):
    
    #根据卷积的不同阶段设置阈值
    t_left = i-len(g)+1
    t_right = i+1
    
    #卷积的前半部分j的初始值
    if t_left < 0:
        t_left = 0
    #卷积的后半部分j的上限
    if t_right > len(y):
        t_right = len(y)
    
    for j in range(t_left,t_right):

        f_2[i] = y[j]*g[i-j] + f_2[i]

#显示        
plt.plot(f_2)
plt.show()