Tensorflow2.0 梯度下降与激活函数

最新推荐文章于 2023-03-02 19:08:14 发布

廷益--飞鸟

最新推荐文章于 2023-03-02 19:08:14 发布

阅读量285

点赞数

分类专栏： TensorFlow

本文链接：https://blog.csdn.net/weixin_45875105/article/details/108623261

版权

TensorFlow 专栏收录该内容

70 篇文章 11 订阅

订阅专栏

Tensorflow2.0 梯度下降与激活函数

"""
    梯度下降与激活函数
"""

import tensorflow as tf

print("==" * 30, "sigmoid")

# sigmoid
# 原始数据
x = tf.linspace(-10.0, 10.0, 10)

# 梯度计算
with tf.GradientTape() as tape:
    tape.watch(x) 
    y = tf.sigmoid(x)  # 函数处理
    grads = tape.gradient(y, x)  # 梯度计算

print("原始数据 x:\n", x.numpy())
print("Sigmoid函数处理后 y:\n", y.numpy())
print("对y求导 y:\n", grads.numpy())

print("==" * 30, "Tanh")
# tanh
x = tf.linspace(-5.0, 5.0, 10)

with tf.GradientTape() as tape:
    tape.watch(x)
    y = tf.tanh(x)
    grads = tape.gradient(y, x)

print("原始数据 x:\n", x.numpy())
print("Tanh函数处理后 y:\n", y.numpy())
print("对y求导 y:\n", grads.numpy())

# relu
print("==" * 30, "Relu")
x = tf.linspace(-1.0, 1.0, 10)
with tf.GradientTape(persistent=True) as tape:
    tape.watch(x)

    y = tf.nn.relu(x)
    y_leaky = tf.nn.leaky_relu(x)
    grads_y = tape.gradient(y, x)
    grads_y_leaky = tape.gradient(y_leaky, x)

print("原始数据 x:\n", x.numpy())
print("relu函数处理后 y:\n", y.numpy())
print("leaky_relu函数处理后 y:\n", y_leaky.numpy())
print("relu结果求导 grads:\n", grads_y.numpy())
print("leaky_relu结果求导 grads:\n", grads_y_leaky.numpy())