tensorflow自定义激活函数(带有条件判断)

tensorflow-cpu 2.8.0
python=3.10

@tf.function
def custom_sigmoid(x):
    res = 10 / (1 + tf.math.exp(-x))
    return tf.where(res <= 600, res, 600)

对应:
c u s t o m _ s i g m o i d = 10 1 + exp ⁡ ( − x ) custom\_sigmoid = \cfrac{10}{1 + \exp{(-x)}} custom_sigmoid=1+exp(x)10

参考:
https://tensorflow.google.cn/versions/r2.8/api_docs/python/tf
https://tensorflow.google.cn/versions/r2.8/api_docs/python/tf/where
https://tensorflow.google.cn/versions/r2.8/api_docs/python/tf/math/exp
https://blog.csdn.net/bjay/article/details/115494010

你可能感兴趣的:(深度学习,tensorflow,python,深度学习)