[英]In tf.layers.conv2d, with use_bias=True, are the biases tied or untied?
One more question: 还有一个问题:
If they are tied biases, how can I implement untied biases? 如果它们是有偏见的偏见,我该如何实施不受束缚的偏见?
I am using tensorflow 1.10.0 in python. 我在python中使用tensorflow 1.10.0。
tied biases is used in tf.layers.conv2d
. 捆绑偏差用于tf.layers.conv2d
。
If you want united biases, just turn off use_bias
and create bias variable manually with tf.Variable
or tf.get_variable
same shape with following feature map, finally sum them up. 如果您想要统一的偏置,只需关闭use_bias
并使用tf.Variable
或tf.get_variable
相同形状并带有以下特征图的手动创建偏置变量,最后将它们汇总即可。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.