Tensorflow中的时间分布卷积层 [英] Time Distributed Convolutional Layers in Tensorflow
本文介绍了Tensorflow中的时间分布卷积层的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！
问题描述
在Keras中，有一个时间分布的包装器，该包装器将一层应用于输入的每个时间片. Tensorflow中有类似的包装器吗?或者如何在Tensorflow中构建时间分布的图层?
In Keras, there is a time distributed wrapper that applies a layer to every temporal slice of an input. Is there a similar wrapper that is available in Tensorflow? Or how do I build time distributed layers in Tensorflow?
推荐答案
您可以尝试进行基本的重塑:
You could try a basic reshape:
- 采用形状为
(?, timeSteps, ...other_dimensions...)的张量 - 将其重塑为
(-1, ...other_dimensions...) - 应用图层/操作
- 将其重塑回
(-1, timeSteps, ...other_dimensions...)
- Take a tensor of shape
(?, timeSteps, ...other_dimensions...) - Reshape it to
(-1, ...other_dimensions...) - Apply the layer/operation
- Reshape it back to
(-1, timeSteps, ...other_dimensions...)
这篇关于Tensorflow中的时间分布卷积层的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
查看全文
