Custom deep learning layer after a sequence input layer

2 次查看（过去 30 天）

Anouar Yatribi 2020-8-3

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/574606-custom-deep-learning-layer-after-a-sequence-input-layer

编辑： Anouar Yatribi 2020-8-3

I have an issue after implementing a custom DNN in MATLAB 2019b.

The issue is that when I implemented a custom layer after a sequence input layer, I got the following error message (Layer_1 is the name of my custom layer):

Invalid network. Layer_1 does not support sequence input. Try using a sequence folding layer before the layer.

I tried to add a sequence folding layer between the sequence input layer and the custom layer but it doesn't work. However, without sequence folding, when I call the "predict" function of my custom layer with a sequence input (one training sample for example), it gives the expected output. So it seems that the DNN architecture is valid, however, the problem remains in the sequence input layer which is incompatible with the custom layer.

I would be happy to hear some suggestions about this issue.