I am using the following architecture:
[code removed - moderator]
with the following compiler
[code removed - moderator]
I tried many other architectures and optimizers and cant get less than 2 mae.
I am using the following architecture:
[code removed - moderator]
with the following compiler
[code removed - moderator]
I tried many other architectures and optimizers and cant get less than 2 mae.
For LSTM, consider increasing number of units as you go deeper in the network. Start with a much smaller network and see if things take shape.
Please see this link as well.