Closed nmheim closed 5 years ago
In addition to normalizing the data, adding a bias vector to the ESNCell improves prediction a lot (commit 9830b5cccf0c19eeed24ef340694b9c8ec917176). Still have to run over the whole dataset to produce some concrete numbers.
Check what the internal TanhRNNCell of pytorch does. There is a possible second bias term for the reservoir layer, which could mean that the input and reservoir layers have two seperate activation functions...
@jamesavery prediction performance should be fixed with 181dc9b0cbcb0ff0b08153716b04baef2590a4a1
In hyperparameter optimization check for best tikhonov beta within every iteration
So far the data was not normalized... is that the reason the predictions are not that good?