hw1: final touch on training procedure description

This commit is contained in:
Claudio Maggioni 2021-05-07 17:12:59 +02:00
parent 85090ae316
commit 68b64da919
3 changed files with 5 additions and 0 deletions

2
.gitignore vendored
View file

@ -422,3 +422,5 @@ TSWLatexianTemp*
# option is specified. Footnotes are the stored in a file with suffix Notes.bib.
# Uncomment the next line to have this generated file ignored.
#*Notes.bib
*.zip

View file

@ -112,6 +112,9 @@ The model I've chosen is a feed-forward neural network with 2 hidden layers:
Finally, the output neuron has a linear activation function. As described
before, the validation set was used to manually tailor the NNs architecture.
The network was trained using the `adam` optimizer with 5000 maximum number of
epochs and with an early stopping procedure that has a patience of 120 epochs.
The fitted model has these final performance parameters:
- Final MSE on validation set $\approx 0.0143344$