Open rw86347 opened 6 years ago
Do you mean import the trained data into the NN on an Arduino (or wherever)?
In principal, sure; instead of using randomizeWeights
to initialize the edge weights, you'd just write the values you want to _net.Layer[l]->Weight[i][j]
. If you get it working, feel free to make a pull request, it'd be a cool feature :)
Is there someway to do the training offline and then export the NN as a char array?