while True: learn()

while True: learn()

View Stats:
Chaining perceptrons (MLP)
I'm defining a DLL and I have a color perceptron where each output is linked to a second color perceptron and then each of the outputs of the second level of perceptrons is linked to a third layer. When I leave it running for a few minutes, all 3 outputs reach ~91% accuracy. That seems wrong, it should be almost zero (9% ^ 3 = 0.0729%), right? I tried replacing the second layer with genetic algorithms (who have the same error rate) but that didn't help. Is this by design?
< >
Showing 1-1 of 1 comments
GospodinNoob  [developer] 4 Feb @ 1:57am 
HI. Show your scheme, please
< >
Showing 1-1 of 1 comments
Per page: 1530 50