#7 Bug on multilayer perceptron

[virginia estellers]

Hi,

i think there's a bug on the code for the mlp as the network doesn't seem to learn anything after the first epoch and the output values do not change.
Besdies, the sumId() function does not work, as the output labels are repeated after calling that function, but are unique before.
Is it possible that i'm using an old version or too big an input (40 real-valued inputs and some 15000 training examples?). Even if i reduce those sizes, the network performs in the same way...so i guess there might be a trouble on the library.
Any help would be much appreciated.
Thanks

Virginia

[Pablo Alvarado]

Logged In: YES
user_id=430480
Originator: NO

Hello,

I ported the MLP of the ltilib to the ltilib-2 last week and they worked just fine. Maybe you have a problem in the parameterization of the algorithm. Have you tried the example that comes embedded into the documentation?

It would help to have a code example that reproduces the error you are having. (As I said, they are working fine here).

Regards,

Pablo