RapidMiner Neural Network

Harumpel

Hello,

which learning enhancements does the Rapid Miner Neural Network operator implement to speed-up learning and improve generalization?

Kind Regards
Theo

land

Hi Theo,
what do you mean by learning enhancements? Sorry, but I'm not too familiar with the slang around Neural Networks.

Greetings,
  Sebastian

Harumpel

Hi Sebastian,

afaik in Neural Networks the basic learning algorithm is backpropagation of the squared error + adjusting the weights according to dE/dw. This so called "naive backpropagation" is quite inefficient, since it takes forever to converge to the optimal configuration in many cases.

Now my question is how exactly (backpropagation) learning is implemented in RapidMiner. I guess I have to take a look into the Joona documentation...

Kind Regards
Theo

land

Hi Theo,
either that or I could ask Ingo who implemented the operator. But currently he's on the CeBit, so that's not an option for now.

Greetings,
  Sebastian