RNA training phase - Artificial Neural Network in Weka

Asked

Viewed 335 times

0

I am developing software that enables users to create new Rnas.

When creating a new Multilayerperceptron RNA in Weka, the testing phase is already performed by lib routines or it is necessary to develop training routines separately?

  • Could you rephrase your question? What testing phase do you refer to? The calculation of weights, learning rate, gradient descent...?

1 answer

0


Testing

Step five is network testing. During this phase the test set is used to determine network performance with data that has not been previously used. The performance of the network, measured at this stage, is a good indication of its actual performance.

Further tests such as analysis of network behavior using special inputs and analysis of current network weights should be considered, as if very small values exist, the associated connections can be considered insignificant and thus be eliminated (prunning). Conversely, substantively higher values than others could indicate that there was over-training of the network.

http://conteudo.icmc.usp.br/pessoas/andre/research/neural/desenv.htm

Browser other questions tagged

You are not signed in. Login or sign up in order to post.