I'm going to start a project on identifying single nucleotide invariant in genome sequence, using ANN. Which one is worth to spend more time on: Deep Neural Networks or two/three layer Neural Network? Which one results in accurate identifying?
I agree with Leonid on the basis of the Occam's razor / parsimony principle. Of two models with the same explanatory power, one must use the simplest. A predictor is a model. If you want to use a neural network of some sort as a predictor, then you should use the simplest first and climb the ladder of complexity in the direction that addresses your interpretation of the results.
I agree with Leonid and Herve. Moreover, consider that you need a lot of data to train a deep neural network.
I agree with you all !!
"I agree with" is not an answer, it's gust a comment to say something ! :)
- Badges
- Science topic