machine learning - Number of hidden layers in a neural network model -
would able explain me or point me resources of why (or situations where) more 1 hidden layer necessary or useful in neural network?
- that's more similar way brain works (which might not computational advantage, lot of people researching nn gain insight way mind works, rather solve real world problems.
- its easier achieve kinds of invariance using more layers. example, image classifier works regardless of in image object found, or object's size. see bouvrie, j. , l. rosasco, , t. poggio. "on invariance in hierarchical models". advances in neural information processing systems (nips) 22, 2009.
Comments
Post a Comment