Query about application of Deep learning models

AI Discussions
1

I am wondering about two queries:

  1. Hidden layers versus visible layers: as visible layers are coded by the programer but how do hidden layers evolve and how to know that these are in 10s, or 1000s or million or billions. Is there any mechanism to assess it? If no then how to say that there are any hidden layers in the neural network?
  2. Input to output mapping: as in regression analysis we provide those inputs on which output functionally depends (for example demand function) but in neural network when many input variables’ data is supplied to generate an output, how can we say that the resultant output is the intended variable outcome (say quantity demanded) but not something else (say quantity supplied)
2

Only the designer of a NN knows how many hidden layers it has.

The number and size of the hidden layers is determined by experimentation and optimization, to give “good enough” performance.

A NN learns from its training set. Whatever inputs and outputs you give it during training, it will make a prediction based on the inputs you give it.

1 Like
3

I got it!