how to decide number of layers and number of neurons

https://stats.stackexchange.com/questions/181/how-to-choose-the-number-of-hidden-layers-and-nodes-in-a-feedforward-neural-netw

How to choose the number of hidden layers and nodes in a feedforward neural network?

 

Determining its size (number of neurons) is simple

 

The Input Layer

shape of your training data

 

Specifically, the number of neurons comprising that layer is equal to the number of features (columns) in your data. Some NN configurations add one additional node for a bias term.

 

 

The Output Layer

Machine mode: returns a class label (e.g., "Premium Account"/"Basic Account"). Regression Mode returns a value (e.g., price).

If the NN is a regressor, then the output layer has a single node.

If the NN is a classifier, then it also has a single node unless softmax is used in which case the output layer has one node per class label in your model.

 

The Hidden Layers

How many hidden layers? Well if your data is linearly separable (which you often know by the time you begin coding a NN) then you don't need any hidden layers at all. Of course, you don't need an NN to resolve your data either, but it will still do the job.

 

see the insanely thorough and insightful NN FAQ for an excellent summary of that commentary

comp.ai.neural-nets FAQ, Part 1 of 7: Introduction

 

the situations in which performance improves with a second (or third, etc.) hidden layer are very few. One hidden layer is sufficient for the large majority of problems.

 

'the optimal size of the hidden layer is usually between the size of the input and size of the output layers'. Jeff Heaton, author of Introduction to Neural Networks in Java offers a few more.

Heaton Research Bookstore

 

In sum, for most problems, one could probably get decent performance (even without a second optimization step) by setting the hidden layer configuration using just two rules: (i) number of hidden layers equals one; and (ii) the number of neurons in that layer is the mean of the neurons in the input and output layers.

 

You state that for the majority of problems need only one hidden layer. Perhaps it is better to say that NNs with more hidden layers are extremly hard to train