Thread: neural nets input layer question

  1. #1
    Registered User
    Join Date
    Apr 2009
    Posts
    2

    neural nets input layer question

    hey!

    im learning about neural nets and this may sound simple but its ambigious in all the diagrams i have seen.



    Does this network have one input layer(the inputs), one hidden layer(the 2 neurons) and one output layer(the last neuron).

    Or does it just have one input layer(the 2 inputs and the 2 neurons, and one output layer with no hidden layer?

  2. #2
    Guest Sebastiani's Avatar
    Join Date
    Aug 2001
    Location
    Waterloo, Texas
    Posts
    5,708
    The second statement is correct.
    Code:
    #include <cmath>
    #include <complex>
    bool euler_flip(bool value)
    {
        return std::pow
        (
            std::complex<float>(std::exp(1.0)), 
            std::complex<float>(0, 1) 
            * std::complex<float>(std::atan(1.0)
            *(1 << (value + 2)))
        ).real() < 0;
    }

  3. #3
    Registered User
    Join Date
    Apr 2009
    Posts
    2
    thank you for the reply.

    Now I have to experiment with finding the optimum number of neurons in the hidden layers, and since i don't have to find the optimum number in the input layer, is there some standard amount that is usually expected to be in the input layer?

    Such as one neuron per input?

    I've seen a lot of material on how many neurons should be in the hidden layer, but nothing about the input layer.

  4. #4
    Malum in se abachler's Avatar
    Join Date
    Apr 2007
    Posts
    3,195
    Quote Originally Posted by Sebastiani View Post
    The second statement is correct.
    No it isn't. The network does have one input layer, one intermediate layer, and one output layer, but the intermediate layer isnt the 2 neurons, its the outputs of the two neurons. Specifically a layer is a set of results, which may be inputs from some set of sensors, intermediate results from a layer of neurons, or the outputs fromt eh last set fo neurons.

    The number of inputs is irrelevant to the optimal number of neurons in the hidden layer(s). in theory any set of inputs can be mapped to any set fo outputs if you use one neuron per example per output. This works for simple associations between the input and output sets where a single intermediate layer is sufficient to produce an arbitrarily accurate mapping. However, given the finite accuracy of floating point values and possibility of non-linear relationships between input and output sets, multiple intermediate layers are usually necessary in practical applications.
    Last edited by abachler; 04-29-2009 at 11:26 AM.

Popular pages Recent additions subscribe to a feed

Similar Threads

  1. input stream question
    By lord in forum C++ Programming
    Replies: 4
    Last Post: 09-11-2008, 08:27 PM
  2. String input reliability question
    By WinterMute in forum C++ Programming
    Replies: 2
    Last Post: 06-15-2007, 11:58 PM
  3. Replies: 2
    Last Post: 03-15-2002, 01:45 PM
  4. simple input and string manipulation question
    By Stig in forum C Programming
    Replies: 1
    Last Post: 12-15-2001, 01:33 PM
  5. Handling input errors, general question
    By Unregistered in forum C++ Programming
    Replies: 2
    Last Post: 11-08-2001, 06:21 PM