Options

Neural Networks

[Deleted User][Deleted User] Posts: 0 Learner III
in Help
Hello

I should make 9 Neural Networks then import the result of all of them to one Neural Network. Is RM able to do this?

Thank you 
mbs
Tagged:

Best Answer

Answers

  • Options
    hughesfleming68hughesfleming68 Member Posts: 323 Unicorn
    What is the problem exactly? With regression in Tensorflow for example, it is common to run several loops and average the results. You often get much better results than if you tried to remove all the randomness. Embrace the randomness as they say. The problem is time as usual. Rapidminer has ensemble tools for grouping models but simple averaging goes long way and is also easy to understand. 
  • Options
    [Deleted User][Deleted User] Posts: 0 Learner III
    edited February 2020
    @hughesfleming68

    Hello

    Thank you for the answer. I have 9 different big data set and I have to use 9 networks for them then the result of all of them should import to an other network. The question is that is RM able to do this?( this is a new project and I try to make data set from long time ago and I should try it first then I should use python for that, so if I can use RM for that I can test it easily)

    Also ensemble learning is not useful here.

    Thank you
    mbs 
  • Options
    hughesfleming68hughesfleming68 Member Posts: 323 Unicorn
    Hi @mbs, which neural net operator are you using?
  • Options
    [Deleted User][Deleted User] Posts: 0 Learner III
    @hughesfleming68

    Two of them:
    1.Deep learning
    2.Neural network

    I try both of them.

    Thank you for the answer
    mbs
  • Options
    hughesfleming68hughesfleming68 Member Posts: 323 Unicorn
    I guess if we are talking about example sets then the last NN is making a prediction of a prediction from the nine inputs sets. If it is models then one way or another this is still some sort of stacking problem. It is quite hard to visualise what you are trying to do. Maybe sketch out a diagram or perhaps someone else can jump in if I have not quite understood the problem. @varunm1 feel free anytime.
  • Options
    [Deleted User][Deleted User] Posts: 0 Learner III
    edited February 2020
    what is your idea to create an operator for that or write a library in R then use it? ;)
    ( Data is a bit complex )
  • Options
    hughesfleming68hughesfleming68 Member Posts: 323 Unicorn
    @mbs, I am glad that you did. Were your three options equally as good or was one way much better than the others? 

    regards,

    Alex
  • Options
    [Deleted User][Deleted User] Posts: 0 Learner III
    @hughesfleming68

    Three different ways with training part and handling data or make one new library in R. All of them are useful. :)

    Thank you for the time that you spend on my question :)
    mbs
Sign In or Register to comment.