






 futureai 
      posted 5/25/2016 23:30           Hi,
I am new to this forum, but not quite new to the world of AI. It's been about a year now since I began and so far I have quite understanding of how neural networks are made and function. So far I know how to feedforward, backpropagate, and train the network. Now it seems as I am stuck in the next process. That is, I want to learn how after I have trained a network I can predict the next output based in data that was already learned. Just wondering if anyone can guide me towards the right direction for this. I am familiar with Python and Java, not so much in C.
Mostly, I am looking to learn the next mathematical algorithm in predicting an output. Is statistics involved in this?
Thanks.
  
   keghn 
      posted 5/27/2016 20:40           For forecasting or stocks prediction, in a un academic way, you take
you parallel inputs, 10 or more, and train a NN to make a prediction. Since this
method does not look at temporal length of the data it is
Markov chain statistical prediction machine.
Or you can turn this NN sideways and have it look at lengths of temporal
Input one would be ten day ago input two would be nine days ago, all
the way up two zero days ago. This NN would slide on the through
the temporal date to train it. Then you would have ten outputs.
That would predict one day a head then the next output would be two
days a head and so on. The more day you predicted ahead and strain
the NN it will act more like a Markov chain machine. And more extreme
prediction it will act like a babbling in efficient prediction machine. It world give a value and an error range + 100% of what is possible instead of something, like you want, of a exact point value and + 5% of Markov play.
The academia does these using RNN and LSTM.
  
   keghn 
      posted 6/2/2016 20:56           Prediction, Jeff Heaton 4:
https://www.youtube.com/watch?v=52IK5wHSkrA
   prediction, Jeff Heaton   
   rouncer 
      posted 8/7/2016 23:44           kegs bang on correct.
u can advance things to take segment symbols and you can get it to sorta be 'dynamic length' even in a fixed size.
  
   AiHasBeenSolved 
  



