What type of advanced neural network example would you like to see?

jeffheaton's picture

Comments

converting input and output to and from Neural Network

David B. Moore's picture

How do you go about converting input and output to and from Neural Network say to control a device? Is there a perfered method or process?

What type of NN

sandbergc9's picture

I voted for a financial predictor, however I'm just stuck knowing how to load several streams of data into an array of temporal data points. That is if I have 30 days of stock a, b c and I want day 1 to be in the first temporal point etc., I am confused how to load the data!

Thanks for all your great work!

CJS

pick a window

jeffheaton's picture

Say you have three stocks, a, b and c. You need to pick a window size, say 5 days. You want to pick the direction of stock b based on the these 5 days. You will calculate the percent change from day to day for stocks a, b and c. You will then use a neural network with 15 input neurons and one output neuron. I get 15 from the 5 days for a, b and c. 3 times 5 = 15. Then train on the percent move for b for the next day. Then you will use the last 5 days from a, b and c to attempt to predict the percent move in b for the next day.

to control a device

jeffheaton's picture

It depends on what sort of device you want to control. The Jordan type of neural network is often used for this. If you are talking about a sort of robot you may have one output neuron that specifies to move forward, and two others that indicate left and right turns. The input neurons would be some sort of sensory input, or perhaps a "collision indicator". I really have not done much with device control, mostly a "software guy".

Prediction Forex Market

hapiri's picture

I would like to see and implement of PNN/GRNN and also SVM for c#/java (quite resamble to RBF what You implemented)
If You can do this I will be more than appreciative.


Copyright 2005 - 2012 by Heaton Research, Inc.. Heaton Research™ and Encog™ are trademarks of Heaton Research. Click here for copyright, license and trademark information.