Home › Neural Network Forums › Encog Project › Using Encog in Java ›

Multiple input neurons

pcasey99's picture
Submitted by pcasey99 on Fri, 07/17/2009 - 13:21
in

How do I add more than one input neuron type in encog-2.0.0? I see how to do it in the java book, but cannot figure what to do in the new framework.

Thanks in advance

Pat Casey

‹ Missing File TrainHopfield does not calculate weight matrix correctly ›
jeffheaton's picture
Submitted by jeffheaton on Mon, 07/20/2009 - 06:04.

What do you mean by add more than one input neuron type? Do you mean activation function? More than one activation function per layer is not supported by Encog, or the book.

pcasey99's picture
Submitted by pcasey99 on Mon, 07/20/2009 - 07:15.

I am referring to the example in the book where the sp500 and prime rate is used as input to the network.

jeffheaton's picture
Submitted by jeffheaton on Tue, 07/21/2009 - 17:49.

The example from the book could be fairly easily converted to work with Encog. Basically it is just stitching the S&P500 & prime interest data together as two neurons times the input window size. The training array produced by this example could be fed right into an Encog BasicNeuralDataSet class.

However, the best way to add the prime interest rate would be to MarketNeuralDataSet. However, it is only designed to incorporate input neurons from Yahoo Finance. It would be a good enhancement to MarketNeuralDataSet to allow it to pull data from multiple sources. This will be considered for a future enhancement.

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