0

I am using a set of time series data with positive and negative values. They are percent changes. I try to predict the next number in the series, but all outputs are positive. Am I doing something wrong, or is this a limitation of tensorflow?

Andy B
  • 39
  • 1
  • 3
  • What is your function to predict outputs? If you are using something like a softmax output layer, then the results are bounded between 0-1. – user3543300 Jan 07 '19 at 22:34
  • https://stackoverflow.com/help/mcve – Luke DeLuccia Jan 07 '19 at 22:38
  • Yes, I'm using softmax. Thanks so much for this response. What function do you recommend for negative outputs? – Andy B Jan 12 '19 at 08:33
  • You can simply use linear activation function for regression problem. Ref: https://stackoverflow.com/questions/46680481/activation-function-for-output-layer-for-regression-models-in-neural-networks – keineahnung2345 Jan 15 '19 at 15:18

0 Answers0