Neuron Network input variables-Missing data

What's the best approach to create a Neural Network with missing data? I have 8 inputs and 2 output variables.

My input data missing is significant (roughly 60%). I do not wish to make a regression analysis to predict the missing values.

Can I create A Neural Network capable of utilizing the available information in each test?

E.g. test 1: 5 inputs available---> The network functions with 5 inputs

test 2: 3 inputs available---> The network functions with 3 inputs

etc.

Thanks.

Answer 1

[Greg Heath]

Sorry: You have to predict the missing data as best you can.

Greg