Method for Imputing Missing Values for NN #2
Description
from user fchollet: https://www.kaggle.com/c/predict-west-nile-virus/forums/t/13725/achieve-lb-0-684-in-seconds-with-a-deep-net-using-keras-python/74709#post74709
"With neural networks, if there are missing input values, I recommend filling them with 0s (after the input is preprocessed). A 0 input will be effectively ignored by a neural network (since the first stage is a linear transformation)."
Adding this to remind myself to try this and to document the reference. Also, for others if needed