After running step 2-3 (after the training, to generate the synthetic records) with count variables, all the values of my synthetic records were between 0 and 1, thus binary after rounding. It seems that the default value of medgan.py is --data_type="binary" so we need to add the argument --data_type="count" in the final step (if we have count variables instead of binary ones).
After running step 2-3 (after the training, to generate the synthetic records) with count variables, all the values of my synthetic records were between 0 and 1, thus binary after rounding. It seems that the default value of
medgan.py
is--data_type="binary"
so we need to add the argument--data_type="count"
in the final step (if we have count variables instead of binary ones).