Preventing encoder-decoder mismatch when energy values are too large to be

represented by the laplace encoder (would have a probability of zero due
to finite precision)