Categorical prediction range change

[Travis Green]

As part of our efforts to make the API's results more consistent, tomorrow at noon PST we will begin testing a system to automatically normalize predicted values from any new categorical models to range between 0 and 1. Models trained before that time will not be affected, but you are welcome to retrain your old model to take advantage of the new normalization. For both old and new models, higher scores will continue to indicate higher likelihood.

If your code interprets prediction results as integers rather than floats/doubles, your code might break. We apologize for any inconvenience this may cause, and hope the new, consistent prediction range improves your use of the API.

-Travis