I now want to normalize my data but I find that she is all negative and I want to know if he is correct.allcombos-22-0-1.fsm
As said you end in a range of ]0..1] after you have normalized your rewards.
You can
- filter negative values.
- never compute negative values.
- add an offset to your not normalized values to receive only an unsigned range.
After you have learned so much more in the past few month, it is a good idea to build a new model instead of adjusting your current working model now. This seems weird, but in fact you will build a different model, because you can consider your better knowledge.
13 People are following this question.
