Is it not a good idea to normalize/standardize the observation space used in the environment for training our model?
The observation space often consists of a much higher number of features [i.e 100000, 180,181.5,180.2......], which can lead the model to be mistrained.
Additionally, computing with higher numbers requires more computational power
Is it not a good idea to normalize/standardize the observation space used in the environment for training our model? The observation space often consists of a much higher number of features [i.e 100000, 180,181.5,180.2......], which can lead the model to be mistrained.
Additionally, computing with higher numbers requires more computational power
Thanks