I have an idea clear in my mind, but don’t know if I am able to express properly and if it is useful to anyone:

– It is clear that a PREDICTION represents the maximum probability to confirm a configuration from a criterion generating it.

– This otherwise means that the prediction in the domain of configurations represents the lowest possible increase of entropy, matching the instance of the criterion with the observation.

– BUT THIS ALSO MEANS THAT IT REPRESENTS THE MINIMUM FLUCTUATION (I.E. GRADIENT) OF THE ENTROPY IN THE DOMAIN OF PREDICTIONS, REPRESENTING THE CRITERION WHICH ORIGINATED THE PREDICTION ITSELF.

**Equivalent to say that the prediction is the derivative domain of the observation domain.**

NOTE: In this sense when I talk about **orthogonality of DOF by definition** I mean that the hosting model has to be big enough to contain all possible independent observable configurations, with the impossibility to project each on each other.