Statistical analysis:
- evaluation function, based on things like: length, energy, forbidden states, etc
- reinforcement based on curve dependend on standard deviation and distance from mean value

Time series analysis:
- most frequent patterns (Apriori-TD) - used as base for plans
- new patterns are remembered for "some time" (adaptive fuzzy)
- patterns are forgotten if they stop ocuring for N rounds (adaptive fuzzy)

Rough set analysis:
- discovering similarity relation for states basing on actions
- grouping goals
- classification of new states and goals
- knowledge generalisation

Memory system:
- information degradation: both knowledge from reinforcement and time-series patterns degrade, this should speed up adaptation to new situations, keep system "clearer", less memory consuming, and will possibly speed up learning, by negative reinforcement of unused knowledge