GlobalOccurance condition

[ramack]

the weights are assinged linearly with the amount of accurance, but this doesn't look reasonable. There are usually a few activities which are done very often (sleeping, way to work, etc.) while the value from the current occurance weighting decreases very fast.

-> use some kind of logarithmic scaleing.

[ramack]

I don't see right now how this would really help. Maybe it is not so bad as it is rigth now...