ABSTRACT
The definition of macroscopic observables with a microscopic foundation for complex systems is one of the approaches taken to understand such systems and predict their future behaviour. In the biological evolution, the notion of information and entropy has been proposed as a candidate for such a quantity. We will argue that its definition should be hierarchical, that is it will depend on the level of abstraction at which the system is observed. We will propose several definitions of information for different levels of evolution and support our claim with simulations from evolutionary algorithms and structure optimization, where we will concentrate on neural systems.
