The MES are special evolutive systems introduced to model autonomous natural systems, such as biological, neural or social systems. These systems are open, self-regulated, with a hierarchy of complexity levels; they are able to record their experiences in a flexible way, and to adapt to a variable environment.
The architecture of a MES is a compromise between a parallel distributed system or multi-agents system, and a hierarchical associative net. Indeed, its dynamics is modulated by the cooperation and/or competition between a net of internal partial centers of regulation, the CoRegulators (CR), each of which acts at its own complexity level and according to its own temporality. Their dialectics through functional loops preserving their structural and temporal constraints can account for the emergence of adaptive complex phenomena, such as the development of higher order cognitive processes for a neural system.
Formally, a Memory Evolutive System (MES) is a hierarchical evolutive system on a continuous timescale, in which each link is weighted by a propagation delay, and in which the following sub-evolutive systems are distinguished:
- A hierarchical Memory (Mem) which is progressively developed, with possible formation of increasing levels (through successive complexifications). It represents a record of the previous experiences of the system, allowing for quicker and more efficient responses in presence of the same situations, but flexible enough to be modified if the responses are no more adapted.
- A net of sub-evolutive systems, called Coregulators (CR, or Center of Regulation) whose components (the agents) have differing complexity levels depending on the CR. Each CR has its own timescale, consisting of a sequence of times of the timescale of the MES. And it operates a stepwise trial-and-error learning process, using its differential access to the central Mem at the development of which it participates.
The lower levels CRs represent specialized modules, possibly in relation with the surroundings; in the higher levels, there are associative CRs, with longer delays, which coordinate the action of a number of lower CRs either directly, or indirectly through the constraints they impose.. The net of CRs is not a real hierarchy because several CRs may have the same level (they are said to be parallel), possibly with different timescales (e.g. the temporality is different for countryside or town). An agent can belong to several CRs.
For instance, in a neural system, among the lower CRs we distinguish a CR 'color', a CR 'shape', ... while the cortex has more and more complex associative CRs.