DEFUZZICATION

Defuzzication is a very important part of the Fuzzy Adaptive Intervention Plan as it is the procedure that leads to decisions about the configuration of space and the architectural properties of different objects and spaces. Our speculation is that we need to develop another Neuro-Fuzzy controller that will take up the part of weighing and tuning different rules and proposals regarding the interpretation of the fuzzy output which results from the inference mechanism.

A key element for the defuzzication is to define the structure and function of the n x m matrix that constitutes the platform where both input and output are presented and projected. The information entered in the matrix could be shape (composition of cells), colours (that express different properties) and evidently, position (the grid is a map of Piraeus). As we have already mentioned, the output is expressed in terms of the containment degree, meaning that each cell or pixel in the matrix, takes a number between 0 and 1 that represents the containment of the global to the local. So, the containment degree would depict the degree to which the global pattern of shapes and colours related to a specific position pertains to the local choice expressed in the input. What the defuzzication actually does is to generate decisions about matching this fuzzy number with specific shapes and properties, specific architectural entities.

The structure of the matrix is in a way multi-layered. Each cell corresponds to a "universe of discourse", namely to a fuzzy matrix that assigns different qualities (colours) to the cell, to different degrees. On the other hand, each matrix is contained to another "universe of discourse", to a hyper-matrix that assigns different qualities to the matrix as a whole, and which is shaped according to the users' suggestions. More precisely we can say that the activated cells within the matrix form a new configuration of shapes and colours that define fuzzy neighbourhoods and therefore new matrices.

Depending on the type of information allocated to the input and the way it is conducted, we have two orientations for the development of the defuzzication. In the first alternative the input is a (binary) shape of activated cells that accumulates different intentions. The quality (colour) assigned to the shape is not explicit and therefore it is subject to investigation- formed and interpreted according to the preceding images. In the second alternative the input is a binary allocation of specific qualities which are processed as such all the way. In each case there is a bi-directional communication among the input and the preceding images and the rules that lead to the interpretation of the outputs are changing in time according to the interaction of the local and global factors.