CumInCAD is a Cumulative Index about publications in Computer Aided Architectural Design
supported by the sibling associations ACADIA, CAADRIA, eCAADe, SIGraDi, ASCAAD and CAAD futures

id d481
authors Langley P, Derix C and Coates P
year 2007
title Meta-Cognitive Mappings: Growing Neural Networks for Generative Urbanism
source Generative Arts conference, Milan, 2007
summary This paper examines the use of dynamic learning systems and adaptive topologies within neural networks models, and their implications as a tool for architectural mappings. The principal investigation is the ability of such systems to identify/ map/ model/ represent flows within dynamic data sets and identify topological relationships between these flows. A growing neural network [GNN] model is proposed, able to map dynamic data inputs over time. It is based on Kohonen’s early self-organising feature maps [SOM] and takes as its starting point previous work by CECA with neural networks in an architectural context, as well as other examples of neural gases, and GNNs, in order to develop a model capable of ‘autopoietic’ behaviour and ‘meta – learning’. The principal investigation is the ability of such a system to identify/ map/ model/ represent flows within dynamic data sets and identify topological relationships between these flows.

As a case study, the proposed neural network model has been used to map ‘urban territory’, as part of an on going architectural research project, based in North London. The project takes the notion of ‘urban territories’ rather than ‘urban space’ as the field for interrogation, as a description of temporal spatial occupation space, rather than spatial physical permanence. Furthermore, the GNN may be used to identify the relationships between unused and vacant sites along the street. In this way, the GNN may become a means of proposing architectural interventions for these spaces, so that the territories of those that occupy it and the negotiations between them are not lost.

keywords neural networks, adaptive topology, urban planning, generative design
series other
type normal paper
email christian.derix@aedas.com
more http://www.generativeart.com/
full text file.pdf (2,637,530 bytes)
references Content-type: text/plain
last changed 2012/09/24 17:38
HOMELOGIN (you are user _anon_815308 from group guest) Works Powered by SciX Open Publishing Services 1.002