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 ijac20076104
authors Benoudjit, Mohamed Amine; Coates, Paul S.
year 2008
title Artificial networks for spatial analysis
source International Journal of Architectural Computing vol. 6 - no. 1, pp. 59-78
summary This paper tests digital representation techniques which can be used by artificial neural networks in a computer-aided design (CAD) environment to analyze and classify architectural spaces. We developed two techniques for encoding volumetric data: vertex representation and feature space representation, as input for artificial neural networks. We tested how two different kinds of artificial neural networks, perceptron networks and self-organizing maps, could recognize given shapes in these representational formats. We have found that a one-layer perceptron can be used to classify shapes even when presented with input vectors composed of real numbers. These spatial representation techniques provide a method for using ANNs for architectural purposes.
series journal
full text file.pdf (2,037,815 bytes)
references Content-type: text/plain
last changed 2008/06/18 06:12
HOMELOGIN (you are user _anon_977485 from group guest) Works Powered by SciX Open Publishing Services 1.002