Professional Documents
Culture Documents
Downloaded 20 Sep 2009 to 128.227.79.245. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
Table 1. Five most frequently cited articles in the Journal of Computing for a classification of the alternative modes of neuron operation兲.
in Civil Engineering 共Thomson Corporation 2006兲 To an extent, inspiration for the design of these higher-order
Number network structures can be gained by studying neural subsystems
Title of citations in the central nervous system. For example, the self-organizing
Neural networks in civil engineering, Parts I and II 131 subsystems that solve certain early processing tasks in the visual
共Flood and Kartam 1994兲 system 共such as, line detection兲 are understood well enough
Neural networks for river flow prediction 97 to be replicated using artificial neural networks 共Sirosh and
共Karunanithi et al. 1994兲 Miikkulainen 1997兲. However, the organization of structures that
Genetic algorithms in pipeline optimization 74 perform higher-order operations than this, later in the visual sys-
共Goldberg and Kuo 1987兲 tem 共such as face recognition兲 are not well understood. Moreover,
Genetic algorithms in discrete optimization of steel 37
most engineering problems do not have analogs in nature with
truss roofs 共Koumousis and Georgiou 1994兲
readily available solutions provided by biological neural systems.
Damage detection in structures based on 35
feature-sensitive neural networks 共Szewczyk and Given this, and the fact that suitable higher-order structures can-
Hajela 1994兲 not normally be derived by design, it seems that we must turn to
techniques such as genetic algorithms for their development. In-
terestingly, of the top five most frequently cited articles in the
networks of greater complexity. More precisely, there is an appar- Journal of Computing in Civil Engineering 共Table 1兲, three were
ent correlation between the cognitive skills of a species and its on neural networks and two were concerned with genetic
encephalization quotient 共EQ兲 共the ratio of actual brain size to algorithms.
expected brain size needed for the basic control and monitoring of While genetic algorithms and related optimization techniques
the body兲 共Jerison 2000兲. It also seems that an increase in the have been used to develop artificial neural network applications
number of neurons alone is not sufficient to provide higher levels for many years 共over 15 years in the case of civil engineering兲,
of cognitive skills, but that these must also be formed into a
this has been restricted to developing single network units rather
structured system of higher-order functional units such as is found
than the higher-order structures proposed here. Such an approach
in the visual system 共Sirosh and Miikkulainen 1997兲. The sticking
point in the development of applied artificial neural networks has will require a sophisticated genetic coding system and corre-
not been an inability to construct systems comprising many mil- sponding set of objective functions to facilitate development of
lions of neurons 共this is perfectly feasible with today’s technol- appropriate neural systems at the macro-level 共the connectivity
ogy兲; rather, it has been, at least in part, a lack of knowledge between the higher level neural units兲, the meso-level 共the
about how to organize neurons into appropriate higher-order net- connectivity between the neurons within a given unit兲, and the
work structures. Note that the term “higher-order structures” is micro-level 共the mode of operation of the individual neurons兲. In
used in this editorial to refer to the connection structure of a addition, the method of development might be designed to evolve
network and should not be confused with the term “higher-order automatic learning responses in a network when subjected to
neural networks,” which is commonly used in the literature to input data, including learning of connection weights and self-
refer to neural networks composed of Sigma-Pi neurons. organization of the connection structure. In the longer term, other
The contention of this editorial is that, there is much that can processing mechanisms operating at levels below and above that
be done to advance the scope of application and utility of artificial of the neuron may be considered, particularly as we come to
neural networks by shifting research towards the development of
understand these processes from biological studies.
higher-order structures. Three basic forms of connection structur-
In summary, over the last decade, artificial neural networks
ing can be identified, namely, concurrent, serial, and hierarchic
organizations of higher-order units 共the latter being an integration have been a primary focus of interest for research in computer-
of the output of several units to provide a summary or abstraction based civil engineering. As far as they have been applied, they
of their results兲. The boundaries of these higher-order units may have provided convenient and often highly accurate solutions to
be very distinct with relatively few connections between the con- problems from all branches of civil engineering. However, the
stituent neurons and those in other units, or they may be indistinct extent of this application has rarely ventured beyond rudimentary
with many cross-connections. Moreover, any unit may feedback problems such as simple function modeling and pattern classifi-
recursively to other units in the structure including itself. For cation. In order to achieve the full potential of this technology, as
versatility in application, the structuring of a network will often promised by biological neural systems, researchers must take on
have to allow for a variable format in the configuration of input the challenge of developing networks that are vastly more com-
data—this is to allow the network to function in an environment plex than have been developed to date. Complexity, in this sense,
where there is spatial variance in the definition of a problem requires not just an increase in the number of neurons, but also a
共resulting from scaling, rotation, translation and/or shearing of the rich higher-order structuring of the network, such as is found in
data format, for example兲, temporal variance 共resulting from the
biological neural systems. Promising approaches to the develop-
use of arbitrary starting points in input data streams, changes in
ment of these structures are genetic algorithms and related meth-
the rate of data flow, and/or gradual shifts in the nature of a
problem over time兲 and stochastic variance 共resulting from noise ods. While these will require the design of sophisticated genetic
and/or missing data values兲. Complementary to the structure of a coding mechanisms, the potential payoff is considerable in terms
network is the mode of operation of the neurons and their of broadening the scope of application of computing to civil en-
connections. Here, no restrictions are proposed: studies might gineering. We are, of course, a very long way from being able to
consider anything from neurons that act as simple logic gates replicate or even go beyond human cognitive skills using artificial
共effectively making the network operate as a digital circuit兲 neural networks, but it is time at least to take the next tentative
through to pulse frequency coded units 共see Kartam et al. 1997兲 step towards this ambition.
Downloaded 20 Sep 2009 to 128.227.79.245. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
References “Neural networks for river flow prediction.” J. Comput. Civ. Eng.,
8共2兲, 201–220.
ASCE. 共2006兲. Research Library, Reston, Va., 具http://ascelibrary.aip.org/典. Koumousis, V. K., and Georgiou, P. G. 共1994兲. “Genetic algorithms in
Bullock, T. H., Bennett, M. V. L., Johnston, D., Josephson, R., Marder, discrete optimization of steel truss roofs.” J. Comput. Civ. Eng., 8共3兲,
E., Fields, R. D. 共2005兲. “The neuron doctrine, redux.” Science, 310, 309–325.
791–793. Rosenblatt, F. 共1958兲. “The perceptron: A probabilistic model for infor-
Flood, I., and Kartam, N. 共1994兲. “Neural networks in civil engineering.
mation storage and organization in the brain.” Psychol. Rev., 65共6兲,
I: Principles and understanding; II: Systems and application,” J. Com-
put. Civ. Eng., 8共2兲, 131–162. 386–408.
Flood, I., and Kartam, N. 共1997兲. “Systems.” Artificial neural networks Sirosh, J., and Miikkulainen, R. 共1997兲. “Topographic receptive fields and
for civil engineers: Fundamentals and applications, N. Kartam, I. patterned lateral interaction in a self-organizing model of the primary
Flood, and J. H. Garrett, Jr., eds., ASCE, Reston, Va., 19–43. visual cortex.” Neural Comput., 9, 577–594.
Goldberg, D. E., and Kuo, C. H. 共1987兲. “Genetic algorithms in pipeline Szewczyk, Z. P., and Hajela, P. 共1994兲. “Damage detection in structures
optimization.” J. Comput. Civ. Eng., 1共2兲, 128–141. based on feature-sensitive neural networks.” J. Comput. Civ. Eng.,
Jerison, H. J. 共2000兲. “The evolution of intelligence.” Handbook of intel- 8共2兲, 163–178.
ligence, R. H. Sternberg, ed., Cambridge University Press, New York. Thomson Corporation. 共2006兲. “ISI web of knowledge, web of science.”
Karunanithi, N., Grenney, W. J., Whitley, D., and Bovee, K. 共1994兲. 具http://portal.isiknowledge.com/典.
Downloaded 20 Sep 2009 to 128.227.79.245. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright