A Review Article[1]
Applications of
Artificial Neural Networks
in Industrial Engineering
Stefanus Rachmat Djajasaputra
Institute Technology Bandung
E-mail: steverd@technologist.com
Abstract
This article gives a survey of applications of Artificial Neural Networks in Industrial Engineering and its related disciplines, such as: operation research, manufacturing, and management. Artificial Neural Networks had been employed for optimation/resource allocation, pattern-recognition (classification, clustering) and prediction. The main advantages of Artificial Neural Networks are its adaptivity, fault tolerant, and noise-resistant. Some Neuralnets have non linear capability, which can be an advantage to overcome complex problems. Several researches showed that Artificial Neural Networks was better in term of accuracy compared with conventional approaches such as linear programming, discrete optimization, regression and statistics.
1. Introduction
The industrial needs which were getting more & more advances and the competitions which were getting tougher, required us to take the benefits of the most advanced technique in problem solving. Artificial Neural Networks was one of the most recent developed computational method. Artificial Neural Networks had been employed for optimation/resource allocation, pattern-recognition (classification, clustering) and prediction (L.Burke, 1992). Several researches showed that Artificial Neural Networks was better in term of accuracy compared with conventional approaches for classification, clustering and predictioan (Rao, 1994). Artificial Neural Networks is able to identify a system adaptively, thus eases a complex-system modeling, such as stock-market behaviour which was so difficuly to do with conventional methods. The main advantages of Artificial Neural Networks are its adaptivity, fault tolerant, noise-resistant and its non linearity that can be useful to overcome industrial complex problems. The other advantage of Artificial Neural Networks is its parallel processing capability, whuch can be implemented on hardware easily in several Artificial Neural Networks (e.g. Hopfield model.) This article serves as an introduction of applications of Artificial Neural Networks in Industrial Engineering and its related disciplines, while references were provided to seek further. The discussion was made as brief as possible and thus we avoid mathematics details.
2. Artificial Neural Networks
Artificial Neural Networks is a computational method which got its idea from
the neural structure of. living creatures. By mimicking the original neural
model, we hope Artificial Neural Networks will be able to reconstruct
inteligence phenomena which can be useful for classification task, system
identification, assosiative memory, etc. Artificial Neural Networks can be
described in a form of learning algorithms
and a computational model, i.e. using graph theory, as interconnected
nerve nodes, along with multiply-factors (weights) representing the interaction
levels between them. One of the most popular form of the Artificial Neural
Networks is multilayer perceptron, as
show in figure 1 (represented in graph theory.) The number of its hidden layers
depends on the complexity of the problem to be solved.
Figure 1. An Artificial Neural Networks model:
Multilayer Perceptron
From learning method point of view, Artificial
Neural Networks can be classified into
two big groups: supervised learning and unsupervised learning. A supervised
learning method needs a supervision during its training phase, thus the users
shoulds provide the output targets along with their inputs. Artificial Neural
Networks adaptively will search a weight configuration which
suitable for mapping inputs to their targets by changing the weights using a
learning algorithm. This learning process will be assumed to be finished after
the performance fulfill a specific criteria (e.g. the distance between the
outputs and the corresponding targets.) After this training process, we hope
that this Artificial Neural Networks will have a generalization capability, i.e.
be able to recognignize (map) the inputs which are not exactly the same with
the training data. On the other hand, a
unsupervised method doesn’t need a target for training. During training, this
neuralnets will map/arrange its weights so that it will be capable to cluster the input patterns. In the following paragraphs, we will have a
short discussion about some popular Artificial Neural Networks. For further
details (including comparisons between networknets models) it’s recommended to
refer to literatures by Freeman and Skapura (1991), Sanchez-Sinecio and Lau
(1992), Haykin (1994), Hetch-Nielsen (1988) and Lippman (1987.)
Backpropagation
Backpropagation
is the most popular supervised learning method which is implemented at
multilayer percepteron model. This method is popular because relatively easy to
implement, while it had proved powerful for many applications. Backpropagation
is suitable for classification/pattern recognition, prediction and control. The
weakness of this method is its slowness, therefore researchers developed faster
algorithms, such as RPPROP (Riedmiller, 1993), Cascade Correlation (S.Fallman,
1988) or Conjugate Gradient (Johansson et.al.,
1990.)
Competitive Networks
In this unsupervised learning method, the weights of
Artificial Neural Networks were represented as topological/spatial
mapping. After the inputs came, these
weights would compete and the winner is the weight closest[2]
to the input pattern.Then this winneer weight will be modified to get closer to
the input-pattern. In some Neuralnets types, the close neighboring weights also
will be modified. After the training, the weights will form groups based on the input-patterns, this
process is called clustering. Some Artificial Neural Networks in Competitive
Networks category, namely: Kohonen model, Self Organizing Map (SOM) and
Learning Vector Quantization (LVQ).
Hopfield Networks
This method is an optimization algorithm which was
developed from statistical physics approach: a learning algorithm will perform
self-organizing on Artificial Neural Networks weights to minimize the energy/
lyapunov function. Thus, this algorithm is appropriate for constraint
optimization problems on matematical programming. The general problem on matematical programming is to find the
minimum (or maximum) of a cost-function given several constraints. In Hopfield model, the cost function and
constraint were translated in form of an energy function. When this energy
function is minimum, the information on the Artificial Neural Networks weights
represents the optimal solution. Hopfield model is a recurrent networks, where the Network’s outpus were feeded to its
inputs. However, Hopfield model has some weaknesses, such as trapped on
sub-optimal solution (local-minima problem) and the limitation of its
information capacity. Several approaches were developed to improve Hopfield
model, for example Boltzman-machine, which is based on simulated annealing from
physics, to avoid local minima.
Adaptive Resonance Theory (ART)
Some people regard this unsupervised Artificial Neural Networks as the most similar neural model comparing with a real brain. The applications of ART mainly are for classification and pattern recognition. ART is better regarding its speed and accuracy compared with other Artificial Neural Networks. Another its advantage is its plastiscity, ART is able to remember new input patterns without forgeting the previous/old input patterns. Nevertheless ART is not so popular due to its high complexity, and unclueness to tune the learning parameter. ART consists of two layers: recognition layer and comparison layer. The input patterns will be saved at recognition layer (short term memory), then the patterns at this recognition layer will be compared with the patterns at comparison layer (long term memory). If a matching[3] pattern was not found, this input pattern will be classified as a new pattern. The development of ART model leads to ART-2, ART-3, Fuzzy ART, ARTMAP (supervised) and Multichannel Fuzzy ART.
The others popular Artificial Neural Networks
models, namely: Probabilistic Based Neural Networks (Specht, 1990), General
Regression Neural Networks (Specht, 1991), BAM (Kosko, 1987), Least Mean Square
(LMS), CMAC (for control application), Hidden Markov Neural Networks, Bayesian
Neural Networks, Gram-Charlier Neural Networks, Wavelet Networks, Radial Basis
Neural Networks dan Temporal Diference Neural Networks.
3. The Application of Artificial Neural Networks in Industrial Engineering
Here are several applications of Artificial Neural Networks in Industrial Engineering and its related disciplines:
Operation Research
On operation research, Artificial Neural Networks is useful as an alternative for conventional techniques (e.g. mathematical programming & statistics) for optimization & decision support. Neural Networks paradigm (e.g. reinforcement learning) can also be used to improved dynamic programming, it’s called Neuro-Dynamic-Programming (Bertsekas & Tsitsiklis, 1996). On the other hand, several operation research techniques (such as linear programming or dynamic programming) can be also implemented to improve Artificial Neural Networks learning algorithms.
One of the most important problems on operation
research is constraint optimization problems, i.e. finding optimal solution
that minimize (or maximize) a cost function under several constraints. In constraint optimization problems, the
most important problem is to model the system, formulate the cost function as
well as its constraints. Conventionally, this problem was solved using
mathematical programming. Looi (1992) did a comparison of several methods
(included Hopfield model and its derivatives) for travelling salesman problem,
a problem which was acknowledged as a benchmark standard for studying
optimization-algorithms. This study along with other study by Burke &
Ignizio (1992) showed that Artificial Neural Networks (Hopfield model) wasn’t
superior compared with conventional methods for constraint optimization
problems. Researchers tried to improve
its performance by improving formulation of energy function and learning
algorithms (C. Dagli, 1994.) Several constrained optimization problems which
can be solved Hopfield model (or its derivatives), namely:
Þ Finding the shortest path
for facilities layout (Tsuchiya, Brathikar, Takefuji, 1996.) Almost similar
approach was applied for reducing the size of
IC layout (Wang, Chen, Lai, Liu, 1992.)
Þ Automatic transport-vehicles
system: comprising the shortest-route, schedule optimization and decision of
vehiche types (Potvin, Shen, Rousseau, 1992.)
Þ Johnston & Adorf (1992)
studied the operation schedule of Hubble telescope: concerning modeling of its
scheduling system, formulation of the cost function along with its constraints,
and implementation using Guarded
–Discrete-Stochastic Neural Networks.
Þ Chiu, Maa & Shanblatt
(1990) studied applying Hopfield model for dynamic programming.
Other applications of Artificial Neural Networks for
example:
Þ Optimation of ticket-machine
maintanance using fuzzy-neural (Liu, Sin, 1997.) First, the input data (e.g. machine activity, weather) was
fuzzified before Artificial Neural Networks (e.g. a backpropagation-like)
processed this data. Finally, the output from Neuralnets was defuzzificatied
resulting the final output, e.g.the machine-maintanance priorities.
Þ Distance estimation using a
combination of backpropagation, regression neural networks and parametrict
method (Alpaydin, Altinel, Aras, 1996)
Þ Optimation of routing &
scheduling of a train-networks using backpropagation to classify service-demand
patterns (Martinelli, Teng, 1995)
Þ Optimazion of communication networks: packet scheduling(
D.B.Schwartz, 1993,) routing (M.Goudreau & L.Giles, 1992), & traffic
control (A.Hiramatsu, 1991.) Planning the celluar mobile-communication networks
to optimize the service-covering area (T.Fritsch,S.Hanshans, 1993)
More detail discussions can be found on journals
which published special issue about this topic, for example: European Journal of
Operations Research, vol.93 or Computers in Operations Research, vol.19.
Manufacturing
Þ
Monitoring , Diagnosis and Maintanance
The pattern recognition
power of Artificial Neural Networks (e.g. backpropagation, ART) fcan be used
for monitoring, vision, and diagnosis of manufacturing system. The result from
this monitoring can be used for production control, e.g. for robot-length
movement control, conveyor-belt speed control, or quality control. The
diagnosis result can be used for maintanance and repairment of equipments, as
well as for alerting system (e.g. at chemical reactor.)
Þ
Control
Artificial Neural
Networks (e.g. LMS, CMAC) can be applied also for adaptive control (Nguyen,
Widrow, 1990.) On neuro-fuzzy system,
Artificial Neural Networks was used for optimizing the parameters of a fuzzy logic system adaptively. There are
already several industrial equipments which had took benefits from
Fuzzy-Neuralnets for pattern recognition and control, namely: air conditioner,
refrigerator, and camera. More detailed discussions about the applications of
fuzzy-neural in industry can be found on publications by von Altlock (1994),
dan Yem, Langari, Zadeh (1995).
Þ
Component Desain
The proses for creating a
new component is very expensive, so we should find a method to reuse the design of existing components. According
to Gunn (1982), actually only 20% of the new components which was really needed for new designs, the rests
can be made/modified from old components. Boeing saved cost of new components production by using a multi-stages
ART Artificial Neural Networks for classification and saving component design
patterns (Smith, Escobedo, Anderson, Caudell, 1997)
Þ
Group Technology
Component grouping based
on “family” will help to make a production system more efficient (Kusiak,
1990.) Several researchers used ART & its derivatives for classification of
machine-component family. (Dagli, Huggahali, 1995; Burke, Kamal, 1992; Malave,
Ramachandran, 1994)
Þ
Feature Identification
Perceptron can be used to
determinine the most suitable component-type, based on feature identification
(Henderson, 1994.)
Þ
Proces Planning
Artificial Neural
Networks (e.g. backpropagation) was used for identifying component-types and
determining the process (e.g. method selection, machine, tools.) For planning a
complex industrial system, Artificial Neural Networks (e.g. backpropagation)
was used to model nonlinear parameters, e.g. durability of components,
chemical-process on reactor, etc.
Þ
Facility Layout
The process of
optimization a facility layout mainly was influenced by material handling, thus
the routing problem for material handling is important. Hopfield model and its
derivatives was used for finding the shortest route of this routing (Tsuchiya,
Brathikar, Takefuji, 1996; Yip, Pao, 1993)
Management:
Marketing and Finance
Þ
Market Segmentation (O’Brien, 1994, Heide, 1995)
Kohonen Artificial Neural
Networks was used for clustering, e.g. inspecting the grouping pattern of
customers based on their age, sex, salary, etc.
Þ
Customer Ranking
(Mooer, Burbach, Heeler, 1995)
Backpropagation was used
to determine the rank of targeted customers based on their business, salary,
ages, etc.
Þ
Pricing, e.g. for real estate industry (Worzala,
Lenk, Silva, 1995)
Backpropagation or radial
basis neural networks was used to model a formula for pricing based on product
attributes (e.g. total area of the house) and market-condition (e.g. interest
rate.)
Þ
Credit-card fraud detetion (Dorronsoro, Ginel,
Sanchez, Cruz, 1997)
Nonlinear discriminant
neural networks was used for fraud-detection based on input data from the
card-holder profiles ( e.g. employment, salary) and the types of the
service/stuffs that their bought.
Þ
Bankrupt Prediction (Odom, Shardo, 1990)
Backpropagation was used
for bankrupt justification, using Neuralnets inputs such as: capital, profit,
total sales, market volume, etc.
Þ
Time-series Forecasting
Some problems which
included in this category for example: stock-market forecasting (Kimoto et.al.,
1990), currency rates predition (Refenes, et.al., 1993,) and sales forecasting
(Duke, Long ,1992.) Refenes (1997) wrote a review article about this
application.
4. Ending
Artificial Neural Networks offers new computation
techniques, which can be applied for diverse problems in engineering, included
industrial engineering and operation
research Several researches showed that Artificial Neural Networks was better
in term of accuracy compared with conventional approaches such as linear
programming, discrete optimization,
regression and statistics. (Fang, Li, 1990; Chiu, Maa, Shanblatt, 1990.) On
intelligent manufacturing, Artificial Neural Networks has more advantages
compared with previous approaches (C. Dagli, 1994). Compared with expert
systems, Artificial Neural Networks is more powerful in dealing with
uncertainty inherent in industrial nature, due to fuzzy properties innate in
Artificial Neural Networks (Kosko, 1992). However, other study showed that
Artificial Neural Networks (Hopfield model) wasn’t superior compared with
conventional methods for constraint optimization problems (L.I.Burke dan J.P.Ignizio,
1992.) This mean that Artificial Neural Networks is not a universal-toolbox
which is the best to solve any problems in the world. There are some problems
which can be solved better using more-exact methods such as mathematical
programming.
In applying Artificial Neural Networks, we must
consider the model which is best matching with the problem on hand. For
example, for clustering and classification it’s better to use ART or kohonen,
however for prediction it’s better to use backpropagation. In the system
design, it’s importsnt to select the appropriate inputs, and it has to include
preprocessing stage also (i.e. normalisation, cleaning, filtering.)
Finally, as an interesting remark: many researchers now see Artificial Neural Networks only as an ordinary statistical/mathematical programming technique which was reinvented by neuro research. (Sarle, 1995.) Don’t be so fanatic/over-confidence with Artificial Neural Networks capabilities, it’s just a tool not a religion.
About the writer: Stefanus Rachmat Djajasaputra (ITB), a researcher on Artificial Neural Networks, Fuzzy logic dan Non-Linear Systems. He had published several papers on national and international events. This paper was also published at his homepage http://go.to/steverd
References
Þ C. von Altrock, 1995, Fuzzy
Logic and Neurofuzzy Applications Explained, Prentice Hall, Inc., Englewood
Cliffs
Þ J.A. Anderson, E. Rosenfeld,
1988, Neurocomputing: Foundation Of
Research, MIT, Boston
Þ E.Alpaydin, I.K.Altinel,
N.Aras, 1996, Parametric Disatance Function vs. Nonparametric Neural Networks
for Estimating Road Travel Distances, European
Journal of Operations Research, vol.92, pg.230-243
Þ D.P. Bertsekas, J.N.
Tsitsiklis, 1996,Neuro-Dynamic Programming,Athena Scientific,Boston
Þ J.P. Bigus. 1996, Data Mining with Neural Networks,
McGraw-Hill, New York
Þ L.I.Burke, S.Kamal, 1992,
Fuzzy ART for Cellular Manufacturing, Proceeding
of ANNIE’92, ASME
Þ L.I. Burke, J.P. Ignizio,
1992, Neural Networks and Operations Research: An Overview, Computers in Operations Research,
vol.19, no.3/4, pg.179-189
Þ G.A.Carpenter, S.Grossberg,
1987, A Massively Parallel Architecture for a Self-Organizing Neural Pattern
Recognition Machine, Computer Vision,
Graphics and Image Processing, vol.37, pg.54-115
Þ G.A.Carpenter, S.Grossberg,
1987,ART2:Stable Self-Organization of Pattern Recognition Codes for Analog
Input Patterns, Apllied Optics,
vol.36, pg.4919-4930
Þ G.A.Carpenter, S.Grossberg,
J.H.Reynolds, 1991, ARTMAP: Supervised Realtime Learning and Classification for
Non-stationary Data by Self-Organizing Neural Networks, Neural Networks, vol.4, pg.565-588
Þ G.A.Carpenter, S.Grossberg,
N.Markuzon, J.H.Reynolds, D.B.Rosen, 1991, Fuzzy ARTMAP: A Neural Networks
Architecture for Incremental Supervise Learning of Analog Multidimensional Map, IEEE Trans. Neural Networks vol.3, pg.
698-713
Þ G.A.Carpenter,
S.Grossberg,.B.Rosen, 1991,Fuzzy ART: Fast Stable Learning and Categorization
of Analog Pattern by an Adaprive Resonance Systems, Neural Networks, vol.4, pg.759-771
Þ G.A.Carpenter,
S.Grossberg,.1990, ART 3:Hierarchical Search Using Chemical Transmitters in
Self-Organizing Pattern Recognition Architecture, Neural Networks, vol.3, pg.129-152
Þ A. Cichocki,R. Unbehauen,
1994,Neural Networks for Optimization and Signal Processing, John Wiley, New York
Þ C. Chiu, C. Maa, M.A.
Shanblatt, 1990, An Artificial Neural Network Algorithm for Dynamic
Programming, International Journal of
Neural Systems, vol.1, no.3, pg. 211-220
Þ J.Y.Cheung, 1994,
Scheduling, Artificial Neural Networks
for Intelligent Manufacturing, Chapman & Hall, London
Þ D.F. Cook, A.D. Whittaker,
1994, Artificial Neural Networks in Manufacturing, Handbook of Design,
Manufacturing and Automation, ed. R.C. Dorf, A. Kusiak, John Wiley And
Sons, New York
Þ C. Dagli, R. Huggahalli,
1995, Int. J. Prod. Res., vol.33,
no.4, pg.893-913
Þ C.H.Dagli, R. Huggahalli,
1993, A Neural Network Approach to Group Technology, Neural Networks in Design & Manufacturing, World Scientific
Pub., Singapore
Þ J.R. Dorronsoro, F.Ginel,
C.Sanchez, C.S.Cruz, 1997, Neural Fraud Detection in Credit Card Operations, IEEE Trans. Neural Networks, vol.8,
no.4, pg.827-833
Þ C.H.Dagli, M.K.Vellanki,
1994, Automated assembly Systems, Artificial
Neural Networks for Intelligent Manufacturing, Chapman & Hall, London
Þ C.H.Dagli, 1994, Artificial Neural Networks for Intelligent
Manufacturing, Chapman & Hall, London
Þ L.S.Duke, J.A.Long, 1992,
Neural Networks Futures Trading, A Feasibility Study, Adaptive Intelligence Systems-Proceedings of BANKAI Workshop,
pg.121-132
Þ S.E.Fallman, 1988, An Empirical Study of Learning Speed in
Backpropagation Networks, Technical Report, Canergie Mellon University,
Pittsburgh
Þ S.E.Fahlman, C.Lebiere,
1990, The Cascade-Correlation Learning
Architechture, Technical Report, Canergie Mellon University, Pittsburgh
Þ L.Fang, T. Li, 1990, Design
of Competition-Based Neural Networks for Combinatorial Optimization, International Journal of Neural Systems,
vol.1, no.3, pg. 221-235
Þ T.Fritsch,S.Hanshans, 1993,
An integrated approach to cellular mobile communication plArtificial Neural
Networksing by s selft organizing feature map, at Proc. IEEE ICNN, vol
2, pg.822
Þ L. Fu, 1994, Neural Networks in Computer Intelligence,
McGraw-Hill, Inc, Singapore.
Þ J.Ghosh, Vision Based
Inspection, Artificial Neural Networks
for Intelligent Manufacturing, Chapman & Hall, London
Þ J. Gibb, 1996, Backpropagation Family Album, Technical
Report, Macquire University
Þ M.Goudreau, L.Giles, 1992,
Routing in random stage interconnection networks: comparing the exhaustive
search, greedy and neural networks approaches, at IJNN, vol.3, no.2, pg.125
Þ M.Guillot, R.Azouzi, M.Cote,
1994, Process Monitoring and Control, Artificial
Neural Networks for Intelligent Manufacturing, Chapman & Hall, London
Þ S. Haykin, 1994, Neural Networks: A Comprehensive Foundation,
Macmillan College Publishing Company-IEEE Press, New York.
Þ M.R.Henderson, 1994,
Manufacturingn Feature Indentification, Artificial
Neural Networks for Intelligent Manufacturing, Chapman & Hall, London
Þ R. Hetch-Nielsen, 1988,
Neurocomputting: picking the human brain, IEEE
Spectrum, 25(3)
Þ P.E. Hick, 1994, Industrial Engineering and Management: A New
Perspective, McGraw-Hill, New York
Þ F.S. Hillier, G.J.
Lieberman, 1995, Introduction to
Operations Research, McGraw-Hill, New York
Þ A.Hiramatsu, 1991, ATM
communication network control by neuralnetworks, at IEEE Trans. Neural
network, vol.1, pg.122
Þ J.J.Hopfield, D.W.Tank,
1985, Neural Computation for Decision in Optimization Problems, Biological Cybernetics, vol52,
pg.141-152
Þ J.J.Hopfield, 1982, Neural
Networks and Physical Systems with Emergent Collective Computational
Abillities, Proceeding of National
Academy of Science, vol.79, pg.2554-2558
Þ E.M.Johansson, F.U.Dowla,
D.M.Goodman, 1990, Backpropagation
Learning for Multi-Layer Feedforward Neural Networks Using the Conjugate
Gradient Method, Technical Report, Lawrence Livermore National Laboratory,
Berkeley
Þ M.D. Johnston, H.M. Adorf,
1992, Schedulling with neural Networks-The Case of The Hubble Space Telescope, Computers in Operations Research,
vol.19, no.3/4, pg.209-240
Þ S.V. Kamarthi, S.R.T. Kumara,
1993, Neural Networks in Conceptual Design, Neural
Networks in Design & Manufacturing, World Scientific Pub., Singapore
Þ T.Kimoto, K.Asakawa, M.Yoda,
M.Takeoka, 1990, Stock Market Prediction System with Modula Neural Network, IEEE- IJCNN, pg.11-16
Þ T.Kohonen, 1988, Self-Organization and Assosiative Memory,
Soringer-Verlag, New York
Þ B.Kosko, 1987,
Bi-directional Associative Memories, IEEE Trans. Systems, Man, and Cybernetics,
vol .18, no.1, pg.49-60
Þ B.Kosko, 1997, Fuzzy
Engineering, Prentice Hall, Inc., Englewood Cliffs
Þ B. Kosko, 1992, Neural Networks and Fuzzy Systems,
Prentice Hall International, Englewood Cliffs
Þ S.R.T.Kumara, N.S.Merchawi,
S.V.Karmarthi, M.Thazhutaveetil, 1993,
Neural Networks in Design and Manufacturing, Chapman and Hall
Þ A. Kusiak, 1990, Intelligent Manufacturing Systems,
Prentice Hall, Inc., Englewood Cliffs
Þ S.Y. Kung, 1993, Digital Neural Networks. Prentice Hall,
Inc., Englewood Cliffs
Þ J.N.K.Liu, K.Y.Sin, 1997,
Fuzzy Neural Networks for Maintenance in Mass Rapid Transit Railway System, IEEE Trans. Neural Networks, vol.8,
no.4, pg.932-941
Þ C. Looi, 1994, Neural
Network Methods in Combinatorial Optimization, Computers in Operations Research, vol.19, no.3/4, pg.191-208
Þ R.P. Lippman, An Introduction to Computing with Neural
Nets, IEEE ASSP Magazine,
halaman.4-22
Þ C.O.Malave, S.Ramachandran,
1994, Machine-part Family Formation, Artificial
Neural Networks for Intelligent Manufacturing, Chapman & Hall, London
Þ D.R. Martinelli, H.Teng,
1996, Optimization of Railway Operations using Neural Networks, Transpn. Res.C, vol.4, np.1, pg.33-49
Þ T. Masters, 1996, Advanced Algorithms For Neural Networks,
John Wiley And Sons, New York
Þ N.S. Merchawi, S.R.T.
Kumara, 1994, Neural Networks in Continuous Process Diagnostics, Artificial Neural Networks for Intelligent
Manufacturing, Chapman & Hall, London
Þ F.E. Meyers,1993, Plant
Layout and Material Handling, Prentice Hall, Inc., Englewood Cliffs
Þ K.Moore, R.Burbach,
R.Heeler, 1995, Using Neural Networks to Analyze Qualitative Data, Marketing research, vol.7, no.1,
pg.34-39
Þ M.Odom, R.Shardo, 1993, A
Neural Network Model for Bankruptcy Prediction, Decision Support Sytems, vol.9, no.6
Þ T.O’Brien, 19945, Neural
Nets for Direct Marketers, Marketing
Research, vol6, no.1, pg.21-26
Þ M.Posani, C.H.Dagli, 1994,
Process PlArtificial Neural Networksing, Artificial
Neural Networks for Intelligent Manufacturing, Chapman & Hall, London
Þ J.Potvin, Y.Shen,
J.Rousseau, 1992, Neural Networks for Automated Vehicle Dispatching, Computers in Operations Research,
vol.19, no.3/4, pg.267-276
Þ V.B. Rao, H.C.Rao, 1993, C++ Neural Networks And Fuzzy Logic, MIS
Press, New York
Þ A.N. Refenes, M.
Azema-Barac, L.Chen, S.A. Karoyussos, 1993, Currency Exchange Rate Prediction
and Neural Network Design, Neural
Computing and Application, vol.1, pg.46-58
Þ A.P.N. Refenes, A.N.
Nurgess, Y.Bentz, 1997, Neural Networks in Financial Engineering: A Study in
Methodology, IEEE Trans. Neural Networks,
vol.8, no.6, pg.1222-1267
Þ Refenes, Azema-Barac, Chen,
Karoussos, 1993, Currency Exchange Rate Prediction and Neural Network Design
Strategies, Neural Computaation and
Application, vol.1, hal. 46-58
Þ M.Riedmiller, H.Braun, 1993,
A Direct Adaptive Method For Faster Backpropagation Learning: The RPPROP
Algorithm, Proceedings Of The IEEE
International Conference On Neural Networks
Þ D.E. Rumelhart, G.E.Hinton,
R.J. Williams, 1986, “Learning Representation By Back-Propagation Errors”, Parallel Distributed Processing, Vol.1,
MIT Press, Cambridge
Þ W.Sarle, 1997, Neural Networks FAQ, Newsgroup: Comp.Ai.Neural-Nets
Þ W.Sarle, 1995, Neural
Networks And Statistical Models, Proceeding
Of SAS Conference
Þ D.B.Schwartz, 1993, ATM
scheduling with queueing delay prediction,
Proc. SIGCOM, vol 1.
Þ D.F.ShArtificial Neural
Networkso, 1990, Recent Advances in Numerical Techniques for Large Scale Optimization,
Neural Networks for Control, MIT
Press, Cambridge
Þ R Sharda, J. Wang, 1996,
Neural Networks and Operations Research/Management Science, European Journal of Operations Research,
vol.93, pg.227-229
Þ E.S. Sinncio, C.G.Y. Lau,
1992, Artificial Neural Networks:
Paradigms, Applications And Hardware Implementation, IEEE, New York
Þ S.D.G.Smith, R.Escobedo,
M.Anderson, T.P.Caudell, 1997, A Deployed Engineering Design Retrieval System
using Neural Networks, IEEE Trans. Neural
Networks, vol.8, no.4, pg.847-851
Þ D.F.Specht, 1990,
Probabilistic Neural Networks and Polynomial Adaline as Complementary
Techniques for Classification, IEEE
Trans. On Neural Networks, vol.1, no.1, pg.111-121
Þ D.F. Specht, 1991, A General
Regression Neural Networks, IEEE Transactions on Neural Networks,
vol.2, no.6, halaman 568-576
Þ Stefanus R. Djajasaputra,
1996, Diagnosis Penyakit Jantung Menggunakan Jaringan Saraf Tiruan, Proceeding Simposium Fisika Nasional XVI,
Bandung
Þ Stefanus R. Djajasaputra,
Tony, Choirul Hadi, The Houw Liong, Sri Hartati, 1998, Neural Network
and Fuzzy Logic for Classifying Remote Sensing Images, accepted in The 5th
Asian Symposium on Visualization
Þ H.A. Taha, 1987, Operation Research: An Introduction,
Macmillan Publishing co., New York
Þ W.C.Turner, J.H. Mize, K.E.
Case, 1987, Introduction to Industrial
and Systems Engineering, Prentice Hall, Inc., Englewood Cliffs
Þ E.Turban, 1995, Decision Support Systems and Expert Systems,
Prentice Hall, Inc., Englewood Cliffs
Þ L.H.Tsoukalas,
A.Ikonomopoulos, R.E.Uhrig, 1994, Fuzzy
Neural Control, Artificial Neural
Networks for Intelligent Manufacturing, Chapman & Hall, London
Þ K.Tsuchiya, S.Bharitkar,
Y.Takefuji, 1996, A Neural Network Approach to Facillity Layout Problems., European Journal of Operations Research,
vol.89, pg.557-563
Þ V. Vemuri, 1990, Artificial Neural Networks: Theoritical
Concepts, IEEE Computer Society Press, Los Alamitos
Þ L.Y. Wang, K.N.Chen, Y.T.
Lai, B.D.Liu, 1992, Neural Network on Two-Dimensional IC Layout Compaction, Electronic Letters, vol.28, no.14,
pg.1297-1298
Þ B.Widrow, S.D.Stearns, 1985,
Adaptive Signal Processing, Prentice Hall, Inc., Englewood Cliffs
Þ E.Worzala, M.Lenk, A.Silva,
1994, An exploration of Neural Networks and Its Applicaton to Real Estate
Valuation, The Journal of Real Estate
Research, vol.10, no.2, pg.185-201
Þ J. Yen, R. Langari, L.A.
Zadeh, 1995, Industrial Applications of
Fuzzy Logic and Intelligent Systems, IEEE Press, New York
Þ P.P.C.Yip, Y.Pao, 1993, A
Parallel and Distributed Processing Algorithm for Facility Layout, Neural Networks in Design &
Manufacturing, , World Scientific Pub., Singapore
Þ L.A. Zadeh, 1988, Fuzzy Logic, IEEE Computer Magazine