2004, Vol. XI, Núm. 2-3"Fuzzy systems: from modelling to knowledge extraction"http://hdl.handle.net/2099/36352020-11-26T04:35:14Z2020-11-26T04:35:14ZEditorial [Workshop "Fuzzy systems: from modelling to knowledge extraction", held at the German Conference on Artificial Intelligence, Hamburg, 2003]Klawonn, FrankKruse, RudolfMikut, RalfRunkler, Thomas A.http://hdl.handle.net/2099/36842017-02-07T16:20:02Z2007-10-17T09:36:11ZEditorial [Workshop "Fuzzy systems: from modelling to knowledge extraction", held at the German Conference on Artificial Intelligence, Hamburg, 2003]
Klawonn, Frank; Kruse, Rudolf; Mikut, Ralf; Runkler, Thomas A.
2007-10-17T09:36:11ZKlawonn, FrankKruse, RudolfMikut, RalfRunkler, Thomas A.Distributed fuzzy decision making for production schedullingRunkler, Thomas A.Sollacher, RudolfReverey, Wendelinhttp://hdl.handle.net/2099/36472017-02-07T16:20:02Z2007-10-05T10:36:51ZDistributed fuzzy decision making for production schedulling
Runkler, Thomas A.; Sollacher, Rudolf; Reverey, Wendelin
In production systems, input materials (educts) pass through multiple
sequential stages until they become a product. The production stages
consist of different machines with various dynamic characteristics. The
coupling of those machines is a non-linear distributed system. With a
distributed control system based on a multi-agent approach, the produc-
tion system can achieve (almost) maximum output, where lot size and lot
sequence are the most important control variables. In most production
processes high throughput and low stock are conflicting goals. In order to
compare and compensate between these multiple goals, a fuzzy decision
making approach is employed here that decides about the material flow
and machine states, based on variables like working load or order queue
length.
2007-10-05T10:36:51ZRunkler, Thomas A.Sollacher, RudolfReverey, WendelinIn production systems, input materials (educts) pass through multiple
sequential stages until they become a product. The production stages
consist of different machines with various dynamic characteristics. The
coupling of those machines is a non-linear distributed system. With a
distributed control system based on a multi-agent approach, the produc-
tion system can achieve (almost) maximum output, where lot size and lot
sequence are the most important control variables. In most production
processes high throughput and low stock are conflicting goals. In order to
compare and compensate between these multiple goals, a fuzzy decision
making approach is employed here that decides about the material flow
and machine states, based on variables like working load or order queue
length.A neuro-fuzzy system for sequence alignment on two levelsWeyde, TilmanDalinghaus, Klaushttp://hdl.handle.net/2099/36462017-02-07T16:20:02Z2007-10-05T10:25:13ZA neuro-fuzzy system for sequence alignment on two levels
Weyde, Tilman; Dalinghaus, Klaus
The similarity judgement of two sequences is often decomposed in similarity
judgements of the sequence events with an alignment process. However, in some
domains like speech or music, sequences have an internal structure which is important
for intelligent processing like similarity judgements. In an alignment task, this structure
can be reflected more appropriately by using two levels instead of aligning event
by event. This idea is related to the structural alignment framework by Markman and
Gentner. Our aim is to align sequences by modelling the segmenting and matching
of groups in an input sequence in relation to a target sequence, detecting variations
or errors. This is realised as an integrated process, using a neuro-fuzzy system. The
selection of segmentations and alignments is based on fuzzy rules which allow the integration
of expert knowledge via feature definitions, rule structure, and rule weights.
The rule weights can be optimised effectively with an algorithm adapted from neural
networks. Thus, the results from the optimisation process are still interpretable. The
system has been implemented and tested successfully in a sample application for the
recognition of musical rhythm patterns.
2007-10-05T10:25:13ZWeyde, TilmanDalinghaus, KlausThe similarity judgement of two sequences is often decomposed in similarity
judgements of the sequence events with an alignment process. However, in some
domains like speech or music, sequences have an internal structure which is important
for intelligent processing like similarity judgements. In an alignment task, this structure
can be reflected more appropriately by using two levels instead of aligning event
by event. This idea is related to the structural alignment framework by Markman and
Gentner. Our aim is to align sequences by modelling the segmenting and matching
of groups in an input sequence in relation to a target sequence, detecting variations
or errors. This is realised as an integrated process, using a neuro-fuzzy system. The
selection of segmentations and alignments is based on fuzzy rules which allow the integration
of expert knowledge via feature definitions, rule structure, and rule weights.
The rule weights can be optimised effectively with an algorithm adapted from neural
networks. Thus, the results from the optimisation process are still interpretable. The
system has been implemented and tested successfully in a sample application for the
recognition of musical rhythm patterns.A cost-sensitive learning algorithm for fuzzy rule-based classifiersBeck, SebastianMikut, RalfJäkel, Jenshttp://hdl.handle.net/2099/36452017-02-07T16:20:02Z2007-10-05T09:45:50ZA cost-sensitive learning algorithm for fuzzy rule-based classifiers
Beck, Sebastian; Mikut, Ralf; Jäkel, Jens
Designing classifiers may follow different goals. Which goal to prefer
among others depends on the given cost situation and the class distribution.
For example, a classifier designed for best accuracy in terms of misclassifica-
tions may fail when the cost of misclassification of one class is much higher
than that of the other. This paper presents a decision-theoretic extension
to make fuzzy rule generation cost-sensitive. Furthermore, it will be shown
how interpretability aspects and the costs of feature acquisition can be ac-
counted for during classifier design. Natural language text is used to explain
the generated fuzzy rules and their design process
2007-10-05T09:45:50ZBeck, SebastianMikut, RalfJäkel, JensDesigning classifiers may follow different goals. Which goal to prefer
among others depends on the given cost situation and the class distribution.
For example, a classifier designed for best accuracy in terms of misclassifica-
tions may fail when the cost of misclassification of one class is much higher
than that of the other. This paper presents a decision-theoretic extension
to make fuzzy rule generation cost-sensitive. Furthermore, it will be shown
how interpretability aspects and the costs of feature acquisition can be ac-
counted for during classifier design. Natural language text is used to explain
the generated fuzzy rules and their design processImproving surface detection for quality assessment of car body panelsDöring, ChristianEichhorn, AndreasGirimonte, DanielaKruse, Rudolfhttp://hdl.handle.net/2099/36442017-02-07T16:20:02Z2007-10-05T09:34:12ZImproving surface detection for quality assessment of car body panels
Döring, Christian; Eichhorn, Andreas; Girimonte, Daniela; Kruse, Rudolf
Surface quality analysis of exterior car body panels was still characterized
by manual detection of local form deviations and subjective evaluation
by experts. The approach presented in this paper is based on 3-D image
processing. A major step towards automated quality control of produced
panels is the classification of the different kinds of surface form deviations.
In previous studies we compared the performance of different soft computing
techniques for the detection of surface defect types. Although the dataset
was rather small, high dimensional and unbalanced, we achieved promising
results with regard to classification accuracies and interpretability of rule
bases. In this paper we reconsider the collection of training examples and
their assignment to defect types by the quality experts. For improving the
reliability of the defect classification we try to minimize the uncertainty of the
quality experts’ subjective and error-prone labelling. We build refined and
more accurate classification models on the basis of a preprocessed training
set that is more consistent. Improvements in classification accuracy using a
partially supervised learning strategy were achieved.
2007-10-05T09:34:12ZDöring, ChristianEichhorn, AndreasGirimonte, DanielaKruse, RudolfSurface quality analysis of exterior car body panels was still characterized
by manual detection of local form deviations and subjective evaluation
by experts. The approach presented in this paper is based on 3-D image
processing. A major step towards automated quality control of produced
panels is the classification of the different kinds of surface form deviations.
In previous studies we compared the performance of different soft computing
techniques for the detection of surface defect types. Although the dataset
was rather small, high dimensional and unbalanced, we achieved promising
results with regard to classification accuracies and interpretability of rule
bases. In this paper we reconsider the collection of training examples and
their assignment to defect types by the quality experts. For improving the
reliability of the defect classification we try to minimize the uncertainty of the
quality experts’ subjective and error-prone labelling. We build refined and
more accurate classification models on the basis of a preprocessed training
set that is more consistent. Improvements in classification accuracy using a
partially supervised learning strategy were achieved.Learning fuzzy systems: an ojective function-approachHöppner, FrankKlawonn, Frankhttp://hdl.handle.net/2099/36432017-02-07T16:20:02Z2007-10-05T09:24:09ZLearning fuzzy systems: an ojective function-approach
Höppner, Frank; Klawonn, Frank
One of the most important aspects of fuzzy systems is that they are
easily understandable and interpretable. This property, however, does not
come for free but poses some essential constraints on the parameters of a
fuzzy system (like the linguistic terms), which are sometimes overlooked when
learning fuzzy system automatically from data. In this paper, an objective
function-based approach to learn fuzzy systems is developed, taking these
constraints explicitly into account. Starting from fuzzy c-means clustering,
several modifications of the basic algorithm are proposed, affecting the shape
of the membership functions, the partition of individual variables and the
coupling of input space partitioning and local function approximation.
2007-10-05T09:24:09ZHöppner, FrankKlawonn, FrankOne of the most important aspects of fuzzy systems is that they are
easily understandable and interpretable. This property, however, does not
come for free but poses some essential constraints on the parameters of a
fuzzy system (like the linguistic terms), which are sometimes overlooked when
learning fuzzy system automatically from data. In this paper, an objective
function-based approach to learn fuzzy systems is developed, taking these
constraints explicitly into account. Starting from fuzzy c-means clustering,
several modifications of the basic algorithm are proposed, affecting the shape
of the membership functions, the partition of individual variables and the
coupling of input space partitioning and local function approximation.Fuzzy clustering: insights and a new approachKlawonn, Frankhttp://hdl.handle.net/2099/36422017-02-07T16:20:02Z2007-10-05T09:13:18ZFuzzy clustering: insights and a new approach
Klawonn, Frank
Fuzzy clustering extends crisp clustering in the sense that objects can
belong to various clusters with different membership degrees at the same
time, whereas crisp or deterministic clustering assigns each object to a unique
cluster. The standard approach to fuzzy clustering introduces the so-called
fuzzifier which controls how much clusters may overlap. In this paper we
illustrate, how this fuzzifier can help to reduce the number of undesired local
minima of the objective function that is associated with fuzzy clustering.
Apart from this advantage, the fuzzifier has also some drawbacks that are
discussed in this paper. A deeper analysis of the fuzzifier concept leads us to
a more general approach to fuzzy clustering that can overcome the problems
caused by the fuzzifier.
2007-10-05T09:13:18ZKlawonn, FrankFuzzy clustering extends crisp clustering in the sense that objects can
belong to various clusters with different membership degrees at the same
time, whereas crisp or deterministic clustering assigns each object to a unique
cluster. The standard approach to fuzzy clustering introduces the so-called
fuzzifier which controls how much clusters may overlap. In this paper we
illustrate, how this fuzzifier can help to reduce the number of undesired local
minima of the objective function that is associated with fuzzy clustering.
Apart from this advantage, the fuzzifier has also some drawbacks that are
discussed in this paper. A deeper analysis of the fuzzifier concept leads us to
a more general approach to fuzzy clustering that can overcome the problems
caused by the fuzzifier.An evolutionary approach to constraint-regularized learningHüllermeier, EykeRenners, IngoGrauel, Adolfhttp://hdl.handle.net/2099/36412017-02-07T16:20:02Z2007-10-05T09:07:12ZAn evolutionary approach to constraint-regularized learning
Hüllermeier, Eyke; Renners, Ingo; Grauel, Adolf
The success of machine learning methods for inducing models from data
crucially depends on the proper incorporation of background knowledge about
the model to be learned. The idea of constraint-regularized learning is to em-
ploy fuzzy set-based modeling techniques in order to express such knowl-
edge in a flexible way, and to formalize it in terms of fuzzy constraints.
Thus, background knowledge can be used to appropriately bias the learn-
ing process within the regularization framework of inductive inference. After
a brief review of this idea, the paper offers an operationalization of constraint-
regularized learning. The corresponding framework is based on evolutionary
methods for model optimization and employs fuzzy rule bases of the Takagi-
Sugeno type as flexible function approximators.
2007-10-05T09:07:12ZHüllermeier, EykeRenners, IngoGrauel, AdolfThe success of machine learning methods for inducing models from data
crucially depends on the proper incorporation of background knowledge about
the model to be learned. The idea of constraint-regularized learning is to em-
ploy fuzzy set-based modeling techniques in order to express such knowl-
edge in a flexible way, and to formalize it in terms of fuzzy constraints.
Thus, background knowledge can be used to appropriately bias the learn-
ing process within the regularization framework of inductive inference. After
a brief review of this idea, the paper offers an operationalization of constraint-
regularized learning. The corresponding framework is based on evolutionary
methods for model optimization and employs fuzzy rule bases of the Takagi-
Sugeno type as flexible function approximators.Knwoledge revision in Markov networksGebhardt, JörgBogerlt, ChristianKruse, RudolfDetmer, Heinzhttp://hdl.handle.net/2099/36402017-02-07T16:20:02Z2007-10-05T08:55:58ZKnwoledge revision in Markov networks
Gebhardt, Jörg; Bogerlt, Christian; Kruse, Rudolf; Detmer, Heinz
A lot of research in graphical models has been devoted to developing
correct and eficient evidence propagation methods, like join tree propagation
or bucket elimination. With these methods it is possible to condition the
represented probability distribution on given evidence, a reasoning process
that is sometimes also called focusing. In practice, however, there is the
additional need to revise the represented probability distribution in order
to reflect some knowledge changes by satisfying new frame conditions. Pure
evidence propagation methods, as implemented in the known commercial
tools for graphical models, are unsuited for this task. In this paper we develop
a consistent scheme for the important task of revising a Markov network so
that it satisfies given (conditional) marginal distributions for some of the
variables. This task is of high practical relevance as we demonstrate with
a complex application for item planning and capacity management in the
automotive industry at Volkswagen Group.
2007-10-05T08:55:58ZGebhardt, JörgBogerlt, ChristianKruse, RudolfDetmer, HeinzA lot of research in graphical models has been devoted to developing
correct and eficient evidence propagation methods, like join tree propagation
or bucket elimination. With these methods it is possible to condition the
represented probability distribution on given evidence, a reasoning process
that is sometimes also called focusing. In practice, however, there is the
additional need to revise the represented probability distribution in order
to reflect some knowledge changes by satisfying new frame conditions. Pure
evidence propagation methods, as implemented in the known commercial
tools for graphical models, are unsuited for this task. In this paper we develop
a consistent scheme for the important task of revising a Markov network so
that it satisfies given (conditional) marginal distributions for some of the
variables. This task is of high practical relevance as we demonstrate with
a complex application for item planning and capacity management in the
automotive industry at Volkswagen Group.Parametric families of fuzzy consequence operatorsElorza, JorgeBurillo López, Pedrohttp://hdl.handle.net/2099/36392017-02-07T16:20:02Z2007-10-04T12:42:27ZParametric families of fuzzy consequence operators
Elorza, Jorge; Burillo López, Pedro
In a previous paper ([6]) we explored the notion of coherent fuzzy consequence
operator. Since we did not know of any example in the literature of
non-coherent fuzzy consequence operator, we also showed several families of
such operators. It is well-known that the operator induced by a fuzzy preorder
through Zadeh’s compositional rule is always a coherent fuzzy consequence
operator. It is also known that the relation induced by a fuzzy consequence
operator is a fuzzy preorder if such operator is coherent ([5]). The aim of
this paper is to show a parametric family of non-coherent fuzzy consequence
operators which induce a preorder and also a family of non-coherent fuzzy
consequence operators which do not induce a preorder. These families of
operators can be implemented through very simple algorithms.
2007-10-04T12:42:27ZElorza, JorgeBurillo López, PedroIn a previous paper ([6]) we explored the notion of coherent fuzzy consequence
operator. Since we did not know of any example in the literature of
non-coherent fuzzy consequence operator, we also showed several families of
such operators. It is well-known that the operator induced by a fuzzy preorder
through Zadeh’s compositional rule is always a coherent fuzzy consequence
operator. It is also known that the relation induced by a fuzzy consequence
operator is a fuzzy preorder if such operator is coherent ([5]). The aim of
this paper is to show a parametric family of non-coherent fuzzy consequence
operators which induce a preorder and also a family of non-coherent fuzzy
consequence operators which do not induce a preorder. These families of
operators can be implemented through very simple algorithms.