Reports de recerca
http://hdl.handle.net/2117/3688
Mon, 05 Dec 2016 08:40:30 GMT2016-12-05T08:40:30ZDetecció i identificació de falles en una xarxa de distribució d'aigües
http://hdl.handle.net/2117/97744
Detecció i identificació de falles en una xarxa de distribució d'aigües
Escobet Canal, Antoni; Nebot Castells, M. Àngela
This technical report deals with two of the main tasks of Fault
Monitoring
Systems (FMS): fault detection and fault identification. During fault
detection,
the FMS should recognize that the plant behavior is abnormal, and
therefore,
that the plant is not working properly. During fault identification,
the FMS
should conclude which type of failure has occurred. The main goal of
this work
is to consolidate a new fault detection technique, called enveloping,
and to
study the performance of the model acceptability measure as a tool to
enhance
and make more robust the fault identification process in the context of
Fuzzy
Inductive Reasoning Fault Monitoring System (FIRFMS). The enveloping
technique
and the model acceptability measure are applied to a water distribution
network
in order to detect specific faults and to identify them. This report is
focused
as a detailed description of the full work developed in the context of
the water
demand application. It does not pretend only to show the results
obtained by
FIRFMS to the application at hand but to present in detail all the
steps done
to reach these results.
Fri, 02 Dec 2016 19:13:45 GMThttp://hdl.handle.net/2117/977442016-12-02T19:13:45ZEscobet Canal, AntoniNebot Castells, M. ÀngelaThis technical report deals with two of the main tasks of Fault
Monitoring
Systems (FMS): fault detection and fault identification. During fault
detection,
the FMS should recognize that the plant behavior is abnormal, and
therefore,
that the plant is not working properly. During fault identification,
the FMS
should conclude which type of failure has occurred. The main goal of
this work
is to consolidate a new fault detection technique, called enveloping,
and to
study the performance of the model acceptability measure as a tool to
enhance
and make more robust the fault identification process in the context of
Fuzzy
Inductive Reasoning Fault Monitoring System (FIRFMS). The enveloping
technique
and the model acceptability measure are applied to a water distribution
network
in order to detect specific faults and to identify them. This report is
focused
as a detailed description of the full work developed in the context of
the water
demand application. It does not pretend only to show the results
obtained by
FIRFMS to the application at hand but to present in detail all the
steps done
to reach these results.Aplicación de algoritmos de clustering desarrollados en el entorno FIR a la predicción de la concentración de ozono
http://hdl.handle.net/2117/97485
Aplicación de algoritmos de clustering desarrollados en el entorno FIR a la predicción de la concentración de ozono
Gómez Miranda, Pilar; Nebot Castells, M. Àngela; Múgica Álvarez, Francisco
El presente trabajo tiene como objetivo estudiar la aplicación
de diferentes al-goritmos de clustering desarrollados en el entorno de
la metodología FIR al problema de la predicción a largo plazo de las
concentraciones de ozono en la zona centro de la ciudad de México. La
investigación realizada se centra en la identificación de modelos para
la predicción del ozono desde dos perspectivas distintas: modelado
estacional y mode-lado mensual. El modelado estacional tiene como
objetivo identificar modelos para una determinada estación del año
(período no lluvioso, en este caso). El modelado mensual tiene como
objetivo identificar modelos para cada mes del año (mes de enero, en
este caso). Los algoritmos de clustering seleccionados para este estudio
han sido dos jerárquicos, Complete Linkage y Ward Linkage y tres no
jerárquicos, Equal Frequency Partition, K-means y Fuzzy C-means.
Wed, 30 Nov 2016 09:47:17 GMThttp://hdl.handle.net/2117/974852016-11-30T09:47:17ZGómez Miranda, PilarNebot Castells, M. ÀngelaMúgica Álvarez, FranciscoEl presente trabajo tiene como objetivo estudiar la aplicación
de diferentes al-goritmos de clustering desarrollados en el entorno de
la metodología FIR al problema de la predicción a largo plazo de las
concentraciones de ozono en la zona centro de la ciudad de México. La
investigación realizada se centra en la identificación de modelos para
la predicción del ozono desde dos perspectivas distintas: modelado
estacional y mode-lado mensual. El modelado estacional tiene como
objetivo identificar modelos para una determinada estación del año
(período no lluvioso, en este caso). El modelado mensual tiene como
objetivo identificar modelos para cada mes del año (mes de enero, en
este caso). Los algoritmos de clustering seleccionados para este estudio
han sido dos jerárquicos, Complete Linkage y Ward Linkage y tres no
jerárquicos, Equal Frequency Partition, K-means y Fuzzy C-means.Automatic construction of rules fuzzy for modelling and prediction of the central nervous system
http://hdl.handle.net/2117/97441
Automatic construction of rules fuzzy for modelling and prediction of the central nervous system
Múgica Álvarez, Francisco; Nebot Castells, M. Àngela; Gómez Miranda, Pilar
The main goal of this work is to study the performance of
CARFIR (Automatic Construction of Rules in Fuzzy Inductive Reasoning)
methodology for the modelling
and prediction of the human central nervous system (CNS). The CNS
controls the hemodynamical system by generating the regulating signals
for the blood vessels and the heart. The main idea behind CARFIR is to
expand the capacity of the FIR methodology allowing it to work with
classical fuzzy rules. CARFIR is able to automatically construct fuzzy
rules starting from a set of pattern rules obtained by FIR. The new
methodology preserves as much as possible the knowledge of the pattern
rules in a compact fuzzy rule base. The prediction results obtained by
the fuzzy prediction process of CARFIR methodology are compared with
those of other inductive methodologies, i.e. FIR, NARMAX and neural
networks
Tue, 29 Nov 2016 14:46:02 GMThttp://hdl.handle.net/2117/974412016-11-29T14:46:02ZMúgica Álvarez, FranciscoNebot Castells, M. ÀngelaGómez Miranda, PilarThe main goal of this work is to study the performance of
CARFIR (Automatic Construction of Rules in Fuzzy Inductive Reasoning)
methodology for the modelling
and prediction of the human central nervous system (CNS). The CNS
controls the hemodynamical system by generating the regulating signals
for the blood vessels and the heart. The main idea behind CARFIR is to
expand the capacity of the FIR methodology allowing it to work with
classical fuzzy rules. CARFIR is able to automatically construct fuzzy
rules starting from a set of pattern rules obtained by FIR. The new
methodology preserves as much as possible the knowledge of the pattern
rules in a compact fuzzy rule base. The prediction results obtained by
the fuzzy prediction process of CARFIR methodology are compared with
those of other inductive methodologies, i.e. FIR, NARMAX and neural
networksFeature selection algorithms: a survey and experimental evaluation
http://hdl.handle.net/2117/97413
Feature selection algorithms: a survey and experimental evaluation
Molina, Luis; Belanche Muñoz, Luis Antonio; Nebot Castells, M. Àngela
In view of the substantial number of existing feature selection
algorithms, the need arises to count on criteria that
enables to adequately decide which algorithm to use in certain
situations. This work reviews several fundamental algorithms found in the
literature and assesses their performance in a controlled
scenario. A scoring measure ranks the algorithms by
taking into account the amount of relevance, irrelevance
and redundance on sample data sets. This measure computes the
degree of matching between the output given by the algorithm and the known
optimal solution. Sample size effects are also studied.
Tue, 29 Nov 2016 12:22:16 GMThttp://hdl.handle.net/2117/974132016-11-29T12:22:16ZMolina, LuisBelanche Muñoz, Luis AntonioNebot Castells, M. ÀngelaIn view of the substantial number of existing feature selection
algorithms, the need arises to count on criteria that
enables to adequately decide which algorithm to use in certain
situations. This work reviews several fundamental algorithms found in the
literature and assesses their performance in a controlled
scenario. A scoring measure ranks the algorithms by
taking into account the amount of relevance, irrelevance
and redundance on sample data sets. This measure computes the
degree of matching between the output given by the algorithm and the known
optimal solution. Sample size effects are also studied.Evolutionary optimization of heterogeneous problems
http://hdl.handle.net/2117/97407
Evolutionary optimization of heterogeneous problems
Belanche Muñoz, Luis Antonio
A large number of practical optimization problems involve elements of quite diverse nature, described as mixtures of qualitative and quantitative information, and whose description is possibly incomplete. In this work we present an extension of the breeder genetic algorithm that represents and manipulates this heterogeneous information in a natural way.
A large number of practical optimization problems involve elements of quite diverse nature, described as mixtures of qualitative and quantitative information, and whose description is possibly incomplete. In this work we present an extension of the breeder genetic algorithm that represents and manipulates this heterogeneous information in a natural way.
Tue, 29 Nov 2016 11:59:22 GMThttp://hdl.handle.net/2117/974072016-11-29T11:59:22ZBelanche Muñoz, Luis AntonioA large number of practical optimization problems involve elements of quite diverse nature, described as mixtures of qualitative and quantitative information, and whose description is possibly incomplete. In this work we present an extension of the breeder genetic algorithm that represents and manipulates this heterogeneous information in a natural way.Incremental construction of LSTM recurrent neural network
http://hdl.handle.net/2117/97400
Incremental construction of LSTM recurrent neural network
Ribeiro, Evandsa Sabrine Lopes-Lima; Alquézar Mancho, René
Long Short--Term Memory (LSTM) is a recurrent neural network that
uses structures called memory blocks to allow the net remember
significant events distant in the past input sequence in order to
solve long time lag tasks, where other RNN approaches fail.
Throughout this work we have performed experiments using LSTM
networks extended with growing abilities, which we call GLSTM.
Four methods of training growing LSTM has been compared. These
methods include cascade and fully connected hidden layers as well
as two different levels of freezing previous weights in the
cascade case. GLSTM has been applied to a forecasting problem in a biomedical domain, where the input/output behavior of five
controllers of the Central Nervous System control has to be
modelled. We have compared growing LSTM results against other
neural networks approaches, and our work applying conventional
LSTM to the task at hand.
Tue, 29 Nov 2016 11:30:41 GMThttp://hdl.handle.net/2117/974002016-11-29T11:30:41ZRibeiro, Evandsa Sabrine Lopes-LimaAlquézar Mancho, RenéLong Short--Term Memory (LSTM) is a recurrent neural network that
uses structures called memory blocks to allow the net remember
significant events distant in the past input sequence in order to
solve long time lag tasks, where other RNN approaches fail.
Throughout this work we have performed experiments using LSTM
networks extended with growing abilities, which we call GLSTM.
Four methods of training growing LSTM has been compared. These
methods include cascade and fully connected hidden layers as well
as two different levels of freezing previous weights in the
cascade case. GLSTM has been applied to a forecasting problem in a biomedical domain, where the input/output behavior of five
controllers of the Central Nervous System control has to be
modelled. We have compared growing LSTM results against other
neural networks approaches, and our work applying conventional
LSTM to the task at hand.Margin maximization with feed-forward neural networks: a comparative study with support vector machines and AdaBoost
http://hdl.handle.net/2117/97318
Margin maximization with feed-forward neural networks: a comparative study with support vector machines and AdaBoost
Romero Merino, Enrique; Màrquez Villodre, Lluís; Carreras Pérez, Xavier
Feed-forward Neural Networks (FNN) and Support Vector Machines (SVM)
are two machine learning frameworks developed from very different
starting points of view. In this work a new learning model for FNN is
proposed such that, in the linearly separable case, it tends to obtain
the same solution as SVM. The key idea of the model is a weighting of
the sum-of-squares error function, which is inspired by the AdaBoost
algorithm. As in SVM, the hardness of the margin can be controlled, so
that this model can be also used for the non-linearly separable
case. In addition, it is not restricted to the use of kernel
functions, and it allows to deal with multiclass and multilabel
problems as FNN usually do. Finally, it is independent of the
particular algorithm used to minimize the error function. Theoretic
and experimental results, on synthetic and real-world problems, are
shown to confirm these claims. Several empirical comparisons among
this new model, SVM and AdaBoost have been made in order to study the
agreement between the predictions made by the respective
classifiers. The results obtained show that similar performance does
not imply similar predictions, suggesting that different models can be
combined leading to better performance.
Mon, 28 Nov 2016 13:10:54 GMThttp://hdl.handle.net/2117/973182016-11-28T13:10:54ZRomero Merino, EnriqueMàrquez Villodre, LluísCarreras Pérez, XavierFeed-forward Neural Networks (FNN) and Support Vector Machines (SVM)
are two machine learning frameworks developed from very different
starting points of view. In this work a new learning model for FNN is
proposed such that, in the linearly separable case, it tends to obtain
the same solution as SVM. The key idea of the model is a weighting of
the sum-of-squares error function, which is inspired by the AdaBoost
algorithm. As in SVM, the hardness of the margin can be controlled, so
that this model can be also used for the non-linearly separable
case. In addition, it is not restricted to the use of kernel
functions, and it allows to deal with multiclass and multilabel
problems as FNN usually do. Finally, it is independent of the
particular algorithm used to minimize the error function. Theoretic
and experimental results, on synthetic and real-world problems, are
shown to confirm these claims. Several empirical comparisons among
this new model, SVM and AdaBoost have been made in order to study the
agreement between the predictions made by the respective
classifiers. The results obtained show that similar performance does
not imply similar predictions, suggesting that different models can be
combined leading to better performance.Exploiting diversity of margin-based classifiers
http://hdl.handle.net/2117/96843
Exploiting diversity of margin-based classifiers
Romero Merino, Enrique; Carreras Pérez, Xavier; Màrquez Villodre, Lluís
An experimental comparison among Support Vector Machines, AdaBoost and a recently proposed model for maximizing the margin with
Feed-forward Neural Networks has been made on a real-world classification problem, namely Text Categorization. The results obtained
when comparing their agreement on the predictions show that similar performance does not imply similar predictions, suggesting that
different models can be combined to obtain better performance. As a consequence of the study, we derived a very simple confidence
measure of the prediction of the tested margin-based classifiers. This measure is based on the margin curve. The combination of
margin-based classifiers with this confidence measure lead to a marked improvement on the performance of the system, when combined with several well-known combination schemes.
Fri, 18 Nov 2016 14:53:16 GMThttp://hdl.handle.net/2117/968432016-11-18T14:53:16ZRomero Merino, EnriqueCarreras Pérez, XavierMàrquez Villodre, LluísAn experimental comparison among Support Vector Machines, AdaBoost and a recently proposed model for maximizing the margin with
Feed-forward Neural Networks has been made on a real-world classification problem, namely Text Categorization. The results obtained
when comparing their agreement on the predictions show that similar performance does not imply similar predictions, suggesting that
different models can be combined to obtain better performance. As a consequence of the study, we derived a very simple confidence
measure of the prediction of the tested margin-based classifiers. This measure is based on the margin curve. The combination of
margin-based classifiers with this confidence measure lead to a marked improvement on the performance of the system, when combined with several well-known combination schemes.Feature selection for microarray gene expression data using simulated annealing guided by the multivariate joint entropy
http://hdl.handle.net/2117/96707
Feature selection for microarray gene expression data using simulated annealing guided by the multivariate joint entropy
González Navarro, Félix Fernando; Belanche Muñoz, Luis Antonio
In this work a new way to calculate the multivariate joint entropy is presented. This measure is the basis for a fast information-theoretic based evaluation of gene relevance in a Microarray Gene Expression data context. Its low complexity is based on the reuse of previous computations to calculate current feature relevance. The mu-TAFS algorithm --named as such to differentiate it from previous TAFS algorithms-- implements a simulated annealing technique specially designed for feature subset selection. The algorithm is applied to the maximization of gene subset relevance in several public-domain microarray data sets. The experimental results show a notoriously high classification performance and low size subsets formed by biologically meaningful genes.
Wed, 16 Nov 2016 09:35:17 GMThttp://hdl.handle.net/2117/967072016-11-16T09:35:17ZGonzález Navarro, Félix FernandoBelanche Muñoz, Luis AntonioIn this work a new way to calculate the multivariate joint entropy is presented. This measure is the basis for a fast information-theoretic based evaluation of gene relevance in a Microarray Gene Expression data context. Its low complexity is based on the reuse of previous computations to calculate current feature relevance. The mu-TAFS algorithm --named as such to differentiate it from previous TAFS algorithms-- implements a simulated annealing technique specially designed for feature subset selection. The algorithm is applied to the maximization of gene subset relevance in several public-domain microarray data sets. The experimental results show a notoriously high classification performance and low size subsets formed by biologically meaningful genes.Modeling the thermal behavior of biosphere 2 in a non-controlled environment using bond graphs
http://hdl.handle.net/2117/96439
Modeling the thermal behavior of biosphere 2 in a non-controlled environment using bond graphs
Nebot Castells, M. Àngela; Cellier, François E.; Múgica Álvarez, Francisco
Biosphere 2 is a closed ecological system of high complexity built to
deepen the understanding of ecological systems, to study the dynamics
of closed ecologies,
and to learn to control their behavior. The use of modeling and
simulation is crucial in the achievement of these goals. Understanding a
physical system is almost synonymous with possessing a model of its
comportment.
The main goal of this study is the development of a dynamic bond graph
model that represents the thermal behavior of the complex ecological system
under study, Biosphere 2. In this work, a first model that captures the
behavior of the ecological system in a non-controlled environment is
presented.
Wed, 09 Nov 2016 16:28:14 GMThttp://hdl.handle.net/2117/964392016-11-09T16:28:14ZNebot Castells, M. ÀngelaCellier, François E.Múgica Álvarez, FranciscoBiosphere 2 is a closed ecological system of high complexity built to
deepen the understanding of ecological systems, to study the dynamics
of closed ecologies,
and to learn to control their behavior. The use of modeling and
simulation is crucial in the achievement of these goals. Understanding a
physical system is almost synonymous with possessing a model of its
comportment.
The main goal of this study is the development of a dynamic bond graph
model that represents the thermal behavior of the complex ecological system
under study, Biosphere 2. In this work, a first model that captures the
behavior of the ecological system in a non-controlled environment is
presented.