Roberto Tagliaferri | Università degli Studi di Salerno (original) (raw)
Papers by Roberto Tagliaferri
Lecture Notes in Computer Science, 2006
In this paper, we describe a novel approach to fuzzy clustering which organizes the data in clust... more In this paper, we describe a novel approach to fuzzy clustering which organizes the data in clusters on the basis of the input data and builds a'prototype'regression function as a summation of linear local regression models to guide the clustering process. This methodology is shown to be effective in the training of RBFNN's. It is shown that the performance of such networks is better than other types of networks.
In this paper the fundamentals of Bayesian learning techniques are shown, and their application t... more In this paper the fundamentals of Bayesian learning techniques are shown, and their application to neural network modeling is illustrated. Furthermore, it is shown how constraints on weight space can easily be embedded in a Bayesian framework. Finally, the application of these techniques to a complex neural network model for survival analysis is used as a significant example.
Studies in Fuzziness and Soft Computing, 2004
ABSTRACT A method to build a fuzzy neural network based on fuzzy relations with truth values in a... more ABSTRACT A method to build a fuzzy neural network based on fuzzy relations with truth values in a suitable algebraic structure is proposed. The properties of the network are analyzed in detail proving some interesting theorems on the stability of parameter variations. Furthermore the architecture of the fuzzy network is illustrated including both the fuzzification and defuzzification modules.
Lecture Notes in Computer Science, 2006
The paper introduces a generalization of the fuzzy logic connectives AND and OR. To define the lo... more The paper introduces a generalization of the fuzzy logic connectives AND and OR. To define the logical connectives different t-norms and t-conorms are used. To generalize the t-norms (t-conorms) the Ordinal Sums are introduced. To learn the parameters of the builded Ordinal Sums and the of weights of the connectives the Genetic Algorithms are applied. Two experiments using both synthetic and benchmark data are made. From one hand, a 2-dimensional classification problem to show the behavior of the approach is considered ...
Proceedings of the International Joint Conference on Neural Networks, 2003., 2003
Page 1. A hierarchical Bayesian learning scheme for autoregressive neural networks: application t... more Page 1. A hierarchical Bayesian learning scheme for autoregressive neural networks: application to the CATS benchmark Fausto Acernese't, Antonio Eleuteri *t, Leopoldo Milano*t and Roberto Tagliafemzs 'INFN Sez. Napoli, Napoli, Italia. ...
Thinking, Observing and Mining the Universe - Proceedings of the International Conference, 2004
Perspectives in Neural Computing, 1998
Proceedings of International Conference on Neural Networks (ICNN'97), 1997
2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541), 2004
Artificial Neural Networks, 1991
BMC Bioinformatics, 2015
OMICs technologies allow to assay the state of a large number of different features (e.g., mRNA e... more OMICs technologies allow to assay the state of a large number of different features (e.g., mRNA expression, miRNA expression, copy number variation, DNA methylation, etc.) from the same samples. The objective of these experiments is usually to find a reduced set of significant features, which can be used to differentiate the conditions assayed. In terms of development of novel feature selection computational methods, this task is challenging for the lack of fully annotated biological datasets to be used for benchmarking. A possible way to tackle this problem is generating appropriate synthetic datasets, whose composition and behaviour are fully controlled and known a priori. Here we propose a novel method centred on the generation of networks of interactions among different biological molecules, especially involved in regulating gene expression. Synthetic datasets are obtained from ordinary differential equations based models with known parameters. Our results show that the generated datasets are well mimicking the behaviour of real data, for popular data analysis methods are able to selectively identify existing interactions. The proposed method can be used in conjunction to real biological datasets in the assessment of data mining techniques. The main strength of this method consists in the full control on the simulated data while retaining coherence with the real biological processes. The R package MVBioDataSim is freely available to the scientific community at http://neuronelab.unisa.it/?p=1722 .
Proceedings of the International Joint Conference on Neural Networks, 2003., 2003
Angelo Ciaramella Department of Mathematics and Computer Science University of Salerno 84081 Baro... more Angelo Ciaramella Department of Mathematics and Computer Science University of Salerno 84081 Baronssi,Salerno,Italy Email: ciaram ... Go to Step 2 Step 4 -Find an independent set of 0's. ... We mark that in our approach to solve the ambiguity of the amplitude dilation we propose ...
Lecture Notes in Computer Science, 2006
In this paper, a Fuzzy Neural Network based on a fuzzy relational “IF-THEN” reasoning scheme (FRN... more In this paper, a Fuzzy Neural Network based on a fuzzy relational “IF-THEN” reasoning scheme (FRNN) is described. Different experiments on benchmark data from the UCI repository of Machine learning database are proposed for classification and approximation tasks. The model is compared with some other methods known in literature pointing out the fundamental features of the model.
Lecture Notes in Computer Science, 2009
We developed a data-mining and visualization approach to analyze the mode-of-action (MOA) of a se... more We developed a data-mining and visualization approach to analyze the mode-of-action (MOA) of a set of drugs. Starting from widegenome expression data following perturbations with different compounds in a reference data-set, our method realizes an euclidean embedding providing a map of MOAs in which drugs sharing the therapeutic application or a subset of molecular targets lies in close positions. First we build a lowdimensional, visualizable space combining a rank-aggregation method and a recent tool for the analysis of the enrichment of a set of genes in ranked lists (based on the Kolmogorov-Smirnov statistic). This space is obtained using prior knowledge about the data-set composition but with no assumptions about the similarities between different drugs. Then we assess that, despite the complexity and the variety of the experimental conditions, our aim is reached with good performance without across-condition normalization procedures.
The 2010 International Joint Conference on Neural Networks (IJCNN), 2010
... This value needs to be corrected for multiple hypothesis test. ... GO_FuzzyEnrichmentAnalysis... more ... This value needs to be corrected for multiple hypothesis test. ... GO_FuzzyEnrichmentAnalysis(C) 1. k =1 2. nUp = nDown = 0 3. totalGO = ; 4. while nUp < 2,000 and nDown < 2,000 5. compute FUP (C,k) and FDOWN (C,k) 6. compute GOUP (k) and GODOWN (k) 7. if GOUP (k ...
Multimedia Databases and Image Communication - Proceedings of the Workshop on MDIC 2004, 2004
One of the central problems in pattern recognition is that of input data probability density func... more One of the central problems in pattern recognition is that of input data probability density function estimation (pdf), i.e., the construction of a model of a probability distribution given a finite sample of data drawn from that distribution. Probabilistic Principal Surfaces (hereinafter PPS) is a nonlinear latent variable model providing a way to accomplish pdf estimation, and possesses two attractive aspects useful for a wide range of data mining applications: (1) visualization of high dimensional data and (2) their classification. PPS generates a non linear manifold passing through the data points defined in terms of a number of latent variables and of a nonlinear mapping from latent space to data space. Depending upon dimensionality of the latent space (usually at most 3−dimensional) one has 1 − D, 2 − D or 3 − D manifolds. Among the 3 − D manifolds, PPS permits to build a spherical manifold where the latent variables are uniformly arranged on a unit sphere. This particular form of the manifold provides a very effective tool to reduce the problems deriving from curse of dimensionality when data dimension increases. In this paper we concentrate on PPS used as a visualization tool proposing a number of plot options and showing its effectiveness on two complex astronomical data sets.
Lecture Notes in Computer Science, 2002
In this paper the fundamentals of Bayesian learning techniques are shown, and their application t... more In this paper the fundamentals of Bayesian learning techniques are shown, and their application to neural network modeling is illustrated. Furthermore, it is shown how constraints on weight space can easily be embedded in a Bayesian framework. Finally, the application of these techniques to a complex neural network model for survival analysis is used as a significant example.
Journal of biological regulators and homeostatic agents
Size-dependent characteristics of novel engineered nanomaterials might result in unforeseen biolo... more Size-dependent characteristics of novel engineered nanomaterials might result in unforeseen biological responses and toxicity. To address this issue, we used cDNA microarray analysis (13443 genes) coupled with bioinformatics and functional gene annotation studies to investigate the transcriptional profiles of Balb/3T3 cells exposed to a low dose (1 μM) of cobalt nanoparticles (CoNP), microparticles (CoMP) and ions (Co2+). CoNP, CoMP and Co2+ affected 124, 91 and 80 genes, respectively. Hierarchical clustering revealed two main gene clusters, one up-regulated, mainly after Co2+, the other down-regulated, mainly after CoNP and CoMP. The significant Gene Ontology (GO) terms included oxygen binding and transport and hemoglobin binding for Co2+, while the GOs of CoMP and CoNP were related to nucleus and intracellular components. Pathway analysis highlighted: i) mitochondrial dysfunction for Co2+, ii) signaling, activation of innate immunity, and apoptosis for CoNP, and iii) cell met...
Lecture Notes in Computer Science, 2010
Abstract. Clustering is one of the most important unsupervised learn-ing problems and it deals wi... more Abstract. Clustering is one of the most important unsupervised learn-ing problems and it deals with finding a structure in a collection of unla-beled data; however, different clustering algorithms applied to the same data-set produce different solutions. In many applications the ...
Lecture Notes in Computer Science, 2006
In this paper, we describe a novel approach to fuzzy clustering which organizes the data in clust... more In this paper, we describe a novel approach to fuzzy clustering which organizes the data in clusters on the basis of the input data and builds a'prototype'regression function as a summation of linear local regression models to guide the clustering process. This methodology is shown to be effective in the training of RBFNN's. It is shown that the performance of such networks is better than other types of networks.
In this paper the fundamentals of Bayesian learning techniques are shown, and their application t... more In this paper the fundamentals of Bayesian learning techniques are shown, and their application to neural network modeling is illustrated. Furthermore, it is shown how constraints on weight space can easily be embedded in a Bayesian framework. Finally, the application of these techniques to a complex neural network model for survival analysis is used as a significant example.
Studies in Fuzziness and Soft Computing, 2004
ABSTRACT A method to build a fuzzy neural network based on fuzzy relations with truth values in a... more ABSTRACT A method to build a fuzzy neural network based on fuzzy relations with truth values in a suitable algebraic structure is proposed. The properties of the network are analyzed in detail proving some interesting theorems on the stability of parameter variations. Furthermore the architecture of the fuzzy network is illustrated including both the fuzzification and defuzzification modules.
Lecture Notes in Computer Science, 2006
The paper introduces a generalization of the fuzzy logic connectives AND and OR. To define the lo... more The paper introduces a generalization of the fuzzy logic connectives AND and OR. To define the logical connectives different t-norms and t-conorms are used. To generalize the t-norms (t-conorms) the Ordinal Sums are introduced. To learn the parameters of the builded Ordinal Sums and the of weights of the connectives the Genetic Algorithms are applied. Two experiments using both synthetic and benchmark data are made. From one hand, a 2-dimensional classification problem to show the behavior of the approach is considered ...
Proceedings of the International Joint Conference on Neural Networks, 2003., 2003
Page 1. A hierarchical Bayesian learning scheme for autoregressive neural networks: application t... more Page 1. A hierarchical Bayesian learning scheme for autoregressive neural networks: application to the CATS benchmark Fausto Acernese't, Antonio Eleuteri *t, Leopoldo Milano*t and Roberto Tagliafemzs 'INFN Sez. Napoli, Napoli, Italia. ...
Thinking, Observing and Mining the Universe - Proceedings of the International Conference, 2004
Perspectives in Neural Computing, 1998
Proceedings of International Conference on Neural Networks (ICNN'97), 1997
2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541), 2004
Artificial Neural Networks, 1991
BMC Bioinformatics, 2015
OMICs technologies allow to assay the state of a large number of different features (e.g., mRNA e... more OMICs technologies allow to assay the state of a large number of different features (e.g., mRNA expression, miRNA expression, copy number variation, DNA methylation, etc.) from the same samples. The objective of these experiments is usually to find a reduced set of significant features, which can be used to differentiate the conditions assayed. In terms of development of novel feature selection computational methods, this task is challenging for the lack of fully annotated biological datasets to be used for benchmarking. A possible way to tackle this problem is generating appropriate synthetic datasets, whose composition and behaviour are fully controlled and known a priori. Here we propose a novel method centred on the generation of networks of interactions among different biological molecules, especially involved in regulating gene expression. Synthetic datasets are obtained from ordinary differential equations based models with known parameters. Our results show that the generated datasets are well mimicking the behaviour of real data, for popular data analysis methods are able to selectively identify existing interactions. The proposed method can be used in conjunction to real biological datasets in the assessment of data mining techniques. The main strength of this method consists in the full control on the simulated data while retaining coherence with the real biological processes. The R package MVBioDataSim is freely available to the scientific community at http://neuronelab.unisa.it/?p=1722 .
Proceedings of the International Joint Conference on Neural Networks, 2003., 2003
Angelo Ciaramella Department of Mathematics and Computer Science University of Salerno 84081 Baro... more Angelo Ciaramella Department of Mathematics and Computer Science University of Salerno 84081 Baronssi,Salerno,Italy Email: ciaram ... Go to Step 2 Step 4 -Find an independent set of 0&#x27;s. ... We mark that in our approach to solve the ambiguity of the amplitude dilation we propose ...
Lecture Notes in Computer Science, 2006
In this paper, a Fuzzy Neural Network based on a fuzzy relational “IF-THEN” reasoning scheme (FRN... more In this paper, a Fuzzy Neural Network based on a fuzzy relational “IF-THEN” reasoning scheme (FRNN) is described. Different experiments on benchmark data from the UCI repository of Machine learning database are proposed for classification and approximation tasks. The model is compared with some other methods known in literature pointing out the fundamental features of the model.
Lecture Notes in Computer Science, 2009
We developed a data-mining and visualization approach to analyze the mode-of-action (MOA) of a se... more We developed a data-mining and visualization approach to analyze the mode-of-action (MOA) of a set of drugs. Starting from widegenome expression data following perturbations with different compounds in a reference data-set, our method realizes an euclidean embedding providing a map of MOAs in which drugs sharing the therapeutic application or a subset of molecular targets lies in close positions. First we build a lowdimensional, visualizable space combining a rank-aggregation method and a recent tool for the analysis of the enrichment of a set of genes in ranked lists (based on the Kolmogorov-Smirnov statistic). This space is obtained using prior knowledge about the data-set composition but with no assumptions about the similarities between different drugs. Then we assess that, despite the complexity and the variety of the experimental conditions, our aim is reached with good performance without across-condition normalization procedures.
The 2010 International Joint Conference on Neural Networks (IJCNN), 2010
... This value needs to be corrected for multiple hypothesis test. ... GO_FuzzyEnrichmentAnalysis... more ... This value needs to be corrected for multiple hypothesis test. ... GO_FuzzyEnrichmentAnalysis(C) 1. k =1 2. nUp = nDown = 0 3. totalGO = ; 4. while nUp < 2,000 and nDown < 2,000 5. compute FUP (C,k) and FDOWN (C,k) 6. compute GOUP (k) and GODOWN (k) 7. if GOUP (k ...
Multimedia Databases and Image Communication - Proceedings of the Workshop on MDIC 2004, 2004
One of the central problems in pattern recognition is that of input data probability density func... more One of the central problems in pattern recognition is that of input data probability density function estimation (pdf), i.e., the construction of a model of a probability distribution given a finite sample of data drawn from that distribution. Probabilistic Principal Surfaces (hereinafter PPS) is a nonlinear latent variable model providing a way to accomplish pdf estimation, and possesses two attractive aspects useful for a wide range of data mining applications: (1) visualization of high dimensional data and (2) their classification. PPS generates a non linear manifold passing through the data points defined in terms of a number of latent variables and of a nonlinear mapping from latent space to data space. Depending upon dimensionality of the latent space (usually at most 3−dimensional) one has 1 − D, 2 − D or 3 − D manifolds. Among the 3 − D manifolds, PPS permits to build a spherical manifold where the latent variables are uniformly arranged on a unit sphere. This particular form of the manifold provides a very effective tool to reduce the problems deriving from curse of dimensionality when data dimension increases. In this paper we concentrate on PPS used as a visualization tool proposing a number of plot options and showing its effectiveness on two complex astronomical data sets.
Lecture Notes in Computer Science, 2002
In this paper the fundamentals of Bayesian learning techniques are shown, and their application t... more In this paper the fundamentals of Bayesian learning techniques are shown, and their application to neural network modeling is illustrated. Furthermore, it is shown how constraints on weight space can easily be embedded in a Bayesian framework. Finally, the application of these techniques to a complex neural network model for survival analysis is used as a significant example.
Journal of biological regulators and homeostatic agents
Size-dependent characteristics of novel engineered nanomaterials might result in unforeseen biolo... more Size-dependent characteristics of novel engineered nanomaterials might result in unforeseen biological responses and toxicity. To address this issue, we used cDNA microarray analysis (13443 genes) coupled with bioinformatics and functional gene annotation studies to investigate the transcriptional profiles of Balb/3T3 cells exposed to a low dose (1 μM) of cobalt nanoparticles (CoNP), microparticles (CoMP) and ions (Co2+). CoNP, CoMP and Co2+ affected 124, 91 and 80 genes, respectively. Hierarchical clustering revealed two main gene clusters, one up-regulated, mainly after Co2+, the other down-regulated, mainly after CoNP and CoMP. The significant Gene Ontology (GO) terms included oxygen binding and transport and hemoglobin binding for Co2+, while the GOs of CoMP and CoNP were related to nucleus and intracellular components. Pathway analysis highlighted: i) mitochondrial dysfunction for Co2+, ii) signaling, activation of innate immunity, and apoptosis for CoNP, and iii) cell met...
Lecture Notes in Computer Science, 2010
Abstract. Clustering is one of the most important unsupervised learn-ing problems and it deals wi... more Abstract. Clustering is one of the most important unsupervised learn-ing problems and it deals with finding a structure in a collection of unla-beled data; however, different clustering algorithms applied to the same data-set produce different solutions. In many applications the ...