Teaching
Tesi e progetti disponibili
Tipo | Laurea Specialistica |
Argomento | Multiobjective Genetic Fuzzy Systems for mining remotely sensed hyperspectral images |
Parole chiave | Multiobjective Genetic Fuzzy Systems, hyperspectral images, Remote Sensing, Regression/Classification |
Descrizione | Multiobjective Genetic Fuzzy systems (e.g., fuzzy rule-based systems trained by a multiobjective evolutionary algorithm) have been estensively applied to regression and classification problems. However, dealing with high dimensional data sets still remains a challenging problem. Mining remotely sensed hyperspectral images is one of such high dimensional, very interesting, problem. In this thesis we are interested in study new ways to apply to the extraction of information from hyperspectral images. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Multiobjective Genetic Fuzzy Systems for Mining High Dimentional Data |
Parole chiave | Multiobjective Genetic Fuzzy Systems, Regression/Classification problems, High dimentional data |
Descrizione | Scope of the thesis is to develop new technique to speed up the fitness evaluation of fuzzy rule-based systems identified through multiobjective evolutionary algorithms |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Complexity reduction of fuzzy systems trough multi-valued logic minimization tools |
Parole chiave | Complexity Reduction, Fuzzy Rule-Based Systems, Multi-Valued Logic Minimization |
Descrizione | Logic minimization of boolean circuits has been studied for decades. The minimization of multiple-valued logic has also been deeply studied and many tools have been developed. Only recently, however, it has been recognized a parallelism between multi-valued logic minimization and fuzzy systems complexity reduction. The scope of this thesis is to implement and compare several simplification techniques, including binary and multi-valued logic minimization approaches. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Implementation of a fuzzy rule-based classifier for PRTools |
Parole chiave | Fuzzy Rule-Based Classifiers, PRTools |
Descrizione | PRTools is a third-part Matlab toolbox which provide a rich set of functions for Pattern Recognition. Currently PRTools has no built-in Fuzzy Rule-Based Classifiers (FRBCs). Scope of the thesis is to develop a PRTools extention which implements several types of FRBCs. |
Requisiti | Good Matlab programming skills. |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea |
Argomento | Input pruning in Fuzzy Rule-Based Systems based on Orthogonal Transformations |
Parole chiave | Fuzzy Rule-Based Systems, Input pruning, Orthogonal Transformation, t-norms |
Descrizione | When generating Fuzzy Rule-Based Systems (FRBSs) from data either using clustering or Wang and Mendel methods, all the inputs are involved. This can be a problem in high dimensional spaces, where the FRBS can became too complex and rules too specific. Input pruning (i.e., the removal of some inputs) can be very important in such applications. Scope of the thesis is the identification of a t-norm which allows input pruning through orthogonal transformations like Singular Value Decomposition (SVD). The basic idea here is that inputs associated with lower singular values can be pruned first than the ones associated with higher singular values. Input Reduction using orthogonal transformations Parole chiave Feature Selection, Pruning, SVD Descrizione Orthogonal Transformations like SVD (Singular Value Decomposition) have been used in the past to prune rules from a Fuzzy Rule-Based System. In this thesis we want try to exploit this idea to prune less important inputs. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Development of a novel family of Multiobjective Evolutionary Algorithms which exploits the convex hull to rank solutions |
Parole chiave | Fuzzy Rule-Based Classifiers, Receiver Operating Characteristic, Multiobjective Evolutionary Algorithms, Convex Hull |
Descrizione | Typically, the performance of Fuzzy Rule-Based Classifiers (FRBCs) are assessed in the Receiver Operating Characteristic (ROC) space. The ROC space can be regarded as a two-objective space (true positives vs false positives). The use of Multiobjective Evolutionary Algorithm (MEAs) for the optimization of FRBCs is quite natural in this context. Scope of the thesis is to implement new MEAs based on the convex hull of solutions in the objective space and the comparison with well-established MEAs. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Exploiting the use of Convex Hull in Non-linear Goal Programming tackled with evolutionary multiobjective algorithms |
Parole chiave | Multiobjective Evolutionary Algorithms, Non-linear Goal Programming, Convex Hull |
Descrizione | Non-Linear Goal Programming (NLGP) consists of finding the nearest solution to a given target solution in a non-linear scenario. Multiobjective evolutionary algorithm have proved to be very effective in NLGP. Scope of the thesis is to investigate the advantages/disadvantages on using the concept of convex hull solutions within this scenario. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Parallel Multiobjective Evolutionary Optimization |
Parole chiave | Performance Comparison of Parallel Multiobjective Evolutionary Algorithms |
Descrizione | Multiobjective Evolutionary Algorithms (MOEAs) have proved to be powerful tools for solving multiobjective optimization problem. However, real optimization problems often require a huge amount of calculus. In such cases, resorting to parallel version can be a solution. Parallel MOEAs can be implemented either as a classical parallel evolutionary algorithm or exploiting the multiobjective aspect of this particular case. Scope of the thesis is to compare exiting Parallel MOEAs among those which exploit the multiobjective nature of the problem. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea Specialistica |
Argomento | Evolving Fuzzy Rule-Based Classifiers |
Parole chiave | Evolving Systems, Fuzzy Rule-Based Classifiers |
Descrizione | Evolving Fuzzy Systems typically are fuzzy rule-based systems with a time-varying rule-base. Their use is particularly promising in online data mining process where new data are collected over the time. Scope of the thesis is the implementation of a new evolving fuzzy rule-based classifier, after reviewing the state-of-the art of evolving fuzzy systems. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |
Tipo | Laurea |
Argomento | Robust and Accurate Regression using TS models and Multiobjective Evolutionary Algorithms |
Parole chiave | Robust Regression, Takagi-Sugeno Fuzzy Models, Multi-Objective Evolutionary Algoritms |
Descrizione | Regression deals with the approximation of a function known by input-out data points. Typically, a parametric model is employed to solve this kind of problems. Fuzzy Rule-Based Systems of the Takagi-Sugeno type are very effective parametric models known to be universal approximators. However, obtaining both an accurate and a robust to noise model is often required. Scope of the thesis is to optimize both the accuracy and the robustness of the model(s) to noise, using multiobjective evolutionary algorithms, after a study of the state-of-the-art of the available metrics used to assess the robustness of a model. |
Requisiti | |
Informazioni | B. Lazzerini - F. Marcelloni |