Dr. Giovanna Castellano is Assistant Professor at the Computer Science Department of the University of Bari (Italy). She received the laurea (Ms.Sci degree) in Computer Science in 1993 at the University of Bari and the Ph.D. in 2001 at the same University, defending a thesis entitled "A Neurofuzzy Methodology for Predictive Modeling".
The reseach activity of Giovanna Castellano falls in the area of Computational Intelligence. Her research interests include Neural Networks, Fuzzy systems, Neuro-Fuzzy systems, Granular Computing, Meta-learning systems, Computational Web Intelligence, Fuzzy Image Processing.
Giovanna Castellano is author of about 160 international, peer-reviewed scientific publications; she is reviewer for high impact-factor journals, she serves as member of the editorial board for several scientific journals and she attended to several international conferences, where she has been speaker or session chair. She is Associate Editor of the Elsevier journal Information Sciences.
Giovanna Castellano has been involved in several national-wide research projects. She is teacher of "Computer Architectures" in academic undergraduate courses.
Abstract: Emotion recognition has become a fundamental task in human-computer interaction systems. In this paper we propose an emotion recognition approach based on biologically inspired methods. Specifically, emotion classification is performed using a Long Short-Term Memory (LSTM) recurrent neural network which is able to recognize long-range dependencies between successive temporal patterns. We propose to represent data using features derived from two different models: Mel-Frequency Cepstral Coefficients (MFCC) and the Lyon cochlear model. In the experimental phase, results obtained from the LSTM network and the two different feature sets are compared showing that features derived from the Lyon cochlear model give better recognition results in comparison with those obtained with the traditional MFCC representation.
Abstract: Recommender systems are systems capable of assisting users by quickly providing them with relevant resources according to their interests or preferences. The efficacy of a recommender system is strictly connected with the possibility of creating meaningful user profiles, including information about user preferences, interests, goals, usage data and interactive behavior. In particular, analysis of user preferences is important to predict user behaviors and make appropriate recommendations. In this paper, we present a fuzzy framework to represent, learn and update user profiles. The representation of a user profile is based on a structured model of user cognitive states, including a competence profile, a preference profile and an acquaintance profile. The strategy for deriving and updating profiles is to record the sequence of accessed resources by each user, and to update preference profiles accordingly, so as to suggest similar resources at next user accesses. The adaption of the preference profile is performed continuously, but in earlier stages it is more sensitive to updates (plastic phase) while in later stages it is less sensitive (stable phase) to allow resource recommendation. Simulation results are reported to show the effectiveness of the proposed approach.
Abstract: The problem of determining the proper size of an artificial neural network is recognized to be crucial, especially
for its practical implications in such important issues as learning and generalization. One popular approach tackling this problem is commonly known as pruning and consists of training a larger than necessary network and then removing unnecessary weights/nodes. In this paper, a new pruning method is developed, based on the idea of iteratively eliminating units and adjusting the remaining weights in such a way that the network performance does not worsen over the entire training set. The pruning problem is formulated in terms of solving a system of linear equations, and a very efficient conjugate gradient algorithm is used for solving it, in the least-squares sense. The algorithm also provides a simple criterion for choosing the units to be removed, which has proved to work well in practice. The results obtained over various test problems demonstrate the effectiveness of the proposed approach.
Abstract: In this paper, we present a shape labeling approach for automatic image annotation. A fuzzy clustering process is applied to shapes represented by Fourier descriptors in order to derive a set of shape prototypes. Then, prototypes are manually annotated by textual labels corresponding to semantic categories. Based on the labeled prototypes, a new shape is automatically labeled by associating a fuzzy set that provides membership degrees of the shape to all semantic classes. Preliminary results show the suitability of the proposed approach to image annotation by encouraging its application in wider application contexts.
Notes: Proceedings of the 9th International Workshop on Fuzzy Logic and Applications (WILF2011)
Abstract: This work presents an approach based on image texture analysis to obtain a description of oocyte cytoplasm which could aid the clinicians in the selection of oocytes to be used in the assisted insemination
process. More specifically, we address the problem of providing a description of the oocyte cytoplasm in terms of regular patterns of granularity which are related to oocyte quality. To this aim, we perform a texture analysis on the cytoplasm region and apply a spatial fuzzy clustering to segment the cytoplasm into different granular regions. Preliminary experimental results on a collection of light microscope images of oocytes are reported to show the effectiveness of the proposed approach.
Notes: Proceedings of the 9th International Workshop on Fuzzy Logic and Applications (WILF 2011)
Abstract: DC* (Double Clustering by A*) is an algorithm for interpretable fuzzy information granulation of data. It is mainly based on two clustering steps. The first step applies the LVQ1 algorithm to find a suitable representation of data relationships. The second clustering step is based on the A* search strategy and is aimed at finding an optimal number of fuzzy granules that can be labeled with linguistic terms. As a result, DC* is able to linguistically describe hidden relationships among available data. In this paper we propose an extension of the DC* algorithm, called DC , which improves the generalization ability of the original DC* by modifying the A* search procedure. This variation, inspired by Support Vector Machines, results empirically effective as reported in experimental results.
Abstract: In this work, a dermoscopic image analysis technique is proposed. A novel approach, based on the detection of gray areas using image analysis techniques is explored. To this aim, a statistical histogram analysis is carried out using the HSB color space to derive the relationship between the skewness and the mean of the brightness color plane histogram. The derived framework is used for adaptive thresholding of gray area regions within a skin lesion image.
Abstract: Fuzzy mathematical morphology has been widely applied to process real-world images characterized by vagueness and imprecision. The aim of this work is to evaluate the
effect of applying fuzzy morphological operators to detect soft edges in biological images of oocytes so as to make more easy their segmentation. The main concepts from fuzzy mathematical morphology are briefly introduced and the results of applying fuzzy morphological operators are reported in low-contrast images of human oocytes.
Abstract: Exclusion/inclusion hyperbox classification has demonstrated significant advantages in terms of its ability to cover topologically complex data structures with a relatively few hyperboxes thus resulting in the superior interpretability of classification results. However, the size of exclusion hyperboxes may occasionally become prohibitive if the data classes are grouped in a particularly unfavorable way in the pattern space. In this study we consider adaptation of the maximum size of hyperboxes in response to the ratio of the exclusion to inclusion hyperboxes. Two alternative adaptation strategies are being considered: (i) the adaptation of the size of all hyperboxes and (ii) the adaptation of the size of hyperboxes that fall within the previously identified exclusion area. The tradeoff between the number and the complexity of the classification rules implied by the two strategies is assessed on a set of sample classification problems.