Machine Learning and Neural Networks group

Department of Systems and Computer Science
University of Florence
Via Santa Marta 3
50139 Firenze - Italy
Tel:+39 055 4796361
Fax:+39 055 4796363



Paolo Frasconi, Giovanni Soda

Postdoctoral fellows

Fabrizio Costa, Andrea Passerini

PhD students

Marco Lippi, Marc Vincent, Matthieu Labbé

Former Members

Sauro Menchetti, Alessio Ceroni Alessandro Vullo, Simone Marinai, Marco Gori, Marco Maggini, Monica Bianchini, Michelangelo Diligenti, Massimiliano Pontil, Franco Scarselli, Edmondo Trentin

Research Projects

Protein Structure Prediction

Proteins are polypeptide chains carrying out most of the basic functions of life at the molecular level. These linear chains fold in complex 3D structures whose shape is responsible of proteins' behavior. Thanks to several genome sequencing projects, the entire DNA sequence of many organisms has been experimentally determined, and millions of proteins sequences identified. However, physical methods, such as NMR and crystallography, are not able to cope with this incredible number of structures, and most of these proteins' 3D structures and functions remain unknown. Therefore, it is becoming increasingly important to predict protein's tertiary structure ab initio from its amino acid sequence, using insights obtained from already known structures.

Adaptive processing of data structures

Connectionist and probabilistic models for machine learning are often limited to processing relatively poor data types. While some domains naturally lead to static data representations, other domains are better described by variable length representations. However, data in the real world often come with a much richer structure and serial order can be generalized to more complex relations. These relations can be naturally represented by graphs. While neural network and probabilistic architectures have been extensively studied for both static and sequential data types, architectures for data structures have received relatively little attention until recently.

Kernel Methods for Machine Learning

Machine learning for the Web

Incremental processing in Natural Language

Automated Services

FI-Pred: a Protein Secondary Structure Predictor

A Recursive Neural Networks based server for the prediction of secondary structures of proteins.

DISULFIND: Cysteines Bonding state and Connectivity Predictor

A server for the prediction of the bonding state and connectivity pattern of the cysteines in a given amino acid sequence.


SVM-Dlight Kernel

A modified version of T. Joachims' SVM-Light that allows dynamic loading of plugins for arbitrary data types and kernels

Demon Kernel

SVM-Dlight plugin to work with kernel matrices

Weighted Decompositional Kernel

C++ source code for the Weighted Decompositional Kernel

3D Decompositional Kernel

C++ source code for the 3D Decompositional Kernel

Open Positions

Ph.D positions available at Machine Learning and Neural Networks Group [link]



Related Sites on Neural Networks and Machine Learning
SSPro: RNN Based protein structure prediction
TC3 - Neural Networks and Computational Intelligence: Technical Committee of the IAPR
METAe: The Metadata Engine

Hosted sites

AI*IA Notizie (bulletin of the Italian Association for Artificial Intelligence)
IEEE Neural Networks Council Italian Regional Interest Group
SIREN (Italian Society for Neural Networks)

29th April 2003. Machine Learning and Neural Networks Group. For questions and comments: . Nedstat Basic - Free web site statistics