Arrow Research search

Author name cluster

Lorenza Saitta

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

17 papers
2 author rows

Possible papers

17

IJCAI Conference 2007 Conference Paper

  • Ugo Galassi
  • Attilio Giordana
  • Lorenza Saitta

This paper presents an algorithm for inferring a Structured Hidden Markov Model (S-HMM) from a set of sequences. The S-HMMs are a sub-class of the Hierarchical Hidden Markov Models and are well suited to problems of process/user profiling. The learning algorithm is unsupervised, and follows a mixed bottom-up/top-down strategy, in which elementary facts in the sequences (motifs) are progressively grouped, thus building up the abstraction hierarchy of a S-HMM, layer after layer. The algorithm is validated on a suite of artificial datasets, where the challenge for the learning algorithm is to reconstruct the model that generated the data. Then, an application to a real problem of molecular biology is briefly described.

PDF Details

IJCAI Conference 2005 Conference Paper

Learning Complex Event Descriptions by Abstraction

  • Ugo Galassi
  • Attilio Giordana
  • Lorenza Saitta
  • Marco

The presence of long gaps dramatically increases the difficulty of detecting and characterizing complex events hidden in long sequences. In order to cope with this problem, a learning algorithm based on an abstraction mechanism is proposed: it can infer a Hierarchical Hidden Markov Model, from a learning set of sequences. The induction algorithm proceeds bottom-up, progressively coarsening the sequence granularity, and letting correlations between subsequences, separated by long gaps, naturally emerge. As a case study, the method is evaluated on an application of user profiling. The results show that the proposed algorithm is suitable for developing real applications in network security and monitoring.

PDF

IJCAI Conference 2003 Conference Paper

Monte Carlo Theory as an Explanation of Bagging and Boosting

  • Roberto Esposito
  • Lorenza Saitta

In this paper we propose the framework of Monte Carlo algorithms as a useful one to analyze ensemble learning. In particular, this framework allows one to guess when bagging will be useful, explains why increasing the margin improves performances, and suggests a new way of performing ensemble learning and error estimation.

PDF

JMLR Journal 2003 Journal Article

Relational Learning as Search in a Critical Region

  • Marco Botta
  • Attilio Giordana
  • Lorenza Saitta
  • Michèle Sebag

Machine learning strongly relies on the covering test to assess whether a candidate hypothesis covers training examples. The present paper investigates learning relational concepts from examples, termed relational learning or inductive logic programming. In particular, it investigates the chances of success and the computational cost of relational learning, which appears to be severely affected by the presence of a phase transition in the covering test. To this aim, three up-to-date relational learners have been applied to a wide range of artificial, fully relational learning problems. A first experimental observation is that the phase transition behaves as an attractor for relational learning; no matter which region the learning problem belongs to, all three learners produce hypotheses lying within or close to the phase transition region. Second, a failure region appears. All three learners fail to learn any accurate hypothesis in this region. Quite surprisingly, the probability of failure does not systematically increase with the size of the underlying target concept: under some circumstances, longer concepts may be easier to accurately approximate than shorter ones. Some interpretations for these findings are proposed and discussed. [abs] [ pdf ][ ps.gz ][ ps ]

PDF Details

IJCAI Conference 1999 Conference Paper

An Experimental Study of Phase Transitions in Matching

  • Attilio Giordano
  • Marco Bona
  • Lorenza Saitta

Finding models of a predicate logic formula is a well-known hard problem, whose complexity is exponential in the number of variables. However, even though this number is kept constant, substantial differences in complexity arise when searching for solutions in different problem instances. Such a behavior appears to be quite general, according to recent results reported in the literature; in fact, several classes of hard problems exhibit a narrow phase transition with respect to some order parameter, in correspondence of which the complexity dramatically rises up, still remaining tractable elsewhere. In this paper we provide an extensive experimental study on the emergence of a phase transition in the problem of matching a Horn clause to a universe, searching for a model of the clause or for a proof that no such model exists. As it turns out, phase transition in the matching problem depends in an essential way on two order parameters, one capturing syntactic aspects of the clause structure (intensional aspect), while the other related to the structure of the universe (extensional aspect).

PDF

AIIM Journal 1989 Journal Article

Dealing with uncertain knowledge in medical decision-making: A case study in hepatology

  • Leonardo Lesmo
  • Lorenza Saitta
  • Pietro Torasso

It has widely been recognized that knowledge-based expert systems need efficient mechanisms to model the uncertainty associated with many decision-making activities. Such a need is particularly urgent in medicine. In this paper, we present an approach based on fuzzy logic to give a possible solution to this problem; its pros and cons are discussed by taking into account the experience gained in developing LITO1 and LITO2, two expert systems devoted to the assessment of the liver function and to the diagnosis of hepatic diseases. The advantages of mixing fuzzy production rules with frame-like structures (introduced for representing the clinical data) are discussed. In particular, the use of fuzzy linguistic variables for modeling the possible values of the clinical data is described: this allows, for a clear and perspicuous description of the correspondence between quantitative and qualitative expressions. Furthermore, different alternatives for evaluating and combining evidence are reviewed. Finally, the need of introducing frame structures also for representing diagnostic hypotheses is discussed, together with the problem of evaluating the fuzzy match between the prototypical description of a diagnostic hypothesis and the data describing the status of the specific patient under examination.

Details DOI

AAAI Conference 1982 Conference Paper

An Expert System for Interpreting Speech Patterns

  • Renato De Mori
  • Lorenza Saitta

Efficient syllabic hypothesization in continuous speech has been so far an unsolved problem. A novel solution based on the extraction of acoustic cues is proposed in this paper. This extraction is performed by parallel processes implementing an expert system represented by a grammar of frames.

PDF Details
v2026.09.13