... Page 3. 150 H.Comon proves the decidability of the ~:l-fragment in the AC-case. ... 6: A ~ #/\ / x# y zt 7: .... D i~/tt >,//\ 9~i ~y?\ ~ Xy zt >..,-Z, #I yZ #x yZ x III In III I fi~]ure2 --- B ,ViEI rAZe\ x I... more
... Page 3. 150 H.Comon proves the decidability of the ~:l-fragment in the AC-case. ... 6: A ~ #/\ / x# y zt 7: .... D i~/tt >,//\ 9~i ~y?\ ~ Xy zt >..,-Z, #I yZ #x yZ x III In III I fi~]ure2 --- B ,ViEI rAZe\ x I y,~i) ztu # # "~- B ,ViEI AA l\ xi zf,,o vw Y >. ,/\ t\ xy #z #tu > .</c\~= .v,~, z u > x/Z, A/\ i\ ...
Research Interests:
Research Interests:
Research Interests:
ABSTRACT This paper proposes a new effective filtering mechanism for pruning the uninteresting nodes implied in the SLCA-based (Smallest LCA – Lowest Common Ancestor) fragments for XML keyword search. Its fundamental concept is the valid... more
ABSTRACT This paper proposes a new effective filtering mechanism for pruning the uninteresting nodes implied in the SLCA-based (Smallest LCA – Lowest Common Ancestor) fragments for XML keyword search. Its fundamental concept is the valid contributor. Given two nodes v and u, and u is v’s parent, the child v is a valid contributor to its parent u, if (1) v’s label is unique among all u’s children; or (2) for the siblings with same label as v, v’s content is not covered by any of them. The new filtering mechanism can be described as following: every node in each retrieved fragment should be valid contributor to its parent.
Research Interests:
Search all the public and authenticated articles in CiteULike. Include unauthenticated results too (may include "spam") Enter a search phrase. You can also specify a CiteULike article id (123456),. a DOI (doi:10.1234/12345678).... more
Search all the public and authenticated articles in CiteULike. Include unauthenticated results too (may include "spam") Enter a search phrase. You can also specify a CiteULike article id (123456),. a DOI (doi:10.1234/12345678). or a PubMed Id (pmid:12345678). ...
Research Interests:
Research Interests:
... the set of rules has no critical pair (ie there are no overlappings between left-hand sides), which ensures the confluence only in the left-linear case (Huet [14], Huet & Levy ... Let S be a left-linear trf rewrite system,... more
... the set of rules has no critical pair (ie there are no overlappings between left-hand sides), which ensures the confluence only in the left-linear case (Huet [14], Huet & Levy ... Let S be a left-linear trf rewrite system, k = max{depth(l) / I ~ rE S}+1 and Ty.(X,k) be the set of k-normal trees ...
Research Interests:
Research Interests:
... Ganzinger and Waldmann [BGW93] show that the class of positive set constraints (with projec ... The monadic class is the class of first order formulas without function symbols, with unary pred ... AKVW93] presents several results on... more
... Ganzinger and Waldmann [BGW93] show that the class of positive set constraints (with projec ... The monadic class is the class of first order formulas without function symbols, with unary pred ... AKVW93] presents several results on the computational complex-ity of solving systems of ...
Research Interests:
We dene a new PAC learning model. In this model, examples aredrawn according to the universal distribution m(: j f) of SolomomooeLevin,where f is the target concept. The consequence is that the simpleexamples of the target concept have a... more
We dene a new PAC learning model. In this model, examples aredrawn according to the universal distribution m(: j f) of SolomomooeLevin,where f is the target concept. The consequence is that the simpleexamples of the target concept have a high probability to be provided tothe learning algorithm. We prove an Occam's Razor theorem. We showthat the class of poly-term DNF
Research Interests:
Research Interests:
Research Interests:
... We notice that we associate with a system of set constraints a simple first-order formula and that we ... Satisfiability of systems of set constraints is decidable, if a system of set constraints is satisfiable ... ie a tuple of... more
... We notice that we associate with a system of set constraints a simple first-order formula and that we ... Satisfiability of systems of set constraints is decidable, if a system of set constraints is satisfiable ... ie a tuple of regular tree languages), a minimal and a maximal solution which are ...
Research Interests:
ABSTRACT A positive set constraint is of the form exp 1 Í\subseteq exp 2, a negative set constraint is of the form exp 1 Í\subseteq exp 2 where exp 1 and exp 2 are set expressions constructed using set variables, function symbols, and the... more
ABSTRACT A positive set constraint is of the form exp 1 Í\subseteq exp 2, a negative set constraint is of the form exp 1 Í\subseteq exp 2 where exp 1 and exp 2 are set expressions constructed using set variables, function symbols, and the set union, intersection and complement symbols. Decision algorithms for satisfiability of systems of positive and negative set constraints were given by Gilleron et al. [GTT93b], Aiken et al. [AKW93], and Charatonik and Pacholski [CP94]. In this paper, we study properties of the set of solutions of such systems and properties of solutions of interest for applications. The main decidability results are: for positive and negative set constraints, equivalence of systems is decidable, it is decidable whether or not a system has a unique solution; for positive set constraints, it is decidable whether or not a system has a least solution, it is decidable whether or not a system has a finite solution (i.e. the interpretation maps each set variable on a finite set).
Research Interests:
Research Interests:
Research Interests:
In many learning problems, labeled examples are rare or expensive while numerous unlabeled and positive examples are available. However, most learning algorithms only use labeled examples. Thus we address the problem of learning with the... more
In many learning problems, labeled examples are rare or expensive while numerous unlabeled and positive examples are available. However, most learning algorithms only use labeled examples. Thus we address the problem of learning with the help of positive and unlabeled data given a small number of labeled examples. We present both theoretical and empirical arguments showing that learning algorithms can
Research Interests:
Research Interests:
Research Interests:
ABSTRACT This paper studies the problem of learning from a set of input graphs, each of them representing a different relation over the same set of nodes. Our goal is to merge those input graphs by embedding them into an Euclidean space... more
ABSTRACT This paper studies the problem of learning from a set of input graphs, each of them representing a different relation over the same set of nodes. Our goal is to merge those input graphs by embedding them into an Euclidean space related to the commute time distance in the original graphs. This is done with the help of a small number of labeled nodes. Our algorithm output a combined kernel that can be used for different graph learning tasks. We consider two combination methods: the (classical) linear combination and the sigmoid combination. We compare the combination methods on node classification tasks using different semi-supervised graph learning algorithms. We note that the sigmoid combination method exhibits very positive results.
Research Interests:
ABSTRACT Adapting keyword search to XML data has been attractive recently, generalized as XML Keyword Search (XKS). Its fundamental task is to retrieve meaningful and concise result for the given keyword query, and [1] is the latest work... more
ABSTRACT Adapting keyword search to XML data has been attractive recently, generalized as XML Keyword Search (XKS). Its fundamental task is to retrieve meaningful and concise result for the given keyword query, and [1] is the latest work which returns the fragments rooted at the SLCA (Smallest LCA - Lowest Common Ancestor) nodes. To guarantee the fragments only containing meaningful nodes, [1] proposed a contributor-based filtering mechanism in its MaxMatch algorithm. However, the filtering mechanism is not sufficient. It will commit the false positive problem (discarding interesting nodes) and the redundancy problem (keeping uninteresting nodes). In this paper, we propose a new filtering mechanism to overcome those two problems. The fundamental concept is valid contributor. A child v is a valid contributor to its parent u, if (1) v's label is unique among all u's children; or (2) for the siblings with same label as v, v's content is not covered by any of them. Our new filtering mechanism is: all the nodes in each retrieved fragment should be valid contributors to their parents. By doing so, it not only satisfies the axiomatic properties proposed by [1], but also ensures the filtered fragment more meaningful and concise. We implement our proposal in ValidMatch, and compare ValidMatch with MaxMatch on real and synthetic XML data. The result verifies our claims, and shows the effectiveness of our valid-contributor-based filtering mechanism.
Research Interests:
Research Interests:
Machine Learning, Data Mining, Semi-supervised Learning, Learning problems, Theoretical Computer Science, and 14 moreInformation Processing, Mathematical Sciences, Supervised Learning, Classification, Binary Classification, Learning Model, Probabilistic reasoning, Naive Bayes, PAC Learning, Lower Bound, Communication Complexity, Real-world Application, Unlabeled Data, and Decision tree Induction
Research Interests:
Page 1. PAC Learning under Helpful Distributions Fran cois Denis, R emi Gilleron LIFL, URA 369 CNRS, Universit e de Lille Iy ... This research was partially supported by "Motricit e et Cognition : Contrat par objectifs r egion... more
Page 1. PAC Learning under Helpful Distributions Fran cois Denis, R emi Gilleron LIFL, URA 369 CNRS, Universit e de Lille Iy ... This research was partially supported by "Motricit e et Cognition : Contrat par objectifs r egion Nord/Pas-de-Calais yLIFL, Bat. ...