A. Srinivasan and R. C. Camacho. Numerical reasoning with an ILP program capable of lazy evaluation and customised search. Journal of Logic Programming, 40(2,3):185-214, 1999. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
From Propositional to First Order Logic in Machine Learning and.. - Van Laer (2002) (1 citation) (Correct)
....the future. In general the complete language specification is (still) a challenging research area (cf. the Aladin project and papers like [Knobbe et al. 2000] Several other techniques have been proposed for handling numerical data in ILP systems, which are very different from our approach. In [Srinivasan and Camacho, 1999] for instance, one extends PROGOL with lazy evaluation of literals. The idea is to introduce existentially quantified variables (or skolem constants) in literals of the bottom clause, at places where constants should appear. These skolem constants are left unevaluated until refinement. At that ....
....background predicates return the molecular weight mweight (a real number) and logP (logarithm of the octanol water partition coefficient) for each compound. The representation and background knowledge is very similar to those used in mutagenicity and toxicity prediction [Srinivasan et at. 1996; Srinivasan et at. 1999]. In [D2eroski et al. 1999] the relational data (denoted R) and 2 derived propositional versions of the data (denoted P1 and P2) are used in the exper BG BG3 4 BG4 B . 0 2 4 6 8 10 12 14 16 18 20 (a) 96 94 92 o BG2 BG3 4 BG4 B . 2 4 6 8 10 12 14 16 18 20 ....
[Article contains additional citation context not shown here]
A. Srinivasan and R. C. Camacho. Numerical reasoning with an ILP program capable of lazy evaluation and customised search. Journal of Logic Programming, 40(2,3):185-214, 1999.
Extracting context-sensitive models in Inductive Logic Programming - Srinivasan (2001) (3 citations) Self-citation (Srinivasan) (Correct)
....label to denote unknown . C(HjB;E) is to be read as the cost of H given B and E. As it is phrased, H need not be all possible models for E, but only those considered. The model is therefore only optimal in a restricted sense. To our knowledge, at least one ILP implementation (described in [28]) allows the minimisation of a user de ned cost function. However, the notion in Fig. 3 of providing a set of cost functions does not appear to have been suggested before. This de nes a setting within which to address the problem of varying contexts described in the previous section (with each ....
....before. This de nes a setting within which to address the problem of varying contexts described in the previous section (with each context mapping to an element of C) Implementations that restrict C and H to singleton sets result in special cases. These include: a) the system described in [28], where no restrictions are placed on the cost function; b) systems that construct models to minimise expected error. These use a particular cost function that assigns the same cost for all misclassi cation errors; and (c) systems that construct models for the 2 class case by assigning an in nite ....
A. Srinivasan and R.C. Camacho. Numerical reasoning with an ILP program capable of lazy evaluation and customised search. Journal of Logic Programming, 40(2,3):185-214, 1999.
Four Suggestions and a Rule Concerning the Application of ILP - Srinivasan Self-citation (Srinivasan) (Correct)
....But this is uninteresting, and unsurprising. Of far more interest to the chemical community was a new structural explanation for a subset of the non linearisable data. This same data were used in [31] where a regression model was constructed using new attributes found by an ILP program, and in [30] where an ILP program used regression as background knowledge. In both cases, it became possible to predict the actual numerical activity to a much better degree of precision than had been possible earlier. Thus, while a first order representation is natural when reasoning with complex structured ....
....translation of chemical ring structures were found relatively easily by a non expert. Conversely, the use of opaque procedural encoding of a regression program caused considerable discomfort and required an elaborate sequence of checkpoints and assertions during the initial stages of the work in [30]. The process of settling on correct and adequate background knowledge should not be underestimated. I have been asked about a breakdown of the different activities that comprise a typical application the underlying thinking being, no doubt, that such a question could be answered quite easily ....
[Article contains additional citation context not shown here]
A. Srinivasan and R.C. Camacho. Numerical reasoning with an ILP program capable of lazy evaluation and customised search. Journal of Logic Programming, 40(2,3), 1999.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC