Debray S., Kannan S., Paithane M, Weighted Decision Trees, in Apt K.R.(Ed.), Logic Programming - Proceedings of the Joint International Conference and Symposium on Logic Programming, MIT Press, Cambridge, MA, pp.654-668, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....their complexity class. Plans for the future development of the CHR implementation are the introduction of a priority scheme, realized through a scheduler [UeCh85] that fixes the remaining degrees of freedom with respect to sequencing 4 , and the factorization of common matching instructions [Deb92]. More information about CHR is available at http: www.informatik.uni muenchen.de fruehwir chr intro.html ....

Debray S., Kannan S., Paithane M, Weighted Decision Trees, in Apt K.R.(Ed.), Logic Programming - Proceedings of the Joint International Conference and Symposium on Logic Programming, MIT Press, Cambridge, MA, pp.654-668, 1992.

....never execute a test whose outcome can be predicted. Thus CMAs with these properties would have optimum performance in practical use. The foundational work of Kliger and Shapiro [7, 8] on decision graphs may be directly incorporated into our CMA formalism, as may ideas presented by Debray et al. [2] on exploiting probabilistic information in clause matching. The work of Podelski and Van Roy [10] on incremental clause matching, and the work on term pattern matching in nonstrict functional languages [9] is clearly relevant, although it will be more difficult for us to incorporate. All this ....

....in practice [6] Kliger s algorithms have been influential in the CCL implementation community. Examples include the decision graphs used in the Monaco [14] FGHC compiler and a variant of decision trees developed for determinacy testing in Andorra like language compilation [16] Debray et al. [2] proposed an alternate indexing heuristic, intended to allow optimal native code generation from a decision tree. This heuristic is based on the idea that a more realistic measure of which test should be performed first is to maximize the informationtheoretic entropy dispelled by the test, i.e. ....

S. K. Debray, S. Kannan, and M. Paithane. Weighted Decision Trees. In Joint International Conference and Symposium on Logic Programming, pages 654--668. Washington D.C., MIT Press, November 1992.

....the profiling information, as input, and produce, as output, the distribution of each agent terminating within k steps. Sections 4 and 5 respectively present the related and future work; whereas section 6 presents our conclusions. 2 On line profiling component Like other probabilistic analyses [1], our analysis is based on being able to associate execution frequencies (or, when non zero, normalized, execution probabilities) with the cases defining an agent. Light weight profiling techniques have been developed for logic programs [5] and adapted to concurrent logic programs such as Strand ....

S. Debray, S. Kannan, and M. Paithane. Weighted Decision Trees. In Proceedings of the Joint International Conference and Symposium on Logic Programming, pages 654--668. MIT Press, 1992.

..... An indexing tree depends, not only on the clauses in the definition, but also on the choice of test literals used to label the branches. Ideally, these test literals should be chosen so as to minimise the depth of the tree (or, perhaps, minimise the average cost of selecting a clause in the tree [16]) For instance, preference could be given to test literals that cause the nodes with the same parent to have disjoint candidate clause sets. An indexing tree has been defined for any predicate but, for a primitive determinate predicate, it is always possible to extend an indexing tree for it so ....

Saumya Debray, Sampath Kannan, and Mukul Paithane. Weighted decision trees. In K. Apt, editor, Proceedings of the Joint International Conference on Logic Programming, Washington, USA, pages 654--668, 1992.

.... This strategy moves work out of the normal non suspending code and into the suspension code, so that procedures that do not suspend do not 1 We currently do not have decision tree compilation implemented: we are investigating decision tree compilation in the presence of execution weights, see [5] for details. have to pay a high performance cost for the language s suspension feature. But this strategy does not work for variables that are not parameters, because there is no guarantee that the variable is available when the suspension code gets executed. The reason is that any ....

S. K. Debray, S. Kannan, and M. Paithane, "Weighted Decision Trees", Proc. Joint International Conference and Symposium on Logic Programming, Washington, D.C., Nov. 1992 (this volume).

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