this paper proposes a survivable lightpath provisioning scheme that allows traffic splitting in multi-domain optical networks to minimize the cumulative cost of a set of paths. the proposed scheme, called two-phase li...
详细信息
ISBN:
(纸本)9781479916337
this paper proposes a survivable lightpath provisioning scheme that allows traffic splitting in multi-domain optical networks to minimize the cumulative cost of a set of paths. the proposed scheme, called two-phase lightpath provisioning, employs an integer linear programming (ilp) formulation based on hierarchical path computation with full-mesh topology abstraction. there are two phases in the scheme. the first phase solves the ilp problem on an inter-domain topology and then feeds the results as intra-domain requests. the second phase solves the ilp problem in each related domain. Finally, we concatenate all the intra-domain solutions along routing sequences. three different protection strategies are considered with varying degrees of primary and backup route separation. Furthermore, to support various types of traffic demands, we investigate two cases in terms of the number of requested wavelengths. First, the number of requested wavelengths is less than link wavelength capacity. Second, the number of requested wavelengths is greater than link wavelength capacity. For the latter case, the proposed scheme allows traffic splitting among feasible primary and backup routes. the proposed scheme well supports the implementation of heuristic algorithms for lightpath provisioning since it can provide reference values, including upper and lower bounds, that are useful as benchmarks.
inductivelogicprogramming (ilp) deals withthe problem of finding a hypothesis covering positive examples and excluding negative examples. It uses first-order logic as a uniform representation for examples and hypot...
详细信息
the proceedings contain 36 papers. the topics discussed include: ontology-based information and event extraction for business intelligence;modelling highly symmetrical molecules: linking ontologies and graphs;personal...
ISBN:
(纸本)9783642331848
the proceedings contain 36 papers. the topics discussed include: ontology-based information and event extraction for business intelligence;modelling highly symmetrical molecules: linking ontologies and graphs;personalizing and improving tag-based search in folksonomies;views and synthesis of cognitive maps;identification of the compound subjective rule interestingness measure for rule-based functional description of genes;automatic generation and learning of finite-state controllers;FactForge: data service or the diversity of inferred knowledge over LOD;from path-consistency to global consistency in temporal qualitative constraint networks;rule quality measure-based induction of unordered sets of regression rules;a study on the utility of parametric uniform crossover for adaptation of crossover operator;and decomposition, merging, and refinement approach to boost inductivelogicprogramming algorithms.
this paper studies inductive definitions involving binders, in which aliasing between free and bound names is permitted. Such aliasing occurs in informal specifications of operational semantics, but;is excluded by the...
详细信息
ISBN:
(纸本)9783642005893
this paper studies inductive definitions involving binders, in which aliasing between free and bound names is permitted. Such aliasing occurs in informal specifications of operational semantics, but;is excluded by the common representation of binding as meta-level lambda-abstraction. Drawing upon ideas from functional logicprogramming, we represent. such definitions with aliasing as recursively defined functions in a higher-order typed functional programming language that extends core ML with types for name-binding, a type of "semi-decidable propositions" and existential quantification for types with decidable equality. We show that the representation is sound and complete with respect, to the language's operational semantics, which combines the use of evaluation contexts with constraint programming. We, also give a new and simple proof that the associated constraint problem is NP-complete.
the proceedings contain 26 papers from the conference on inductivelogicprogramming15thinternationalconference, ilp 2005. the topics discussed include: guiding inference through relational reinforcement learning;c...
详细信息
the proceedings contain 26 papers from the conference on inductivelogicprogramming15thinternationalconference, ilp 2005. the topics discussed include: guiding inference through relational reinforcement learning;converting semantic meta-knowledge into inductive bias;distance based generaliztion;automatic induction of abduction and abstraction theories from observations;strategies to parallelize ilp systems;inducing casual laws by regular inferece;spatial clustering of structured objects;predicate selection for structural decision trees;inductive equivalence of logic programs;and a study of applying dimensionality reduction to restrict the size of a hypothesis space.
inductivelogicprogramming (ilp) methods have proven to succesfully acquire knowledge with very different learning paradigms, such as supervised and unsupervised learning or relational reinforcement learning. However...
详细信息
ISBN:
(纸本)9783540738466
inductivelogicprogramming (ilp) methods have proven to succesfully acquire knowledge with very different learning paradigms, such as supervised and unsupervised learning or relational reinforcement learning. However, very little has been done on applying it to General Problem Solving (GPS). One of the ilp-based approaches applied to GPS is HAMLET. this method learns control rules (heuristics) for a non linear planner, PRODIGY4.0, which is integrated into the IPSS system;control rules are used as an effective guide when building the planning search tree. Other learning approaches applied to planning generate macro-operators, building high-level blocks of actions, but increasing the branching factor of the search tree. In this paper, we focus on integrating the two different learning approaches (HAMLET and macro-operators learning), to improve a planning process. the goal is to learn control rules that decide when to use the macro-operators. this process is successfully applied in several classical planning domains.
A consequence of ilp systems being implemented in Prolog or using Prolog libraries is that, usually, these systems use a Prolog internal database to store and manipulate data. However, in real-world problems, the orig...
详细信息
ISBN:
(纸本)9783540738466
A consequence of ilp systems being implemented in Prolog or using Prolog libraries is that, usually, these systems use a Prolog internal database to store and manipulate data. However, in real-world problems, the original data is rarely in Prolog format. In fact, the data is often kept in Relational Database Management Systems (RDBMS) and then converted to a format acceptable by the ilp system. therefore, a more interesting approach is to link the ilp system to the RDBMS and manipulate the data without converting it. this scheme has the advantage of being more scalable since the whole data does not need to be loaded into memory by the ilp system. In this paper we study several approaches of coupling ilp systems with RDBMS systems and evaluate their impact on performance. We propose to use a Deductive Database (DDB) system to transparently translate the hypotheses to relational algebra expressions. the empirical evaluation performed shows that the execution time of ilp algorithms can be effectively reduced using a DDB and that the size of the problems can be increased due to a non-memory storage of the data.
Effectiveness and efficiency are two most important properties of ilp approaches. For both top-down and bottom-up search-based approaches, greater efficiency is usually gained at the expense of effectiveness. In this ...
详细信息
ISBN:
(纸本)9783540696087
Effectiveness and efficiency are two most important properties of ilp approaches. For both top-down and bottom-up search-based approaches, greater efficiency is usually gained at the expense of effectiveness. In this paper, we propose a bottom-up approach, called ilp by instance patterns, for the problem of concept learning in ilp. this approach is based on the observation that each example has its own pieces of description in the background knowledge, and the example together withthese descriptions constitute a instance of the concept subject to learn. Our approach first captures the instance structures by patterns, then constructs the final theory purely from the patterns. On the effectiveness aspect, this approach does not assume determinacy of the learned concept. On the efficiency aspect, this approach is more efficient than existing ones due to its constructive nature, the fact that after the patterns are obtained, boththe background and examples are not needed anymore, and the fact that it does not perform coverage test and needs no theorem prover.
this paper presents a brief introduction of the relation between logicprogramming and machine learning. the area researching the relation is usually called inductivelogicprogramming (ilp, for short). In this paper ...
详细信息
ISBN:
(纸本)9783540692331
this paper presents a brief introduction of the relation between logicprogramming and machine learning. the area researching the relation is usually called inductivelogicprogramming (ilp, for short). In this paper we will give the details of neither ilp systems nor ilptheories. We explain how to substitute concepts used in logicprogramming to items needed in formulating learning theories. We also show some theoretical applications to which the substitution are contributing.
Many domains in the field of inductivelogicprogramming (ilp) involve highly unbalanced data. A common way to measure performance in these domains is to use precision and recall instead of simply using accuracy. the ...
详细信息
Many domains in the field of inductivelogicprogramming (ilp) involve highly unbalanced data. A common way to measure performance in these domains is to use precision and recall instead of simply using accuracy. the goal of our research is to find new approaches within ilp particularly suited for large, highly-skewed domains. We propose Gleaner, a randomized search method that collects good clauses from a broad spectrum of points along the recall dimension in recall-precision curves and employs an "at least L of these K clauses" thresholding method to combine sets of selected clauses. Our research focuses on Multi-Slot Information Extraction (IE), a task that typically involves many more negative examples than positive examples. We formulate this problem into a relational domain, using two large testbeds involving the extraction of important relations from the abstracts of biomedical journal articles. We compare Gleaner to ensembles of standard theories learned by Aleph, finding that Gleaner produces comparable testset results in a fraction of the training time.
暂无评论