|
|
Knowledge and Information Systems |
Abstract. This paper presents a critical review of the most important evolutionary-based multiobjective optimization techniques developed over the years, emphasizing the importance of analyzing their Operations Research roots as a way to motivate the development of new approaches that exploit the search capabilities of evolutionary algorithms. Each technique is briefly described with its advantages and disadvantages, its degree of applicability and some of its known applications. Finally, the future trends in this discipline and some of the open areas of research are also addressed.
Abstract. In this paper we introduce a new multidimensional
index structure called the S-tree. Such indexes are appropriate for a
large variety of pictorial databases such as cartography, satellite
and medical images. The S-tree discussed in this paper is similar in
flavor to the standard R-tree, but accepts mild imbalance in the
resulting tree in return for a significantly reduced area, overlap and
perimeter in the resulting minimum bounding rectangles. In fact, the
S-tree is defined in terms of a parameter which governs the degree to
which this trade-off is allowed. We develop an efficient packing
algorithm based on this parameter. We then analyze the S-tree
analytically, giving theoretical bounds on the degree of imbalance of
the tree. We also analyze the S-tree experimentally. The S-tree does
well in two dimensions, and even better in three dimensions. Indeed,
the S-tree can be expected to do better still as the dimensionality
increases. While the S-tree is extremely effective for static
databases, we outline the extension to dynamic databases as well.
Abstract. Planning for the future is an important activity both
at the individual and organizational levels. Planning consists of
defining alternative actions to handle various events in the future.
The alternatives arise because of different possible outcomes of
events. A plan consists of a sequence of actions to be carried out for
each possible outcome. In the context of database modeling, the
actions are operations on a database. A database management system
should enable its users to define events and alternatives, and also
allow them to interact with the database under different alternatives
(possibly to evaluate different plans). The existing temporal data
models treat the future analogous to the past or present; they provide
for one future path (in the sense that facts valid at some future time
can be stored), but do not provide support for alternatives in the
future. In this paper, we present a model for incorporating events
and alternatives by extending the temporal data model to support
branching time. The extended model permits definitions of events,
their inter-dependencies and associated actions. The events that
affect an object are modeled by a tree, permitting an object to have
different states at the same valid time but under different
alternatives. The branching time paradigm is obtained by superimposing
a linear valid time on the event tree. We extend the temporal
relational algebra and the Temporal SQL2 to support a branching time
data model. The paper also briefly deals with the uncertainties
associated with future planning as well as probabilities of possible
event outcomes. Finally, we sketch an implementation strategy for the
branching time data model.
Abstract. In this paper we study the problem of searching the
Web with online learning algorithms. We consider that web documents
can be represented by vectors of $n$ boolean attributes. A search
engine is viewed as a learner, and a user is viewed as a teacher. We
investigate the number of queries a search engine needs from the user
to search for a collection of web documents. We design several
efficient learning algorithms to search for any collection of
documents represented by a disjunction (or a conjunction) of relevant
attributes with the help of membership queries or equivalence queries.
Abstract. Considering the importance of the domain relationship
in eliminating noisy features in feature selection, we present an
alternate approach to designing a multi-objective fitness function
using multiple correlation for the genetic algorithm (GA), which is
used as a search tool in the problem. Multiple correlation is a simple
statistical technique that uses the multiple correlation coefficients
to measure the relationship between a dependent variable and a set of
independent variables within the domain space. Simulation studies
were conducted on both real-world and controlled data sets to assess
the performance of the proposed fitness function. The comparison
between the traditional fitness function and our proposed function is
also reported. The results show that the proposed fitness function can
perform more satisfactorily than the traditional one in all cases
considered, including different data types, multi-class and
multi-dimensional data.
Handling of Alternatives and Events in Temporal Databases
N.L. Sarda and P.V. Siva Prasada ReddyShort Papers
Searching the Web with Queries
Zhixiang Chen, Xiannong Meng, and Richard H. FowlerFeature Selection Using the Domain Relationship with Genetic
Algorithms
Nidapan Chaikla and Yulu Qi
This page has been accessed times since July 6, 1999.