Information Technology Reference
In-Depth Information
Immune-Inspired
Adaptive Information Filtering
Nikolaos Nanas 1 , Anne de Roeck 1 , and Victoria Uren 2
1 Computing Department, The Open University, Milton Keynes, MK7 6AA, U.K.
{ N.Nanas, A.deRoeck } @open.ac.uk
2 Knowledge Media Institute, The Open University, Milton Keynes, MK7 6AA, U.K.
{ V.S.Uren } @open.ac.uk
Abstract. Adaptive information filtering is a challenging research prob-
lem. It requires the adaptation of a representation of a user's multiple
interests to various changes in them. We investigate the application of an
immune-inspired approach to this problem. Nootropia, is a user profiling
model that has many properties in common with computational mod-
els of the immune system that have been based on Franscisco Varela's
work. In this paper we concentrate on Nootropia's evaluation. We define
an evaluation methodology that uses virtual user's to simulate various
interest changes. The results show that Nootropia exhibits the desirable
adaptive behaviour.
1
Introduction
Information Filtering (IF) systems seek to provide a user with relevant infor-
mation based on a tailored representation of the user's interests, a user profile .
The user interests are considered to be long-term. Consequently, a user may
be interested in more than one topic in parallel. Also, changes in user interests
are inevitable and can vary from modest to radical. In addition to short-term
variations in the level of interest in certain topics, new topics of interest may
gradually emerge and interest in existing topics may wane. Adaptive IF deals
with the problem of adapting the user profile to such interest changes.
Profile adaptation to changes in a user's multiple interests is a fascinating and
challenging problem that has already attracted biologically-inspired approaches.
Evolutionary IF systems maintain a population of profiles (chromosomes) to
represent a user's interests and apply Genetic Algorithms-inspired by natu-
ral evolution-to evolve the population and thus adapt the profiles to changes
in them. These approaches treat profile adaptation as a continuous optimisa-
tion problem and tackle it by performing combined global and local search in a
stochastic, but directed fashion.
Profile adaptation however is not a traditional optimisation problem. As Fil-
lipo Menczer puts it [1], it is a ”multimodal” and time-dependent one, where
convergence to a single optimum should be avoided. A user's multiple and chang-
ing interests translate into an information space where there are multiple optima
that change over time. It has been argued and supported experimentally [2,3],
Search WWH ::




Custom Search