Information Technology Reference
In-Depth Information
5
Summary and Future Research
In adaptive IF, user profile adaptation to changes in the user's interests poses a
challenging and fascinating research problem that invites the application of bio-
logically inspired solutions. In this paper we evaluated the ability of an immune-
inspired user profiling model, called Nootropia, to adapt to various changes in a
user's multiple interests. The evaluation methodology that we employed is based
on the routing subtask of TREC 2001, but extends it in various ways. It modi-
fies it to reflect the multi-modal and time dependent nature of adaptive IF. The
results demonstrate that Nootropia exhibits the wanted adaptive behaviour. It
can learn two topics in parallel and reflect their short-term variations, learn an
emerging topic of interest, and forget a no longer interesting topic. In this later
case negative feedback is not necessary, but it facilitates the process. They also
demonstrate the importance of links and of the network structure that terms
and links compose.
We may argue that this first attempt to apply immune-inspired ideas to the
problem of adaptive IF has been promising and is worth further investigation.
Nootropia, a self-organising network of terms that exhibits dynamics, metady-
namics, endogenous selection and other properties in common with Varela's view
of the immune network, performed satisfactory in the task of adapting to sim-
ulated changes in a user's multiple interests. The evaluation methodology itself
poses a challenging setting for the evaluation of biological inspired algorithms
in general. We wish to improve and standardise this methodology to conduct
comparative experiments between Nootropia and an evolutionary approach. We
seek to provide further evidence that the ability of AIS to boost and maintain
diversity, in contrast to the elitist character of GAs, proves advantageous given
the challenge of adaptive IF. In general, we hope to promote further constructive
interaction between biologically inspired computing and IF.
References
1. Menczer, F.: ARACHNID: Adaptive retrieval agents choosing heuristic neighbor-
hoods for information discovery. In: 14th International Machine Learning Confer-
ence. (1997) 227-235
2. Gaspar, A., Collard, P.: From GAs to artificial immune systems: Improving adap-
tation in time dependent optimization. In: Proceedings of the Congress on Evolu-
tionary Computation. Volume 3., IEEE Press (1999) 1859-1866
3. Simoens, A., Costa, E.: An immune system-based genetic algorithm to deal with
dynamic environments: Diversity and memory. In: Sixth International Conference
on Neural Networks and Genetic Algorithms, Springer (2003) 168-174
4. Nanas, N., Uren, V., de Roeck, A., Domingue, J.: Multi-topic information filtering
with a single user profile. In: 3rd Hellenic Conference on Artificial Intelligence,
(2004) 400-409
5. Nanas, N., Uren, V., de Roeck, A.: Nootropia: a user profiling model based on a
self-organising term network. In: 3rd International Conference on Artificial Immune
Systems, (2004) 146-160
