Techniques, Technologies and Patents Related to Intelligent Educational Systems - E-Infrastructures and Technologies for Lifelong Learning - page 14

learner characteristics and for performing user
model management either in an environment of
collaborating IESs that exchange information to
initialize/update the user model or in a distributed
learning environment (Rishi & Govil, 2008).
Neural networks consist of many simple
interconnected processing units called neurons
(Haykin, 2008). Each connection from neuron
u j to neuron u i is associated with a numerical
weight w ij corresponding to the influence of u j to
u i . The output of a neuron is based on its inputs
and corresponding weights. Training data are used
to train a neural network to perform its desired
function. Different types of neural networks have
been developed such as Back-Propagation Neural
Networks, Radial Basis Function networks, Self-
Organizing Maps (SOMs), Hopfield networks,
Boltzmann machines, ART networks, etc. Neural
networks can be useful in the following situa-
tions: (a) when classification and clustering tasks
need to be performed for which no explanation is
required, (b) when generalization is required, (c)
when training data is available. Neural networks
can be used in various online/offline IES tasks
(e.g. online pedagogical tasks, offline analysis
of collected data). For instance, in (Villaverde,
Godoy & Amandi, 2006) a feed-forward neural
network evaluates a user's learning style based
on his/her actions. Also clustering capabilities
of neural networks (such as those of SOMs) can
be used to cluster learners supporting adaptation
(Legaspi et al., 2008).
Bayesian networks are graphs, where nodes
represent statistical concepts and links represent
mainly causal relations among them (Darwiche,
2009). Each link is assigned a probability, which
represents how certain is that the concept where
the link departs from causes (leads to) the concept
where the link arrives at. Bayesian networks are
either defined by experts or learned from available
data. Bayesian networks enable uncertainty model
development in user modeling and evaluation.
For instance, in (Butz, Hua & Maguire, 2008)
Bayesian networks are used to track learner knowl-

edge regarding each domain concept. Liu (2008)
uses Bayesian networks to represent composite
concepts learning. Besides learner knowledge,
Bayesian networks can be used to evaluate other
user model characteristics such as learning style
(Botsios, Georgiou & Safouris, 2008).
Fuzzy methods are used to incorporate vague-
ness and uncertainty handling into IESs (Jameson,
1995). Fuzzy methods are based on fuzzy sets
to express membership degrees of elements into
these sets. In fuzzy logic, the degree of truth of
a statement can range between [0, 1] and is not
constrained to the two truth values {true, false} as
in classic binary logic (Klir & Yuan, 1995). Fuzzy
expert systems constitute a popular application of
fuzzy logic, which use fuzzy rules to infer conclu-
sions. Fuzzy logic inference process includes three
phases: fuzzification of inputs (via membership
functions), application of fuzzy rules and defuzzi-
fication (to produce the output). Fuzzy concepts
can be effectively incorporated into other methods/
techniques (e.g. fuzzy clustering, fuzzy decision
trees, fuzzy rough sets, fuzzy cognitive maps and
fuzzy Bayesian networks). Fuzzy methods can be
used when corresponding data (e.g. membership
functions, membership degrees, fuzzy rules) can
be obtained and when representation of vagueness
and uncertainty is required. Fuzzy methods can be
useful in various online/offline IES aspects (i.e. do-
main knowledge construction, user modeling and
evaluation, pedagogical tasks and analysis of data
collected during IES operation). The approaches
concerning use of RBR in IES tasks (mentioned
above) could potentially be improved with the
use of fuzzy rules. In (Chen & Duh, 2008) fuzzy
item response theory is used to evaluate learner
ability. In (Jing et al., 2006) fuzzy logic is used
for instructional strategy selection. In (Hwang &
Yang, 2009) fuzzy integrals were used to assess
affective states of the students. Fuzzy cognitive
maps have been used for user modeling (Laureano-
Cruces et al., 2009). In (Baia & Chen, 2008) fuzzy
rules and fuzzy reasoning techniques are used to
automatically construct concept maps and evaluate