Chapter Evolving Connectionist and Fuzzy Connectionist Systems: Theory and Applications for Adaptive, On-line Intelligent Systems

Fig.8. A block diagram of an intelligent, adaptive, on-line system for prediction, decision
making and control
Fig.9. An EFuNN is evolved from the flow F(t), the previous hour flow F(t-1), the moving
average 12 hours MA12h(t) and the moving average 24 hours MA24h(t) data. Here the real
values F(t+1) and the predicted by the evolving EFuNN for one day ahead are plotted as
the EFuNN evolves step by step through the time-series data.
9. ECOS and EFuNNs for evolving intelligent agents
9.1. The evolving, intelligent agents paradigm
Agent-based techniques allow for implementing modular systems that consist of
independent software modules that can communicate with each other and with the
user using a standard protocol, can 'navigate' in a new software environment
searching for relevant data, processing the data and passing results [77]. Intelligent
0
50
100
150
200
250
300
350
400
450
500
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pre
processing
(filtering)
ECOS
EFuNNs
Final
Decision
Making/
Control
Adaptation
Rule extraction,
explanation
Inputs
Outputs

---

136
agents can perform intelligent information processing, such as reasoning with
uncertainties and generalisation. Intelligent agents should be able to adapt to a
possibly changing environment as they work. Such adaptation is crucial for 
a
mobile robot navigation, or for an adequate decision making on operations with a
dynamically changing stock index [6,7,13,31]. Here the latter example is used to
illustrate the potential for using ECOS and 
EFuNNs for building intelligent
agents.