Figure 2: Feedback controllers measure the difference (error) between what we
want (reference) and what we have (output) and make corrections (control) based
on this difference.
For example, in the field of model creation for control purposes, artificial neural
networks have been compared favourably, in certain settings, with first principles
models in the implementation of nonlinear multivariable predictive control Hen-
riques et al. (2002). This neural network approach uses a recurrent Elman network
for capturing the plant’s dynamics, being the learning stage implemented on-line
using a modified version of the back-propagation through time algorithm Elman
(1990); Rumelhart et al. (1986).
All this analysis takes us to the formulation of a second principle of cognitive
system construction:
Principle 2: Model isomorphism — An embodied, situated, cognitive sys-
tem is as good as its internalised models are.
Model quality is measured in terms of some definable isomorphism with the
modelled system as established by the modelling relation. Let’s see why all this
disgression on model learning and quality is relevant for the consciousness en-
deavour.
3 Reactive vs Anticipatory Control
Many control mechanisms follow the well known error-feedback paradigm. This
control structure is so simple and robust that almost all control loops are based on
this approach. The strategy is simple and extremely effective Wiener (1961): mea-
sure the difference between what we want and what we have and make corrections
based on this difference (see Figure 2).
These controllers are very effective but have a serious drawback: they are al-
ways behind the plant, i.e. they cannot make the plant strictly follow a reference
signal without a delay (except for special plants in special circumstances). These
controllers just act as reaction to plant output diverting from what is desired (er-
rors); so they will wait to act until output error is significant.
ASLab.org / Principles for Consciousness / A-2007-011 v 1.0 Final
More intriguing information
1. Has Competition in the Japanese Banking Sector Improved?2. EFFICIENCY LOSS AND TRADABLE PERMITS
3. Work Rich, Time Poor? Time-Use of Women and Men in Ireland
4. The name is absent
5. Second Order Filter Distribution Approximations for Financial Time Series with Extreme Outlier
6. The open method of co-ordination: Some remarks regarding old-age security within an enlarged European Union
7. HACCP AND MEAT AND POULTRY INSPECTION
8. Smith and Rawls Share a Room
9. Ruptures in the probability scale. Calculation of ruptures’ values
10. Education Research Gender, Education and Development - A Partially Annotated and Selective Bibliography