Abstract
There is potential for important progress to be made in
understanding the design and enhancing the computational power of
artificial neural networks, but the approaches need not be biologically
motivated. Moreover, as the basis for automating intelligent behavior,
the manipulation of symbols remains a viable alternative to the neural
paradigm. There remain grand challenges to achieving with neural networks
the computational capabilities afforded by symbol manipulation. Hybrid
approaches coupling symbolic and neural processing have the potential to
overcome apparent deficiencies in the neural arena. Experimental evidence
of computational inefficiency in artificial neural networks may reflect
underlying theoretical limitations, as reported in the recent literature.
Intelligent Decision Aids
Intelligent M4 Systems
Machine Learning
Sensor-Based Systems
Back to NCARAI Library