Conventional computer architectures and artificial neural networks have radically different structures and utilities. While computers are used to quickly perform complex but clearly defined tasks, neural networks are particularly adapted to solve more blurred problems like those encountered in perception, recognition and optimization. The hope is that neural networks might share some of the processing capabilities of the human brain. This is especially interesting in perception problems which are very long to compute by conventional computers because of the need for powerful sequential algorithms. The tradeoff between speed and efficiency implies the use of more powerful computers in order to run the complex algorithms needed to solve real-time perception problems. Nevertheless sequential solutions are not “natural” for this class of problems in which a large amount of simple but repetitive computation is needed.
[1]
J J Hopfield,et al.
Neural networks and physical systems with emergent collective computational abilities.
,
1982,
Proceedings of the National Academy of Sciences of the United States of America.
[2]
Geoffrey E. Hinton,et al.
Connectionist Architectures for Artificial Intelligence
,
1990,
Computer.
[3]
D. O. Hebb,et al.
The organization of behavior
,
1988
.
[4]
Richard P. Lippmann,et al.
An introduction to computing with neural nets
,
1987
.
[5]
Yannis Tsividis,et al.
Analogue circuits for variable-synapse electronic neural networks
,
1987
.
[6]
L.D. Jackel,et al.
An associative memory based on an electronic neural network architecture
,
1987,
IEEE Transactions on Electron Devices.