Neural network construction via back-propagation
Conference
·
OSTI ID:10194019
A method is presented that combines back-propagation with multi-layer neural network construction. Back-propagation is used not only to adjust the weights but also the signal functions. Going from one network to an equivalent one that has additional linear units, the non-linearity of these units and thus their effective presence is then introduced via back-propagation (weight-splitting). The back-propagated error causes the network to include new units in order to minimize the error function. We also show how this formalism allows to escape local minima.
- Research Organization:
- Stanford Linear Accelerator Center, Menlo Park, CA (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC03-76SF00515
- OSTI ID:
- 10194019
- Report Number(s):
- SLAC-PUB-6516; CONF-9406278-1; ON: DE95002867; TRN: 94:023949
- Resource Relation:
- Conference: Stanford PDP research meeting,Stanford, CA (United States),23 Jun 1994; Other Information: PBD: Jun 1994
- Country of Publication:
- United States
- Language:
- English
Similar Records
Improving convergence in neural networks
Cost estimation of timber bridges using neural networks
Prediction of protein tertiary structure from sequences using a very large back-propagation neural network
Miscellaneous
·
Sun Jan 01 00:00:00 EST 1989
·
OSTI ID:10194019
Cost estimation of timber bridges using neural networks
Journal Article
·
Mon May 01 00:00:00 EDT 1995
· Cost Engineering
·
OSTI ID:10194019
Prediction of protein tertiary structure from sequences using a very large back-propagation neural network
Conference
·
Fri Dec 31 00:00:00 EST 1993
·
OSTI ID:10194019