** PRUNE**
Keyword of Binary
Type This keyword indicates which method will be used to optimise
input variables. This option is useful if you want to eliminate redundant
variables. The opruning procedures are described in references Tetko
et al., 1997 and Kovalishyn,
et. al. 1998. In general, sensetivities (or redundances) of input variables
are estimated following ensemble calculations and the least sensetive (or
most redundant) variable is eliminated. The pruning determines set of input
variables that calculates the smallest error for the validation data set.
This parameter should be used if **pruning** option is selected in ANALYSIS. The available pruning options include next sensetivity calculation methods:** wikel** {1} Sensetivity
is estimated as a sum of absolute values of outgoing weights.
**tetko** {2} Similar
to previous approach but using normalized weights.
**obd** {4} Sensetivy
is estimated according to the second derivatives of the network error as
a function of weights. The higher order and off-diagonal elements of the
Hessian matrix are ignored.
More details about the implementation of pruning methods and their comparison
can be found in Tetko et
al, 1996, Kovalishyn
et al, 1998.
__See FAQ__ if you have questions. **How to cite this applet?** Are you looking for a new **job in chemoinformatics?** |