VALID | VALIDATION
Specifies that the DATA= data set is a validation set. The data set must contain the target
variable.
TEST
Specifies that the DATA= data set is a test set. The data set must contain the target variable.
SCORE
Specifies that residuals, error functions, and fit statistics are not produced. The data set does
not have to contain the target variable.
Default:
TEST, except as follows:
TRAIN when the DATA= data set in the PROC statement is the same as
the DATA= data set in the SCORE statement. Specifying TRAIN
with any data set other than the actual training set is an error.
VALID when the DATA= data set in the SCORE statement is the same as
the VALIDATA= data set in the PROC statement.
Copyright 2000 by SAS Institute Inc., Cary, NC, USA. All rights reserved.
The NEURAL Procedure
SET Statement
Sets the value of the weight-list to number. The SET statement does not freeze weights, so
subsequent training may change the values of the weights specified in a SET statement.
Category Action Statement - affects the network or the data sets. Options set in an action statement
affect only that statement.
SET weight-list number;
Required Arguments
weight-list
Specifies the list of weights to be affected or changed.
number
Specifies the number to which the weight-list is set.
Copyright 2000 by SAS Institute Inc., Cary, NC, USA. All rights reserved.
The NEURAL Procedure
SHOW Statement
Prints information about the network.
Category Action Statement - affects the network or the data sets. Options set in an action statement
affect only that statement.
NOTE: At least one option must be specified, but there is no single argument that is required.
SHOWWEIGHTS
STATEMENTS
Options
Specify at least one:
STATEMENTS
Prints statements that can be used with the NEURAL procedure to reproduce the network.
WEIGHTS
Prints the network weights.
Copyright 2000 by SAS Institute Inc., Cary, NC, USA. All rights reserved.
The NEURAL Procedure
TARGET Statement
Defines an output layer.
Category Action Statement - affects the network or the data sets. Options set in an action statement
affect only that statement.
Alias: OUTPUT
TARGET | OUTPUT variable-list /
<
ACT=
activation-function>
<
BIAS|NOBIAS >
<
COMBINE=
combination-function>
<
ERROR=
keyword>
<
ID=
name>
<
LEVEL=
value>
<
MESTA=
number>
<
MESTCON=
number>
<
SIGMA=
number>
<
STD=
method>;
Required Arguments
variable-list
Specifies the target variables.
ID=name
Specifies the identifier for the layer.
Options
ACT=activation-function
Specifies the activation function. See
Activation Functions
.
Default:
Depends on the measurement level, as follows:
If LEVEL=INTERVAL, then the default is IDENTITY.
If LEVEL=ORDINAL then the default is LOGISTIC.
If LEVEL=NOMINAL, then the default is MLOGISTIC
(For Error=MBERNOULLI, MENTROPY, or MULTINOMIAL, the
only activation function allowed is MLOGISTIC.)
BIAS | NOBIAS
Specifies whether to use bias (or not to use bias).
Default:
BIAS
COMBINE=combination-function
Specifies the combination function. See
Combination Functions
.
ERROR=keyword
Specifies the Error function. Default is NORMAL for LEVEL=INTERVAL; otherwise, default is
MBERNOULLI. For more information, see the Error Functions table that follows.
Error Functions
KEYWORD
TARGET
DESCRIPTION
Functions with scale parameters :
NORmal
any
Normal
distribution
CAUchy
any
Cauchy
distribution
LOGistic
any
Logistic
distribution
HUBer
any
Huber M
estimator
BIWeight
any
Biweight M
estimator
WAVe
any
Wave M
estimator
GAMma
>0
Gamma
distribution
POIsson
0
Poisson
distribution
Functions with no scale parameter
BERnoulli
0,1
Bernoulli
distribution
(binomial with
one trial)
BINomial
0
Binomial
distribution
ENTropy
0-1
Cross or relative
entropy for
independent
targets
MBErnoulli
0,1
Multiple
Bernoulli
(multinomial
with one trial)
MULtinomial
0
Multinomial
distribution
MENtropy
0-1
Cross or relative
entropy for
targets that sum
to 1
(Kullback-Leibler
divergence)
LEVEL=value
Specifies the measurement level, where value can be:
NOMINAL|NOM
Nominal.
ORDINAL|ORD
Ordinal.
INTERVAL|INT
Interval.