add_sample_class_mlp🔗
Short description🔗
add_sample_class_mlp — Add a training sample to the training data of a multilayer
perceptron.
Signature🔗
add_sample_class_mlp( class_mlp MLPHandle, real Features, number Target )
Description🔗
add_sample_class_mlp adds a training sample to the
multilayer perceptron (MLP) given by MLPHandle. The
training sample is given by Features and Target.
Features is the feature vector of the sample, and
consequently must be a real vector of length NumInput, as
specified in create_class_mlp. Target is the
target vector of the sample, which must have the length
NumOutput (see create_class_mlp) for all three
types of activation functions of the MLP (exception: see below). If
the MLP is used for regression (function approximation), i.e., if
OutputFunction \(=\) 'linear', Target is
the value of the function at the coordinate Features. In
this case, Target can contain arbitrary real numbers. For
OutputFunction \(=\) 'logistic', Target can
only contain the values 0.0 and 1.0. A value of
1.0 specifies that the attribute in question is present,
while a value of 0.0 specifies that the attribute is
absent. Because in this case the attributes are independent,
arbitrary combinations of 0.0 and 1.0 can be
passed. For OutputFunction \(=\) 'softmax',
Target also can only contain the values 0.0 and
1.0. In contrast to OutputFunction \(=\)
'logistic', the value 1.0 must be present for
exactly one element of the tuple Target. The location in
the tuple designates the class of the sample. For ease of use, a
single integer value may be passed if OutputFunction \(=\)
'softmax'. This value directly designates the class of the
sample, which is counted from 0, i.e., the class must be an integer
between 0 and NumOutput \(-\) 1. The class is converted to a
target vector of length NumOutput internally.
Before the MLP can be trained with train_class_mlp,
all training samples must be added to the MLP with
add_sample_class_mlp.
The number of currently stored training samples can be queried with
get_sample_num_class_mlp. Stored training samples can be
read out again with get_sample_class_mlp.
Normally, it is useful to save the training samples in a file with
write_samples_class_mlp to facilitate reusing the samples,
and to facilitate that, if necessary, new training samples can be
added to the data set, and hence to facilitate that a newly
created MLP can be trained anew with the extended data
set.
Execution information🔗
Execution information
-
Multithreading type: reentrant (runs in parallel with non-exclusive operators).
-
Multithreading scope: global (may be called from any thread).
-
Processed without parallelization.
This operator modifies the state of the following input parameter:
During execution of this operator, access to the value of this parameter must be synchronized if it is used across multiple threads.
Parameters🔗
MLPHandle (input_control, state is modified) class_mlp → (handle)
MLP handle.
Features (input_control) real-array → (real)
Feature vector of the training sample to be stored.
Target (input_control) number(-array) → (integer / real)
Class or target vector of the training sample to be stored.
Result🔗
If the parameters are valid, the operator
add_sample_class_mlp returns the value 2 (H_MSG_TRUE). If necessary,
an exception is raised.
Combinations with other operators🔗
Combinations
Possible predecessors
Possible successors
train_class_mlp, write_samples_class_mlp
Alternatives
See also
clear_samples_class_mlp, get_sample_num_class_mlp, get_sample_class_mlp
Module🔗
Foundation