commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gilles (JIRA)" <j...@apache.org>
Subject [jira] [Created] (MATH-1278) Deep copy of "Network" ("o.a.c.m.ml.neuralnet")
Date Sat, 19 Sep 2015 19:16:04 GMT
Gilles created MATH-1278:
----------------------------

             Summary: Deep copy of "Network" ("o.a.c.m.ml.neuralnet")
                 Key: MATH-1278
                 URL: https://issues.apache.org/jira/browse/MATH-1278
             Project: Commons Math
          Issue Type: Wish
            Reporter: Gilles
            Assignee: Gilles
            Priority: Minor
             Fix For: 4.0, 3.6


It might be useful to track the changes in a ANN during the learning process.

Certain algorithms for assessing the state of a SOFM can take a long time if the training
dataset in very large.
If the state of the {{Network}} changes a lot (due to the continuing training process) during
the evaluation computation, its output will not represent the state at some definite time.

Stopping the training process during the whole evaluation would result in a waste of time.

An alternative is to create a deep copy of the {{Network}} instance: the evaluation will be
performed on the copy while the training can continue on the original instance.




--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message