Classifier Naming System


Description

The ClassX classifier naming system was developed by Anatoly Suchkov of STScI to bring some order to the variety of classifiers generated in the ClassX project. This naming system will most likely evolve in the future as the variety of classifiers grows. For now, a classifier name is formatted as follows:

yyyymmdd[_|a]MzzzzNNs

where

yyyy = Four-digit year from the generation date of the training data sample that the classifier was derived from.

mm = Two-digit month (with leading 0 where needed) from the generation date of the training data sample that the classifier was derived from.

dd = Two-digit day (with leading 0 where needed) from the generation date of the training data sample that the classifier was derived from.

M = A numeric code for approximating the revision of the training sample, since many versions have been used over the course of the project.

zzzz = Four separate character codes representing, in order, X-ray (x), infrared (i), optical (o), and radio (r) correlations used in the classification. If a particular band is not used, it is represented by an e, for "empty".

NN = Number of classes used in the classification, with leading 0 as needed.

s = Variable-length alphanumeric suffix used to distinguish otherwise indistinguishable classifiers.

Example

Consider the following ClassX classifier name:

20030212_5xeor06e

The name may be parsed as follows:

20030212 - The classifier was derived from the 20030212 training data sample, released on 12 February 2003.

5 - This classifier was trained using version 5 of the 20030212 training sample.

xeor - The classifier used X-ray (x), optical (o), and radio (r) data, but not infrared (e instead of i).

06 - The classifier uses 6 distinct classes.

e - This is classifier e of this family of classifiers (20030212_5xeor06-based).


Date of last modification: Monday, December 01, 2003
Responsible NASA official: Tom McGlynn

Curator: Eric Winter