Add support for sparse feature vectors

[GoogleCodeExporter]

SimpleFeatureStore (implements FeatureStore) currently stores feature vectors 
of all instances in an array. For sparse features vectors (with many zeros = 
non-present features), it might be too memory consuming.

-> Add FeatureStore implementation that would internally store feature vectors 
in some hash map format.

(follow up of this discussion at the dev-mailing-list: 
https://groups.google.com/forum/#!topic/dkpro-tc-dev/KfbyLKlFNCQ )

Original issue reported on code.google.com by ivan.hab...@gmail.com on 15 Oct 2014 at 12:26

• Blocked on: #200

[GoogleCodeExporter]

This issue was updated by revision r1165.

Initial implementation of SparseFeatureStore. Needs tests. Default value of 
non-present features is set to 0 -- any other existing solution for that?

Original comment by ivan.hab...@gmail.com on 15 Oct 2014 at 12:33

[GoogleCodeExporter]

This issue was updated by revision r1170.

SparseFeatureStore uses null for empty features; added tests.

Original comment by ivan.hab...@gmail.com on 16 Oct 2014 at 7:19

[GoogleCodeExporter]

Original comment by ivan.hab...@gmail.com on 16 Oct 2014 at 8:46

• Now blocked on: #200

[GoogleCodeExporter]

Fixed by fixing issue 200.

Original comment by ivan.hab...@gmail.com on 16 Oct 2014 at 10:41

• Changed state: Fixed

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

This issue was updated by revision r1176.

Adding check to keep feature vectors consistent

Original comment by ivan.hab...@gmail.com on 16 Oct 2014 at 12:21

[GoogleCodeExporter]

updated by r1226.

Renaming SimpleFeatureStore to DenseFeatureStore. For now, both Dense and 
Sparse FeatureStores remain in the SimpleFeatureStore module.

Original comment by daxenber...@gmail.com on 19 Nov 2014 at 6:16