| commit | bd41da3588d363afda27ed0760b8fbf7490b5e70 | [log] [tgz] |
|---|---|---|
| author | Marc Kupietz <kupietz@ids-mannheim.de> | Fri Nov 24 10:26:43 2017 +0100 |
| committer | Marc Kupietz <kupietz@ids-mannheim.de> | Fri Nov 24 10:26:43 2017 +0100 |
| tree | 7f03dbef39c95e493890d36b2bdb4f8db2426bdc | |
| parent | 4dc270c90638feaf005dbfe0c4cc07087f62011e [diff] |
w2v-server: use conormalized activation sum for auto-focus Best results after trying all alternatives: * sum of gloablly normalized probabilities * sum of column-set normalized probabilities (also co-normalized) "best" meant as a combination of complementarity to max(actiavtion) and usefulness, particularly for low frequency words