A One Hot Vector: Difference between revisions
From Algolit
(→Note that) |
|||
| Line 19: | Line 19: | ||
=Note that= | =Note that= | ||
| − | Words are represented once in a vector. So words with multiple meanings, like "bank", are more difficult to represent. | + | ''Words are represented once in a vector. So words with multiple meanings, like "bank", are more difficult to represent. There is research to multivectors for one word, so that it does not end up in the middle.'' - Richard Socher, for more notes on this lecture visit http://pad.constantvzw.org/public_pad/neural_networks_3 |
| − | |||
| − | |||
| − | |||
[[Category:Algoliterary-Encounters]] | [[Category:Algoliterary-Encounters]] | ||
Revision as of 14:46, 25 October 2017
| Type: | Algoliterary exploration |
| Technique: | word-embeddings |
| Developed by: | Algolit |
one-hot-vectors
"Meaning is this illusive thing that were trying to capture" (Richard Socher in CS224D Lecture 2 - 31st Mar 2016 (Youtube))
[0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0]
with one 0 for each word in a vocabulary where the 1 is representing the place of a word in the vector > In this kind of vector representation: none of the words are similar, they are all a 1.
Note that
Words are represented once in a vector. So words with multiple meanings, like "bank", are more difficult to represent. There is research to multivectors for one word, so that it does not end up in the middle. - Richard Socher, for more notes on this lecture visit http://pad.constantvzw.org/public_pad/neural_networks_3