Difference between revisions of "A One Hot Vector"
From Algolit
(→Algolit one-hot-vector scripts) |
(→Note that) |
||
| Line 88: | Line 88: | ||
Two one-hot-vector scripts were created during one of the Algolit sessions, both creating the same matrix but in a different way. To download and run them, use the following links: [https://gitlab.constantvzw.org/algolit/algolit/blob/master/algoliterary_encounter/one-hot-vector/one-hot-vector_gijs.py one-hot-vector_gijs.py] & [https://gitlab.constantvzw.org/algolit/algolit/blob/master/algoliterary_encounter/one-hot-vector/one-hot-vector_hans.py one-hot-vector_hans.py] | Two one-hot-vector scripts were created during one of the Algolit sessions, both creating the same matrix but in a different way. To download and run them, use the following links: [https://gitlab.constantvzw.org/algolit/algolit/blob/master/algoliterary_encounter/one-hot-vector/one-hot-vector_gijs.py one-hot-vector_gijs.py] & [https://gitlab.constantvzw.org/algolit/algolit/blob/master/algoliterary_encounter/one-hot-vector/one-hot-vector_hans.py one-hot-vector_hans.py] | ||
| − | =Note that= | + | ==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, idem.)] | "''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, idem.)] | ||
Revision as of 15:28, 25 October 2017
| Type: | Algoliterary exploration |
| Technique: | word-embeddings |
| Developed by: | Algolit |
Making a one hot vector
If this is our example sentence ...
"The algoliterary explorers discovered a multidimensional landscape made of words disguised as numbers."
... these are the 14 words we work with ...
a algoliterary as discovered disguised explores landscape made multidimensional numbers of the words .
... a single vector in a one-hot-vector looks like this ...
[0 0 0 0 0 0 0 0 0 0 0 0 0 0]
... and a full fourteen-dimensional matrix like this ...
[[0 0 0 0 0 0 0 0 0 0 0 0 0 0] a [0 0 0 0 0 0 0 0 0 0 0 0 0 0] algoliterary [0 0 0 0 0 0 0 0 0 0 0 0 0 0] as [0 0 0 0 0 0 0 0 0 0 0 0 0 0] discovered [0 0 0 0 0 0 0 0 0 0 0 0 0 0] disguised [0 0 0 0 0 0 0 0 0 0 0 0 0 0] explores [0 0 0 0 0 0 0 0 0 0 0 0 0 0] landscape [0 0 0 0 0 0 0 0 0 0 0 0 0 0] made [0 0 0 0 0 0 0 0 0 0 0 0 0 0] multidimensional [0 0 0 0 0 0 0 0 0 0 0 0 0 0] numbers [0 0 0 0 0 0 0 0 0 0 0 0 0 0] of [0 0 0 0 0 0 0 0 0 0 0 0 0 0] the [0 0 0 0 0 0 0 0 0 0 0 0 0 0] words [0 0 0 0 0 0 0 0 0 0 0 0 0 0]] .
... with one 0 for each unique word in a vocabulary, and a row for each unique word.
The following step is to count how often a word appears next to another ...
"The algoliterary explorers discovered a multidimensional landscape made of words disguised as numbers."
[[0 0 0 1 0 0 0 0 1 0 0 0 0 0] a [0 0 0 0 0 1 0 0 0 0 0 1 0 0] algoliterary [0 0 0 0 1 0 0 0 0 1 0 0 0 0] as [1 0 0 0 0 1 0 0 0 0 0 0 0 0] discovered [0 0 1 0 0 0 0 0 0 0 0 0 1 0] disguised [0 1 0 1 0 0 0 0 0 0 0 0 0 0] explores [0 0 0 0 0 0 0 1 1 0 0 0 0 0] landscape [0 0 0 0 0 0 1 0 0 0 1 0 0 0] made [1 0 0 0 0 0 1 0 0 0 0 0 0 0] multidimensional [0 0 1 0 0 0 0 0 0 0 0 0 0 1] numbers [0 0 0 0 0 0 0 1 0 0 0 0 1 0] of [0 1 0 0 0 0 0 0 0 0 0 0 0 0] the [0 0 0 0 1 0 0 0 0 0 1 0 0 0] words [0 0 0 0 0 0 0 0 0 1 0 0 0 0]] .
Algolit's one-hot-vector scripts
Two one-hot-vector scripts were created during one of the Algolit sessions, both creating the same matrix but in a different way. To download and run them, use the following links: one-hot-vector_gijs.py & one-hot-vector_hans.py
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, idem.)]
For more notes on this lecture visit http://pad.constantvzw.org/public_pad/neural_networks_3