 |
ML Newsnew | past | best | rss | submit | about |
| FastText: Training text classifiers on 1b words in less than 10 minutes (arxiv.org) | |
| 1 point by jbaiter on Jul 7, 2016 | hide | past | pdf | discuss on HN | |
|
In plain words: A simple text classifier turns each document's words into one summary vector and scores classes directly, instead of stacking many layers like deep networks. It matched deep models' accuracy while training on over a billion words in under ten minutes on a normal CPU. Abstract · Bag of Tricks for Efficient Text ClassificationThis paper explores a simple and efficient baseline for text classification. Our experiments show that our fast text classifier fastText is often on par with deep learning classifiers in terms of accuracy, and many orders of magnitude faster for training and evaluation. We can train fastText on more than one billion words in less than ten minutes using a standard multicore~CPU, and classify half a million sentences among~312K classes in less than a minute. | |
|
add comment on HN
Also discussed: Jul 2016 (169 points, 15 comments)
|