Papers with one-sample learning scheme
Batch IS NOT Heavy: Learning Word Representations From All Samples (P18-1)
Copied to clipboard
| Challenge: | Stochastic Gradient Descent with negative sampling is the most prevalent approach to learn word representations. |
| Approach: | They propose a method that uses batch gradient learning to generate word representations from all training samples. |
| Outcome: | The proposed method outperforms sampling-based methods on several benchmark tasks. |