arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1909.01264cs.LGstat.ML
收藏

压缩词嵌入的下游性能研究

On the Downstream Performance of Compressed Word Embeddings

Stanford University(斯坦福大学)

Avner May, Jian Zhang, Tri Dao, Christopher Ré

摘要:

压缩词嵌入对于在内存受限的环境中部署NLP模型非常重要。然而,理解是什么使得压缩嵌入在下游任务上表现良好是具有挑战性的-现有的压缩质量度量通常无法区分表现良好的嵌入和表现不佳的嵌入。因此,我们提出了特征空间重叠分数作为一种新的措施。在线性和逻辑回归的背景下,我们将特征空间重叠分数与下游性能联系起来,通过在此分数方面为压缩嵌入开发泛化边界。然后,我们表明,我们可以下界的特征空间重叠分数的一个简单的均匀量化压缩方法,有助于解释这种方法的强大的经验性能。最后,我们表明,通过使用的特征空间重叠得分作为选择标准之间的嵌入从一个代表性的一组,我们压缩,我们可以有效地识别性能更好的嵌入与高达$2\times$较低的选择出错率比下一个最佳的压缩质量的措施,并避免训练模型的成本为每个感兴趣的任务。

英文摘要:

Compressing word embeddings is important for deploying NLP models in memory-constrained settings. However, understanding what makes compressed embeddings perform well on downstream tasks is challenging---existing measures of compression quality often fail to distinguish between embeddings that perform well and those that do not. We thus propose the eigenspace overlap score as a new measure. We relate the eigenspace overlap score to downstream performance by developing generalization bounds for the compressed embeddings in terms of this score, in the context of linear and logistic regression. We then show that we can lower bound the eigenspace overlap score for a simple uniform quantization compression method, helping to explain the strong empirical performance of this method. Finally, we show that by using the eigenspace overlap score as a selection criterion between embeddings drawn from a representative set we compressed, we can efficiently identify the better performing embedding with up to $2\times$ lower selection error rates than the next best measure of compression quality, and avoid the cost of training a model for each task of interest.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑