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 1803.02839cs.CLcs.AIstat.ML
收藏

NLP中RNN和嵌入的涌现代数结构

The emergent algebraic structure of RNNs and embeddings in NLP

Excella Consulting AI Research(Excella Consulting人工智能研究(公司))

Sean A. Cantrell

摘要:

我们研究了一个单向GRU和词嵌入的代数和几何性质的训练端到端的文本分类任务。在词嵌入维度、GRU隐藏维度和GRU输出的线性组合上执行超参数搜索。我们的结论是,单词自然地嵌入到李群中,RNN形成了李群的非线性表示。呼吁这些结果,我们提出了一类新的递归类神经网络和词嵌入计划。

英文摘要:

We examine the algebraic and geometric properties of a uni-directional GRU and word embeddings trained end-to-end on a text classification task. A hyperparameter search over word embedding dimension, GRU hidden dimension, and a linear combination of the GRU outputs is performed. We conclude that words naturally embed themselves in a Lie group and that RNNs form a nonlinear representation of the group. Appealing to these results, we propose a novel class of recurrent-like neural networks and a word embedding scheme.

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。

↑