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 2002.05095cs.LGstat.ML
收藏

生成网络的压缩学习

Compressive Learning of Generative Networks

ISPGroup, ICTEAM, UCLouvain(ISPGroup, ICTEAM, 比利时天主教鲁汶大学(UCLouvain));Louvain-La-Neuve - Belgium(比利时鲁汶-拉讷夫)

Vincent Schellekens, Laurent Jacques

摘要:

生成网络隐式地从它们的采样中近似复杂的密度,具有令人印象深刻的精度。然而,由于现代数据集的巨大规模,这种训练过程通常在计算上是昂贵的。我们将生成式网络训练引入到最近的压缩学习框架中:我们通过首先将大规模数据集作为单个草图向量进行单次压缩来减少它们的计算负担。然后,我们提出了一个成本函数,它近似于最大均值离散度量,但只需要这个草图,这使得优化具有时间和内存效率。

英文摘要:

Generative networks implicitly approximate complex densities from their sampling with impressive accuracy. However, because of the enormous scale of modern datasets, this training process is often computationally expensive. We cast generative network training into the recent framework of compressive learning: we reduce the computational burden of large-scale datasets by first harshly compressing them in a single pass as a single sketch vector. We then propose a cost function, which approximates the Maximum Mean Discrepancy metric, but requires only this sketch, which makes it time- and memory-efficient to optimize.

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。

↑