arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2104.03305cs.LG
收藏

具有优化熵编码的学习变换压缩

Learned transform compression with optimized entropy encoding

Geneva School of Business Administration(日内瓦商学院);HES-SO University of Applied Sciences of Western Switzerland(瑞士西部应用科学与艺术大学)

Magda Gregorová, Marc Desaules, Alexandros Kalousis

摘要:

我们考虑学习变换压缩的问题,在这里我们学习离散码的变换和概率分布。我们利用软松弛的量化操作,以允许反向传播的梯度和采用矢量(而不是标量)量化的潜在代码。此外,我们在代码概率分配中应用类似的松弛,从而能够直接优化代码熵。据我们所知,这种方法是完全新颖的。我们进行了一组概念验证实验,证实了我们的方法的效力。

英文摘要:

We consider the problem of learned transform compression where we learn both, the transform as well as the probability distribution over the discrete codes. We utilize a soft relaxation of the quantization operation to allow for back-propagation of gradients and employ vector (rather than scalar) quantization of the latent codes. Furthermore, we apply similar relaxation in the code probability assignments enabling direct optimization of the code entropy. To the best of our knowledge, this approach is completely novel. We conduct a set of proof-of concept experiments confirming the potency of our approaches.

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。

↑