arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2210.13554cs.LGcs.AIcs.CV
收藏

权重固定网络

Weight Fixing Networks

University of Southampton(南安普顿大学)

Christopher Subia-Waud, Srinandan Dasmahapatra

摘要:

深度学习模型的现代迭代包含数百万(数十亿)个唯一参数,每个参数由一个b位数字表示。压缩神经网络的流行尝试(如修剪和量化)表明,许多参数是多余的,我们可以删除(修剪)或用少于b位的(量化)来表示,而不会影响性能。在这里,我们希望在最小化网络信息内容方面走得更远。而不是通道或逐层编码,我们期待无损全网络量化,以最小化网络中唯一参数的熵和数量。我们提出了一种新的方法,我们称之为权重固定网络(WFN),我们设计它来实现四个模型结果目标:i)非常少的唯一权重,ii)低熵权重编码,iii)适合节能版本的硬件乘法的唯一权重值,以及iv)无损任务性能。其中一些目标是相互冲突的。为了最好地平衡这些冲突,我们结合了一些新的(和一些常用的)技巧;一个新的正则化术语,(i,ii)将聚类成本视为相对距离变化(i,ii,iv),以及关注整个网络重用权重(i,iii)。我们的Imagenet实验证明了无损压缩使用比SOTA量化方法少56倍的唯一权重和低1.9倍的权重空间熵。

英文摘要:

Modern iterations of deep learning models contain millions (billions) of unique parameters, each represented by a b-bit number. Popular attempts at compressing neural networks (such as pruning and quantisation) have shown that many of the parameters are superfluous, which we can remove (pruning) or express with less than b-bits (quantisation) without hindering performance. Here we look to go much further in minimising the information content of networks. Rather than a channel or layer-wise encoding, we look to lossless whole-network quantisation to minimise the entropy and number of unique parameters in a network. We propose a new method, which we call Weight Fixing Networks (WFN) that we design to realise four model outcome objectives: i) very few unique weights, ii) low-entropy weight encodings, iii) unique weight values which are amenable to energy-saving versions of hardware multiplication, and iv) lossless task-performance. Some of these goals are conflicting. To best balance these conflicts, we combine a few novel (and some well-trodden) tricks; a novel regularisation term, (i, ii) a view of clustering cost as relative distance change (i, ii, iv), and a focus on whole-network re-use of weights (i, iii). Our Imagenet experiments demonstrate lossless compression using 56x fewer unique weights and a 1.9x lower weight-space entropy than SOTA quantisation 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。

↑