“I "woke up" to this sort of approach to understanding deep neural networks when I read this paper”
An exact mapping between the Variational Renormalization Group and Deep Learning
证明变分重整化群与基于 RBM 的深度学习之间存在精确映射
作者指出深度学习与理论物理中的重整化群(RG)密切相关,RG 是一种迭代的粗粒化方案,可在不同尺度上提取相关特征。他们构造了 Kadanoff 提出的变分重整化群与基于受限玻尔兹曼机(RBM)的深度学习架构之间的精确映射,并用一维和二维最近邻 Ising 模型加以说明,提示深度学习算法可能在用一种广义的类 RG 方案从数据中学习相关特征。
它常被引用来说明深度学习与统计物理之间的深层联系,有人说它让自己真正理解了神经网络,也有人觉得它不过很可爱、并没有把问题变成物理,并质疑论据与对比散度训练的比较不足。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及“PS. You will enjoy the following papers too:”
“I don't see any comparison between the approximations that physicists normally use versus the contrastive divergence for training RBM-based networks, or evidence that their results are more similar in practice than any other technique.”
“It's cute, but it does not turn the subject into physics.”
“Additionally, a paper that has everyone excited about deep connections between the mathematical analysis of physical systems and the hierarchical feature learning paradigm speaks of the connection in terms of the Renormalization Group [2].”
“In particular, I've spent the last several months in trying to understand the following two papers in increasing levels of depth”
看全部 9 条 →
“there are also many researchers poking and prodding at deep learning theory [1, 2].”
“it's quite plausible that at the fringe/edge there is research being done that involves differential geometry - hell there's this paper”
“The pyramid strategy loosely tracks with renormalization group theory, which has been formally studied for years as a method of interpreting machine learning models:”
- Deep Learning and Renormalization Group Thery ↗kevinlu310 · 2014-12-08 · 10 分
- An Exact Mapping Between the Variational Renormalization Group and Deep Learning ↗standard4317 · 2015-02-05 · 1 分
- An Exact Mapping Between the Variational Renormalization Group and Deep Learning ↗evanb · 2015-07-31 · 30 分
- Exact mapping between Variational Renormalization Group and Deep Learning (2014) ↗_qc3o · 2016-11-30 · 97 分