“The most persuasive argument I've seen for this view -- our view -- is laid out in this paper by Lin, Tegmark, and Rolnick”
Why does deep and cheap learning work so well?
用物理视角解释深度学习为何有效
论证深度学习的成功不仅取决于数学,也取决于物理:实际关心的函数类往往只需指数级更少的参数就能近似,即所谓廉价学习。对称性、局部性、组合性等物理中常见的性质会转化为极简单的神经网络;当数据由某种层级式过程生成时,深度网络比浅层网络更高效。作者用信息论形式化这些论点,讨论其与重整化群的关系,并证明了若干不可展平定理。
评论者把它当作解释深度学习为何有效的最有说服力的物理视角之一:自然数据的分布受物理定律约束,因而容易学习。
学习路径
读这篇之前,按顺序读书架上的 1 篇 · 参考文献来自 OpenAlex,顺序和要点由 Claude Code 排出候选里只有一篇用神经网络求解量子多体问题的论文与目标论文有较明确的联系,所以路径很短。
- 1
用神经网络表示量子多体波函数,是物理与神经网络结合的实例,与目标论文的物理视角呼应。
读抓用受限玻尔兹曼机表示波函数的想法;训练细节略读。
- 2Why does deep and cheap learning work so well?
用物理视角解释深度学习为何有效。
读抓对称性、局部性、组合性如何使网络只需指数级更少的参数,以及与重整化的类比。
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“you might also enjoy really interesting paper coauthored by the same person”
“Another interesting explanation for deep learning's success in the physical world:”
“there was a paper by Max Tegmark on HN yesterday with some new ideas and results, but I haven't had time to read it yet.”
“He comments on the video linking to two papers on arXiv which relate to the material in the lecture: Why does deep and cheap learning work so well? and [OTHER_PAPER]”
“That does not appear to be true in theory (due to the laws of physics, e.g., see Why does deep and cheap learning work so well? ) nor in practice”
看全部 12 条 →
“3. Why does deep and cheap learning work so well? Physics and Deep Learning are well entwined.”
“there are also many researchers poking and prodding at deep learning theory [1, 2].”
“Perhaps this paper provides an explanation?”
“this concentration of natural data distributions might be a result of the laws of Physics of the universe in which we happen to live: Why does deep and cheap learning work so well?”
“Tegmark & Lin's discussion of how deep learning maps onto the physical world: Why does deep and cheap learning work so well? [1][2]”
“but perhaps it could be of interest to you. Why does deep and cheap learning work so well?”
- Why does deep and cheap learning work so well? ↗okket · 2016-08-31 · 38 分
- Why does deep and cheap learning work so well? ↗mauliknshah · 2018-05-11 · 434 分
- Why does deep and cheap learning work so well? (2017) ↗max_ · 2019-08-15 · 16 分
- ML paper – Why does deep and cheap learning work so well? ↗programd · 2019-10-24 · 4 分