AI 与机器学习2017arxiv.org
将深度学习应用于新颖应用的最佳实践
Best Practices for Applying Deep Learning to Novel Applications
进阶 · 约 35 分钟读完前置知识：深度学习基础、机器学习项目流程、领域知识引入
导读
这篇 arXiv 报告总结了把深度学习应用到新领域/新任务时的实践经验与最佳做法，强调需要领域专家、合理准备数据，并尽量复用已有模型而不是让网络从零拟合。评论中特别提到文中“用网络学习对近似物理模型的修正”这一建议，认为这比直接端到端预测更可靠。
HN 关注点：其可操作的工程建议，讨论焦点落在领域专家/人员能力、生产落地，以及残差修正与领域知识的重要性。
HN 评论精选
这几条评论里：支持 1 质疑 1 补充 1 亲历 1
质疑最大落地风险是人员：所谓领域专家未必真专家，算法强的人常缺乏生产代码交付经验。
总体建议不错，但我认为它没覆盖一些主要陷阱，头号是人员。报告开头假设你拥有或能接触到应用领域的专家；以我经验，对目标读者群来说问题恰恰出在这里——不是没人，而是那个人并非真正 SME。“专家”在 ML 领域本就模糊，排除骗子后，ML 从业者在能力、领域专精度和具体应用能力上差异巨大。我与优秀 ML 人员合作最大能力缺口是：最懂机制/算法/方法的人往往很不擅长交付生产代码；不是没能力，而是时间多花在研究上，做事方式与应用工程师不同。本文是应用导向，所以这点很相关。ML 应用有很多雷区，本文从系统角度列了一些，但我经验中多数是人员和“文化”问题，不是别的领域的 culture fit。
AndrewKemendo查看英文原文
It's good general advice, but frankly I think it doesn't address some of the major pitfalls - namely the top one: personnel. In the very beginning it's stated: "In this report I assume you are (or have access to) a subject matter expert for your application." In my experience this is where it goes off the rails for most of the crowd that she is addressing. Not because they don't have someone, but because who they have isn't really a "subject matter expert." It's a muddy term anyway especially in the field of Machine Learning. Excluding for a moment the huksters and bold faced liars, within ML there is WIDE variance in competence, domain specificity and application specific capability within the field. The biggest capability gap that I have encountered when working with fantastic ML folks is that the ones that understand the mechanisms/algorithms/approaches best, are actually pretty terrible at delivering production code. That's not for lack of capability, it's simply because the bulk of their time has been spent in research - so they approach things very differently than application focused engineers. This is extremely relevant in this case because this is an application specific paper. There are a plethora of mine-fields in applications of ML, some of which are outlined here from a systems approach, but the majority of which are personnel issues in my experience, and "culture" issues - not to be confused with "culture fit" problems that exist elsewhere.
支持把它视为 Domingos 经典 ML 实践文章的 DL 版，认可其建议价值。
看起来是好建议。像是 Domingos 2012《A Few Useful Things to Know about Machine Learning》的更偏深度学习版本。
gwern查看英文原文
Looks like good advice to me. Like a more DL-focused version of "A Few Useful Things to Know about Machine Learning" http://www.datascienceassn.org/sites/default/files/A%20Few%20Useful%20Things%20to%20Know%20about%20Machine%20Learning.pdf , Domingos 2012
亲历复用已有模型并让网络学习残差修正，通常比端到端直接预测更可靠。
文中说：对物理高度近似的复杂过程，可以让网络直接输出结果，也可让网络学习对近似结果的修正；后者几乎肯定优于前者。这符合我的经验（计算化学博士）。当用一个强通用数学补丁去修已有模型时，应尽可能保留原模型；否则补丁很难拟合，甚至可能比原来的还差。哲学上也契合切斯特顿栅栏。
comicjk查看英文原文
> Let’s say you want to improve on a complex process where the physics is highly approximated (i.e., a “spherical cow” situation); you have a choice to input the data into a deep network that will (hopefully) output the desired result or you can train the network to find the correction in the approximate result. The latter method will almost certainly outperform the former. This aligns with my experience (computational chem PhD). When applying a strong, general-purpose mathematical patch to an existing model, use as much of the existing model as possible. Otherwise the patch will have a hard time fitting, and maybe be worse than what you started with. Philosophically, this also comports with my thinking (it's the modeling equivalent of Chesterton's Fence https://en.wikipedia.org/wiki/Wikipedia:Chesterton%27s_fence).
补充领域知识能约束 DL 结果落在物理现实内；材料科学中正尝试端到端与修正两种路线。
物理在状态上是确定性的且总趋向平衡，所以 DL 的新结果可能仍符合某个连续数学模型，却不稳定甚至不真实。领域专长能帮助以让结果落在现实边界内（并希望稳定）的方式准备数据。我正在材料科学目标上尝试两种方法，现在云端强大硬件便宜，很好奇结果。五年前这都不可能，我很感激兴奋。
DrNuke查看英文原文
Physics is deterministic in states and always tends towards an equilibrium, so novel results from DL may still fit some continuum math model without being stable or even real. Domain expertise, on the other hand, helps prepare data for ML algos in such a way that results will come (or not) within the boundaries of reality and hopefully stability. I am trying both approaches for some materials science goals of mine and am curious to see what happens, now that powerful hardware is cheap enough on the cloud to put some ideas to work. Side point is all this was just impossible five years ago, so I am grateful and excited to have this opportunity.
历次讨论
- 2017Hacker News 讨论帖300 分 · 17 评
同领域还值得读
DeepSeek-R1：通过强化学习激励 LLM 的推理能力
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL
研究如何用强化学习激励大语言模型的推理能力，提出 DeepSeek-R1，并探讨 RL 训练对模型推理表现的影响。
AI 与机器学习2025 · 1,351 分 · 1,056 评 · 引用 30进阶 · 约 80 分钟编辑精选1-bit LLM 时代：面向低成本计算的三值参数
The Era of 1-bit LLMs: ternary parameters for cost-effective computing
提出用三值参数 {-1, 0, 1} 替代 LLM 中的浮点权重，即所谓“1-bit LLM”。
AI 与机器学习2024 · 1,040 分 · 447 评 · 引用 19进阶 · 约 45 分钟编辑精选DeepSeek-V3.2：推进开放大语言模型的前沿
DeepSeek-v3.2: Pushing the frontier of open large language models
这是 DeepSeek-V3.2 的技术报告，延续了 DeepSeek 开源权重模型的路线，目标是让开放模型继续逼近甚至追平前沿闭源模型。
AI 与机器学习2025 · 982 分 · 465 评进阶 · 约 90 分钟编辑精选通过自我对弈与通用强化学习掌握国际象棋和将棋（AlphaZero）
Mastering Chess and Shogi by Self-Play with General Reinforcement Learning
DeepMind 的 AlphaZero 论文：用通用强化学习算法（自我博弈 + MCTS + 神经网络），从零开始仅数小时训练便在国际象棋和将棋上击败了世界顶级引擎 Stock…
AI 与机器学习2017 · 539 分 · 270 评 · 引用 13进阶 · 约 90 分钟编辑精选用强化学习发现更快的矩阵乘法算法
Discovering faster matrix multiplication algorithms with reinforcement learning
论文提出 AlphaTensor，用强化学习搜索矩阵乘法算法。
AI 与机器学习2022 · 518 分 · 111 评专业 · 约 90 分钟编辑精选