AI 与机器学习2019paradise.caltech.edu
两个神经元就足以骑自行车 (2004)
It Takes Two Neurons to Ride a Bicycle (2004)
入门 · 约 45 分钟读完前置知识：经典力学基础、控制理论基本概念
导读
这篇 2004 年的论文（作者 Matthew Cook）研究无人自行车能否用极简控制器保持平衡。作者构建了一个精确保持角动量与动能的刚体物理模拟器来模拟自行车，结果发现一个仅含两个神经元的简单电路就足以控制自行车骑行，效果甚至比用键盘操作的人类更好。作者后来进一步将控制器简化到单个神经元乃至普通线性反馈系统。
HN 关注点：标题极具反直觉冲击力（'两个神经元就能骑自行车'），HN 读者围绕控制器的极简性、自行车作为非完整系统的控制难度以及实验可复现性展开讨论。
HN 评论精选
这几条评论里：质疑 1 补充 1 亲历 2
亲历作者本人现身说法：双神经元控制器优于人类，且模拟器精确保持守恒量是关键
这是 NIPS 2004 的演示项目。我创建了一个假设需要学习高阶概念的环境：先做了一个通用刚体物理模拟器来模拟自行车供学习智能体学习骑行。但正如他人在其他场景发现的那样，多数环境并不需要高阶概念——这里一个简单的双神经元电路就足以控制自行车，甚至比用键盘的人类更好。物理模拟器本身就是个大工程：我研究了刚体力学（比多数人想象的复杂），并特意设计模拟器同时精确保持角动量和动能，因为之前做量子力学模拟的经验表明保持守恒量对结果精度至关重要。后来我把控制器简化到一个神经元，再到更简单的线性反馈系统……
floatingatoll查看英文原文
NIPS 2004 Demonstrations: https://web.archive.org/web/20041204162656/http://www.nips.cc/Conferences/2004/Demonstrations/Demonstrations.php Virtual Bicycle — Matthew Cook, California Institute of Technology: http://www.paradise.caltech.edu/cook/papers/index.html In this work I created an environment in which I assumed learning of higher order concepts would be necessary. I created a general purpose physics-based hinged rigid body simulator and used it to simulate a bicycle for a learning agent to learn to ride. But as others have found in other contexts, most environments do not require higher order concepts, and in this case it turned out that a simple two neuron circuit was already sufficient for controlling the bicycle, indeed, better than a human using the keyboard. The physics simulator was a significant project in itself: I studied rigid body mechanics (which is more complex than most of us realize) and designed the simulator explicitly so as to simultaneously exactly preserve both angular momentum and kinetic energy, as previous simulations of mine (in quantum mechanics!) had shown me that preserving conserved quantities can be crucial for getting accurate results. The simulator works nicely, and I later read in Sam Buss's 2001 paper, Accurate and Efficient Simulation of Rigid Body Rotations, that my precautions were well warranted. Later I was able to reduce the controller to just one neuron, and then to an even simpler plain linear feedback system, confirming the finding that many real problems are best solved by a hack. The simulator has been useful in other projects since then. Virtual Bicycle Download Page: http://www.dna.caltech.edu/~cook/
亲历亲历者指出自行车是非完整系统，保证相同初始条件是训练最大难点
'如何确保每次起始条件完全相同？'是那个讨论帖里最好的问题，也是我们用神经网络训练自行车平衡时最大的挑战。自行车难搞是因为它是非完整系统（non-holonomic），不只是找线性化运动方程那么简单。
boxcardavin查看英文原文
"How would you ensure having the exact same starting conditions each time?" is the best question in that thread, and this was the biggest challenge for us in training our bikes to balance using a NN. Bikes are hard because they're a non-holonomic system, it isn't just a matter of finding linearized equations of motion.
质疑质疑标题夸大，实际内容只是无人自行车的模拟路径
标题党了……这实际是'无人自行车带滚动起动的模拟路径'。
tomglynch查看英文原文
Title gore... What this really is, is a Simulated path of an unmanned bicycle with a rolling start.
历次讨论
- 2019Hacker News 讨论帖146 分 · 25 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