How I Learn 我的学习心得
Thoughts
CN-blogs
I record my thoughts and ideas about math, physics, CS and many more here (constantly updating).
Directly exported from Notion, may contain some formatting issues.
1 General Spirits
No. | Date | Ideas |
---|---|---|
41 | 2025-8-17 | Learn by terminology: 我曾经看轻术语的重要性, 事实上每个术语都有自己的一套哲学在里面, 你需要分辨哪些术语是必要的 (总线、分布式, etc), 哪些是不必要的. 当一个领域所有的资料都以术语的方式呈现时, 你必须先掌握所有的术语的意义, 然后才能理解这些资料. So enjoy the art of terminologing! |
40 | 2025-8-3 | 检验一流智力的标准, 就是在头脑中同时存在两种截然相反的想法时仍能保持行动能力. – F. Scott Fitzgerald |
39 | 适当的不合理的假设 (其实就是帮助建立一个模块) 能够帮助快速进展. 一个不合理的假设没有你想象的那么不合理和顽固, 知识本身你都会忘 let alone 不合理的假设. | |
38 | 一个学科形成时的 Paradigm Shift 范式转变: 一个概念的形成需要一条完整的动机链, 其中包含很多个关键的 Paradigm Shift, 学会一个概念很容易, 只需要找到属于你自己的动机链和这些 Paradigm Shift, 你就学会了. 但是也很复杂, 因为这些都要你自己去寻找. 举个栗子: 白云和挂科有什么关系? 只要三个 Paradigm Shift: 白云 -> White Cloud -> WC -> CW -> 挂科. 虽然每个 Paradigm Shift 都是相对显然的. 我不认为一个 new comer 能够「理解」数学里的 Set, Morphism, Limits in categories 抑或者控制系统里的 Phasor, Transfer function, etc. 我也不认为科普文或视频对外行 (或内行) 有任何的实质性帮助. | |
37 | The imageries and the method by which we are storing it (knowledge) all and the way we think about it, could be really, if we get to each other’s head, entirely different. – Richard Feynman | |
36 | 2025-3-8 | Remember what Richard Feynman thought when he went to the dentist? He thought about the entire process of generating the electricity, from the big wheel in the factory, the transmission lines, to the rotor he’s looking at right now. All of these are one realization of Maxwell’s Equations. Isn’t that amazing? 如果你能这么思考一个领域的内容, 你才真正理解了它! 不过这并不难, know how things work只是人类的本能, 只不过这种本能已经在无数的教育框架和所谓模型中渐渐变成了一种邪教, 使人不加思考地相信他们, 以至于搞懂他们都会被表扬. |
35 | 2025-3-8 | 为什么看懂代码能够让你快速学会一个领域的知识(如ML)? 是因为当你看懂程序并且该程序能够跑出正确的结果的时候, 你的脑袋中就有了Ah, this is nothing but …的一句话. 看懂代码是让你得出这句话的很有效的方法, 但或许并不是最高效的方法. 亦或者有更高效率的看懂代码的方法? |
34 | 2025-3-8 | 理解了一个问题也就解决了这个问题. |
33 | 2025-3-6 | 阻碍你进行科研的原因之一是你不清楚其问题的抽象程度, 而大多数问题的抽象程度比较高, 不需要你进行很多原理上的了解, 将它变成纯粹的数学问题进行思考. 比如信道的模型你就当作一个“有趣”的结构, 讨论其正确性的工作有别于你现在研究的问题. |
32 | 2025-1-4 | 如果把一句话里的关键词换成其他很多词也成立,那么这句话本身就是废话。 |
31 | 2024-12-21 | 理解一个东西大致分为:弄懂它讲的是什么、对它建立认知和充分理解。前者需要查阅大量的资料和克服恐惧心理,后者需要充分的独立的自主思考时间。本科科研的本质就是耗费你大量的时间来完成前一个步骤,后一个步骤将很慢地进行。所以有一个shortcut就是通过其他同学来告诉你它是什么、用来干什么的、细节是怎样的。这样你就省去了大量读懂它的时间。 |
30 | 2024-12-4 | When you realize that an idea or a subject could’ve been reinvented if it is lost totally, this feeling is very important because you already know the direction of science. |
29 | 2024-12-4 | Once you know the motivation and the cool aspects of something (like the spirit of phase space), you already possess all of that subject (Everything is just propositions, which will naturally arise with a little bit extra proofs and paper-and-pencil thing. If your brain has enough speed and memory, the whole subject can even appear in your head all of a sudden! Those “propositions” are just used to correct imperfect intuitions!), and you could have brought about the whole scenery yourself, provided that you just know the initial spirit! A new subject is formed exactly when a cool idea is proposed and that idea happens to be true in the most cases and suddenly has rich application for no reason. The rest of the years are just years of proofing and proposition derivations. In other words, a subject is formed right after the axioms are written! By the way, this is how I learn Hamiltonian physics. |
28 | 2024-11-25 | 人类学习的特点:只需要一个例子、或一个类比, 而不需要两个 |
27 | 2024-11-23 | Reinterpreting existing things that no one states like that before is invention. |
26 | 2024-11-18 | 当你知道题目是考什么的后,他就变得不再神圣。 |
25 | 如果你想快速地学好工程中的某个领域,你可以(而且需要)囫囵吞枣,但是你得记得你吞了几个枣。 | |
24 | 2024-11-4 | Doing the same thing, what’s going on in people’s heads can be completely different! Which means it’s totally possible to fall in love with something that you suck at by change what’s going on in your minds. (for example, programming as another way of thinking, or think about its mathematical essence.) |
23 | 2024-10-5 | 学习能力指学任何东西的能力。你一定会低估任何这个词所包含的事情,就想你一定会低估阿列夫1比阿列夫0大的程度、一个范畴中可以包含哪些东西一样。 |
22 | Recognition has a hierarchy. | |
21 | There are two things we should never have to chase: true friends, true love. | |
20 | 2024-10-2 | Cheap Dopamine. |
19 | 2024-9-7 | 一本书、一系列视频中真正有用的观点其实也就那么几个, take it easy! |
18 | 2024-6-25 | 你真的要严格区分your thoughts和政治经济学吗? |
17 | 2024-6-20 | Now I’m talking like a psychologist, and you know, I know nothing about this. —Richard Feynman |
16 | 2024-6-20 | The problem with the world is that intelligent people are full of doubts and the stupid ones are full of confidence. —Charles Bukowski |
15 | 2024-6-20 | You don’t have to be focused while studying, you just need to be relaxed, chilled, and enjoy what you are studying. —YouTube comment |
14 | 2024-6-20 | Maybe it’s not that the way to learn math is different from computer science, it’s that you are more interested in math! Math is for me a field where you have to run before you can walk. |
13 | “If all of human knowledge, everything that’s known, is believed to be an enormous hierarchic structure, then the high country of the mind is found at the uppermost reaches of this structure in the most general, the most abstract considerations of all. Few people travel here. There’s no real profit to be made from wandering through it, yet like this high country of the material world all around us, it has its own austere beauty that to some people make the hardships of traveling through it seem worthwhile.” — ROBERT PIRSIG, Zen and the Art of Motorcycle Maintenance (1974) | |
12 | Find differences first, then find similarities. | |
11 | You cannot experience another realm without stepping out of your current zone. | |
10 | 学习的过程本身就是以片面、孤立的视角看问题。 | |
9 | 你的精神寄托可以是马勒、柴可夫斯基,可以是工作、学习,也可以是星辰大海,唯独不能是人,活着的人。——改编自罗翔 | |
8 | 从你写下以上这些文字以及实际做出的改变,到对记笔记原则的思考与实际行动,反映出了人真正转变固有思维方式与“信仰”和实际行动比你想象的更难改变. | |
7 | 记住专有名词还是有必要的 (它不是目的,而是手段), 特别是你想深入理解一个概念之前, 它不仅加速了学习、理解的速率, 给大脑带来间歇性刺激, 也减小了人脑对知识的过拟合. | |
6 | 自发的、清晰表述的问题能够给自己带来不少的潜在价值. ——王木头 | |
5 | 让我们的学习有进展的, 并非正确的知识, 而是能够被自己认可的解释. | |
4 | 我们登上并非我们所选择的舞台, 演出并非我们所选择的剧本. | |
3 | Okay, why can’t you learn something new? You are the person who likes systematic learning, which does have benefits, but also has fatal weaknesses: You can’t learn fast. However, the reason behind that is you don’t have trees of that particular field in your head, you don’t know where to put newcomers. So build your trees first! | |
2 | If you are in the state of learning, you are not in the status of worrying. (NOTE Learning is generalized!) | |
1 | 知识结构决定学习方法. (如单片机、通信网络这种课非常有别于数学或物理课,其知识有明显的网状特点,应该采用逐步加深而不是扩散的学习方法, 其中逐步加深的过程中出现 “不是很明白” 或 “没有学习的感觉” 的情况是正常的, 要靠时间去磨专业名词等等, 但是一定要花时间去看! 另外这种学科中的实例比定义更有价值!) |
2 How to Learn Math
No. | Date | Ideas |
---|---|---|
184 | 2025-6-30 | 我发现好像名词的定义不再神圣了. 重要的是一个概念的动机和结论. |
183 | 2025-5-10 | currying 是 “\(\text{Hom}\) 与 \(\otimes\) 的伴随性”的程序员版本! |
182 | 2025-5-7 | 你真的懂得方程思想吗 – 大胆假设未知的量或函数, 然后列出他们满足的方程!!!!! |
181 | 2025-4-7 | When visualizing things, you should be clear what space you are looking at, it could be a different space than what you think. (For example, in stereographics projection, what you are looking at is the projective space, not the common complex space.) |
180 | 2025-3-30 | You should always think about a homomorphism whenever there is a quotient. |
179 | 2025-3-30 | The quotient of a group \(G\) by a normal subgroup \(N\) is a means of organizing the group elements according to how they fail—or don’t fail—to satisfy the property required to belong to \(N\). Think about \(\mathbb{Z}/5\mathbb{Z}\). |
178 | 2025-3-10 | The second order derivative commutes is called the Levi-Civita connection! |
177 | 2025-1-31 | I would rather have questions that can’t be answered than answers that can’t be questioned. —Richard Feynman |
176 | 2025-1-31 | Science is the belief in the ignorance of experts. |
175 | 2025-1-31 | Physics is to math what sex is to masturbation. Sure it might give some practical results but that’s not why we do it. — Richard Feynman |
174 | 2025-1-5 | 之前你经常听到群描述了某种对称性,今天通过strongly symmetric channel的概念终于理解了这句话。对称性就是某个群作用到某个集合上,关于这个集合的某个函数不变(G-invariant)。 |
173 | 2025-1-5 | Gaussian 分布在连续意义上是一种“最混乱”的分布,Uniform 分布在离散意义上是一种“最混乱”的分布。 |
172 | 2025-1-5 | 你要分清哪些是数学上tedious的证明,哪些是关键的思想。往往一个理论的思想只有一两个,其他的只是因为人脑的内存不足以至于不能一眼就看出来结论(所以如果人脑推理速度很快的话,这些结论没有很多的价值,比如熵是凸函数就是这样一个结论)。 |
171 | 2025-1-3 | \((u,v)\) 在某些情况下可以写成 \(u+iv\), 其中u和v是某个向量空间中的元素。 |
170 | 2025-1-2 | Hermitian 性总是和实数性联系在一起:\[x(t) \in \mathbb{R} \iff H(-f) = H^*(f) \\ S(f) \in \mathbb{R} \iff R(-t) = R^*(t)\] 但是如何统一 \(f\) is \(\text{P.D.} \iff\) induced matrix F is Hermitian \(\iff\) spectrum of \(F\) is real? |
169 | 2025-1-2 | 傅立叶变换是伸缩的积分变换;卷积是平移的积分变换。 |
168 | 2025-1-2 | Hilbert transform is unique in some sense. (The only one singular integral in dimension 1) |
167 | 2025-1-2 | 当你觉得你学的某个知识是某种东西的特例的时候,你将感到很快乐。 |
166 | 2025-1-1 | 为什么我们不能将一个概念向别人解释地很清楚?一个很重要的点是我们会忘记当时我们怎么学会的(包括参考资料、视频等),当我们向别人解释时,我们大脑里呈现的是那个例子,但是我们并没有把这个例子告诉对方 |
165 | 2025-1-1 | entropy是分布的函数而不是随机变量的函数 (similar to moments) (The random variable does not contain the information on its own distribution, so it is impossible to define the entropy of the distribution —or any moments of the distribution — purely as a function of the random variable!) |
164 | 2024-12-28 | 定理的条件是如何诞生的:先是有一个新颖的观点,然后通过直觉驱动的不严谨的推导,得出一个很有可能正确的结论(这里需要正确的数学审美素养)。然后为了论述这个结论,加上了很多分析限制(比如局部紧的)。产出论文时,把这些限制写在论文开头来防止你很快地看懂你浅陋的思维。 |
163 | 2024-12-26 | Hermitian 跟实数总是有一定的关系!比如 Hermitian 矩阵的特征值值一定是实数,实信号的 Fourier 变换具有 Hermitian 性,那内积、相关函数好像也具有 Hermitian 性,其“实数”性体现在哪呢?跟 Positive definite 有关? |
162 | 随机信号过的不是线性系统,而是线性系统在 \(\text{Hom}(\Omega, -)\) 函子下的像!! | |
161 | 2024-12-17 | 如果你要求一个函数或数列,不要想办法求出具体的某一项,而是先写出整个函数或数列要满足的方程!! |
160 | 2024-12-16 | When you have uncertain things, just think of one of the possible explanations and believe it is true. |
159 | 2024-12-14 | “Kernel” also means weighted sum (or integral). |
158 | 2024-12-3 | Fourier transform is a group algebra homomorphism between the group algebra and endomorphisms of the direct sum of irreducible representations |
157 | 2024-12-2 | Whenever a mathematical statement looks magical, it never actually is. |
156 | 2024-11-20 | Dynamical system is a concept in math, not physics. (But yeah you are right if you disagree) There are two versions of ergodicity, using measuring-preserving transformations, and in the language of stochastic process. |
155 | 2024-11-20 | Symplectic manifolds provide the generalized setting for classical mechanics, where the motion of a mechanical system is described by a geodesic. |
154 | 2024-11-19 | free这个概念和形式定义相关。形式定义是切入一个概念的重要方法。 |
153 | 2024-11-19 | 线性变换的性质已经完美地融合在了矩阵的运算中,但是这种融合过于完美,以至于大多数人都不能从其中抽取出一条适合于人类大脑线性思维的学习轨迹。(矩阵运算的各个理解方式都对应了一些及其重要的性质!) |
152 | 2024-11-1 | You should be aware of how we model real-world things. (how do we think of information sources? —Stochastic processes; how do we imagine channel? —Just a little conditional probability matrix.) |
151 | 2024-10-23 | It’s a funny thing about square matrices: they can represent several very different things, especially two–linear transformations from a space to itself and bilinear forms. |
150 | 2024-10-20 | 数学家想用最高的抽象性描述尽可能多的事物,然而最后描述出来的却没想象的那么多,这不是数学的问题,而是挖掘的问题(我在避免用application这个词,因为人类已经对它麻木了) |
149 | 2024-10-20 | 分布的函数和随机变量的函数是不一样的。 |
148 | 2024-10-17 | It seems that some genius people talking about a very visual problem (say a random walk), what’s going on in their heads is totally different from what he describes. There will always be some abstraction and soul of this particular subject in his head simultaneously, but it’s hard to speak them out. (For example, a sum in a random process is not just adding numbers but random variables.) The described version is always a special case of a more generalized problem. |
147 | 2024-10-15 | 很多概念的学习顺序其实没你想的那么重要,不同的教材中介绍他们的顺序都不一样,你要做的不是去找一个通用的学习顺序,而是去揣摩作者这样写的目的。(比如KL散度和Mutual information 哪个先介绍好一点?)但是往往概念有很多等价的定义,有必要清楚你认为哪一个是出发点! |
146 | 2024-10-14 | 一本好的书应该给你足够的心理准备,即使他没有给你,你自己也应该做好某种心理准备来应对可能的难于理解的问题,关键词:“只不过是”. |
145 | OK there are several things in math that make you feel uncomfortable: repetitive proofs, intimidating notations, lack of motivation for new concepts, etc. Now stop worrying and fix them! | |
144 | Changing perspective is both essential in math and life. | |
143 | 你要相信你独立发现的数学结构之间的相似性一定是一门学问的有效动机! | |
142 | 2024-10-5 | 理解一个概念只需要一个例子!你一定会低估你理解一个概念所需要的例子数量!但是这个例子需要足够的surprising, 并且多余的结构越少越好! |
141 | 确实你有必要刻意地区分一下随机变量、观测值、样本的区别了。 | |
140 | 我不认为说一个东西没有意义有任何的好处。 | |
139 | 2024-9-25 | 如何听课?“这个时刻的值是随机的”,当老师说这句话时,没必要去为严谨性去反驳,当我自己尝试严谨地表达我头脑中理解的画面时,我发现只能这么表达了。然而这句话的背后蕴含了大量对学科本质及思想的理解。 |
138 | 2024-9-25 | 是谁改变了你对数学的看法?Feynman. (你真的要区分应用数学和纯数吗,what is understanding?, 数学真的是过剩的,任何一个将纯数理论用在其他领域的人都是里程碑式的人物,比如纤维丛,然而他们只是理解了这个数学概念,没有那么伟大,他们也不会认为自己很伟大、亦或是有多天才) |
137 | 2024-9-21 | GPT在学习数学中的作用——强化萌生的观点。 |
136 | 2024-9-18 | 认识到你学的东西知识一个宏图的局部是一件很难的事情。(比如香农熵只是One of the methods) |
135 | 当你在学所谓的“严格数学”时,很容易就忘记这样做背后的指导思想,它是什么东西的抽象?定理填补了哪个直觉漏洞?代数是直觉的量化、是为了避免大脑内存溢出时做的信息压缩。万不得已才做代数推导。 | |
134 | 内秉是学习几何的指导思想。 | |
133 | 2024-9-13 | 寻求本质的过程就是抽象的过程。 |
132 | 人的学习和机器的学习差别之一在于人只要少量甚至一个例子就会觉得自己学会了。(这既是人的优点也是人的缺点,但是往往理解就发生于一个非平凡的例子中(如有理数不完备这个事实其实是你对完备性的唯一认识,但是你会觉得你理解了完备性这个概念,而且大多数时候这个理解也够用了,之所以说大多数是因为我没碰到过反例,而且我崇尚认知的最高指导思想)) | |
131 | 如果你认为范性质是特殊的,你也应该立即觉得特殊的东西也许满足范性质。 | |
130 | 2024-9-10 | When you thought two things are analogous, that’s what a normal person should’ve noticed. You should step further by considering their difference. |
129 | 2024-8-25 | 知道对什么的抽象比知道抽象本身更有意义。 |
128 | 2024-8-25 | 作为一个数学民科,你要做的不是会证明一个定理,而是看着一个定理而觉得它很显然。(如果觉得不是很显然,找到为什么觉得不显然具体的原因,然后在证明中找到对应的那一步) |
127 | 2024-8-21 | 如果你只知道一个概念的范畴语言定义,那你完全没有理解它。范畴语言的好处在于你可以将不同概念的交换图进行比较从而对比他们的相似之处。(比如张量积的范性质定义) |
126 | 2024-8-13 | You are learning the story not part of the story. |
125 | 2024-8-10 | 再多的抽象语言不如否定一个trivial的例子(信号,“有某种意义的下降”) |
124 | 2024-8-5 | Let I be an interval in the real line. —-think about the worst condition of I, not the nicest! |
123 | 2024-7-15 | I keep iterating that when you define some structure such as a group or vector space, you should always immediately look for functions that preserve that structure. –All Angles youtube |
122 | Young man, in mathematics you don’t understand things. You just get used to them. —John von Neumann | |
121 | Consider things at the highest abstract layer. (Help to find the essence of it, e.g. 摩尔条纹) | |
120 | Lie groups are kind of “continuous” groups (which can be thought of as a curved surface, or a manifold), and Lie algebra is at the tangent space at the identity of the Lie group, with the “additional bilinear-ish operation” Lie brackets. | |
119 | 知识间的连接比知识本身的内容更重要 | |
118 | A formal definition is needed when dealing abstractly with sequences. —-Jeremy Orloff | |
117 | 表面上对样本值进行的操作其实背后是对概率密度函数进行对应的操作。(如加法对应卷积)–Frequentist | |
116 | 数学发现往往不是靠逻辑完成的(而是对称性、灵感、延拓:如实数域扩展到复数域,然后借用复数域中的结论来反映某种现实意义),靠逻辑完成的是严格化过程。 | |
115 | 数学中引入“space”的原因之一是提供了一个全局的、广阔的的视角来观察所有元素。(如probability space中的sigma algebra归纳列举了所有可测集,top space中的topology归纳列举了所有开集) | |
114 | Inspiration is cheap, but rigor is expensive. –Cassie Kozyrkov | |
113 | It’s always easy to learn a thing when you first have the map, the big picture on your hand, just don’t be overwhelmed by the details. | |
112 | Mathematics is a lattice, not a ladder. –Marcus Pivato | |
111 | It is entirely possible that future generations will look back, from the vantage point of view of a more sophisticated theory, and wonder how we could have so gullible. –David J. Griffiths | |
110 | 你的概率论打开方式不对。 | |
109 | 旋转是缩放的高维延拓。 | |
108 | 数学语言不应是你掩饰无知的工具。 | |
107 | 看山是山,看水是水;看山不是山,看水不是水;看山还是山,看水还是水。(从解构和建构) | |
106 | You gotta have taste! —杨振宁 | |
105 | 如果你发现了好的学习资料,应该感到悲哀,因为你可能因此少了很多独立发现规律的机会与兴奋感。 | |
104 | 概率应该叫做归一测度,随机变量应该叫做归一映射,期望应该叫做一阶原点矩,方差应该叫做二阶中心距,总之概率的一切命名都应该听起来与概率无任何关系! | |
103 | 2023-11-7 | 今天(11-7-2023)我打算采用一个看上去很有希望的思考方式:用封闭性和映射来思考一切事物(注意是一切!) |
102 | 如果你发现初学符号体系或概念非常不可理解(如vector和covector的指标规定),则他一定是对数学体系的高瞻远瞩。 | |
101 | 有些概念是以你目前的阅历难以理解的,那么先让它随风去吧。你不是天才,只是一个追求知识的普通人。 | |
100 | Don’t be lost in math!! | |
99 | The trick is to learn as much math as possible, without becoming a mathematician. (Haha, that’s really accurate!!!) —Jonathan Gardner | |
98 | The derivative is the opposite (or dual) of the boundary!!!!! | |
97 | 能定义自然数结构的系统,具有个体对整体进行概括和抽象的能力。——王木头 | |
96 | 教材是偏序的。——Maki | |
95 | 以教学为目的的学习总是好的。(注意不是以学习为目的,具体方法是写讲义)——Maki | |
94 | 数学家只需要纸和笔,不需要电脑。 | |
93 | 其实大部分的概念你感觉理解了,实际上只是熟悉了它的使用(“磨”,“be comfortable with the idea”)。 | |
92 | 平均值与采样方法有关! | |
91 | 将直线投影到圆上,感觉上原本均匀分布的点不再均匀了,这种感觉就是“非线性”,其中局部的密度之比便是导数。(所谓的对某某量进行微分,d theta, dl) | |
90 | Math is not just about solving problems. It’s about gaining insight, finding connections, and explanations. –Terence Tao | |
89 | 差生文具多。 | |
88 | 纯数学是为了提升我们对数学本身的认识;应用数学是利用数学去提升我们对其他学科的认识。 | |
87 | Don’t be intimated by the math jargon – I just want to get a point across, so bear with this bit. Still, it is very important to see the big picture first, and care about those little details later. | |
86 | 阅读数学书要揣摩作者这样安排概念、定理的理由,寻找a chain that connect them together. | |
85 | The simpler, the better. (任何让你感觉很痛苦的事一定是你没有在恰当的地方认识他,如Lagrange’s Thm, 国内的任何一本抽代教材,去s吧!TM看了一天的国内视频,还没最后youtube10分钟学的多!!!我***) | |
84 | 你要发挥双语者在学习数学中的优势。(我想强调的是对于每一个英文数学概念,如果有的话,尽量要知道其中文名称) | |
83 | 你无需保证第一次学的东西就是正确的。一个相对符合直觉的陈述比一个正确的理论往往在一开始更有价值。 | |
82 | 每当事实与直觉相悖时,不要惊慌,务必弄清 where is the point?? What’s wrong exactly with your intuition??(e.g. 生日悖论) | |
81 | 当你用图形思考数学时,要考虑其最大失效程度是多少。(如用图论中的Cycle思考循环群的失效程度比用开区间理解拓扑空间的失效程度低很多) | |
80 | 自学要假装是在上讨论班(用语言尝试表达出来,毕竟人类文明进化程度太低,对事物的理解还是大部分建立在语言上的)(事实证明这样能极大提高自学效率) | |
79 | 多线程学习+定义链逻辑梳理 | |
78 | If that doesn’t blow your mind, I don’t know what will. (de Rham cohomology) | |
77 | 如何理解定义?——What do you want it to behave? | |
76 | Be humble. | |
75 | 学任何的数学都需要从直观走向严格。(注意是先直观) | |
74 | 理解一个概念时(如Topological Space),尽量用接近这个定义的例子来理解(如如果用R来理解就会不完善,因为R不光有拓扑结构,还有序结构、几何结构) | |
73 | 向量是向量,面积可以是向量,线性泛函也可能是向量。 | |
72 | 你不一定要知道会证明每一个定理、推论或引理,但你需要一种安全感并且认识到他们不是魔法。 | |
71 | 离开实践的认识是不可能的。——毛泽东《实践论》 | |
70 | 教会别人往往不能从完全“正确”的理论开始,而是从不正确到正确,从直觉到严格的过程(教会自己也是同理)(如光滑流形,叶状结构流形的可积性严格化过程)(Accepting some of the mathematics behind the pictures but not enough to overwhelm you with theoretical details. –Fortney Manifolds) | |
69 | 数学喜欢给不同的事物相同的名字,也喜欢给相同的事物不同的名字。(对应一般化思想和不同的数学分支) | |
68 | 一本书只能从头开始读是非常可悲的 | |
67 | 定义链是很有必要研究的(不论是数学还是物理) | |
66 | OMG, I just found that Feynman had said the same thing: What I cannot create, I do not understand. Know how to solve every problem that has been solved. (知识的本体不是书上写的文字) | |
65 | You actually learned something as if you invented them. (You’d be confident using them) | |
64 | 数学对象都在自己的独立空间中,置身与他们自己的空间,设身处地为他们着想(如向量在切空间中,不在光滑流形之中;1-form在余切空间中,不在切空间中) | |
63 | Reading a book is nothing more than buying vegetables, you check them carefully first, package what you want, take away the packages, and digest them afterwards. (NOTE I wanna say the process of packaging is especially important) | |
62 | 要有宏观意识,知道学这个他在试着告诉你什么,它是一些技巧还是核心思想,还是一些不显然的结论,它能否通过计算机技术代替手工解决? | |
61 | 注重它给你带来的feeling,而不是具体细节,细节应当能通过它给你的feeling推自然导出。 | |
60 | It takes a lot of imagination. | |
59 | 学习的饱和现象:当学习一门新技术达到一定广度(或深度)时,总有几个问题无法完全理解,这时需要学习更深入的知识才能打破这种状况。 | |
58 | If you cannot “think something in one way”, then learn something else! | |
57 | 数学是一门赋予不同事物相同名字的艺术,而数学家的这个做法会让学生理所应当地认为他们是一样的,然而他们确实是一样的。(e.g. 向量空间上的乘法和数域上的乘法)(Original: Mathematics is the art of giving the same name to different things. –Henry Poincare) | |
56 | 数学中过多的对称性正在阻碍你的理解 | |
55 | 在好教材泛滥的世界中如何生存?选一本形象、严谨、全面的教材作为主要参考资料,其他的books作为补充(快速了解目录结构,直接跳到想要看的地方学习,看不懂再去看前面) | |
54 | 独立调研和思考探索的价值远远高于任何形式的资料(包括好的教材)。 | |
53 | 数学基础课程必备:线性代数,傅立叶分析,复变函数,概率论,信息论,随机过程,微分几何,数值分析,微分方程,场论,离散数学,最优化理论 | |
52 | 快速入手一样你不太敢下手的学科:干它(直接做题),再学习概念,形成正反馈 | |
51 | 系统性的学习可能不是最有效的(从一本书的中间开始学看起来很扯,而且容易强迫症,但往往这种方法最符合大脑的学习模式) | |
50 | 一门新的学科就是一种全新的思考方式 | |
49 | 参数化的本质是转为独立变量(usu. 单变量或双变量)的积分,因为只有这样才能积出来 | |
48 | 理解的代价是不会写证明,说出这句话的人对“理解”这个概念的理解有待提高 | |
47 | 牛顿发明的那套好像很喜欢处理加权和似的 | |
46 | 几重积分其实是看有几个自由度 | |
45 | 不要试图脱离三维空间来理解高维或抽象空间,恰恰相反,对这些抽象空间的理解一定建立在欧几里得空间中的几何直观(但是事后一定要尝试脱离三维欧式空间来进行思考!) | |
44 | 当你觉得一个概念无法理解时(无法理解它的动机,表达了什么),那么暂时先忽视它(假设没有发明这个概念),再通过实际问题体会为什么要有这个概念 | |
43 | 场论是另一种看待存在的方式 | |
42 | Good explanation>symbolic proof | |
41 | It seems almost cruel of the universe to fill its language with riddles that we either can’t solve, or where we know that any solution would be useless for long-term prediction anyway. It is cruel, but then again, it should also be reassuring. It gives some hope, that the complexity we see in the world can be studied somewhere in this math, and that it’s not hidden away in the mismatch between model and reality. | |
40 | 同一词汇在不同领域内的应用要做区分(如数学上的维度和物理上的维度,权重矩阵和对象矩阵) | |
39 | 不是系统学习才叫学习,以需求为导向的学习才是(如果学一个孤立于应用范围之外的知识,只会增大精神内耗) | |
38 | 如何判断数学书质量高?有习题答案,习题精炼(数量少),文本中的例子数量较多,文本分段,有严格定义和证明,有图。 | |
37 | Perhaps, there isn’t the definition of a matrix, but only, interpretations of the matrices. —See Matrices | |
36 | 学习是为了增强泛化能力,是为了解决从未见过的情况而进行的思维网络构建 | |
35 | 要注意知识的输入范围和输出范围(直接上手解题可以增强这种泛化能力,因为学习过程中往往只能碰到用这种方法能解决的问题,这些问题有什么共同特征却被忽略了,用这种方法不能解决的问题也非常有思考价值) | |
34 | 指令学习和归纳学习并重(先指令学习构建“底稿”,防止过拟合,再用归纳学习建立知识的网状结构,大脑的潜意识会自动拟合,大量的例子来体会知识非常重要!) | |
33 | 学习任何一样东西,都要有方法论! | |
32 | 连接词很重要(引入的过程比知识本身更重要,“good teacher”的本质区别, “motivation”) | |
31 | 明确输入输出 | |
30 | 将要做的课后习题提前打勾 | |
29 | 先把书读厚,再把书读薄。——华罗庚 | |
28 | 如果你不能用简单的语言描述你所学的东西,你就没有真正学会它。——费曼 | |
27 | 有时我们只知道了知识的内容,而忘记了它的motivation | |
26 | 对于数学,学什么都不过分 | |
25 | 一个清晰问题表述的形成所需的时间可能不亚于解决它所花的时间 | |
24 | 一个事物没有本身的样子,只有你看见它的样子 | |
23 | 有时候数学也要先会用,再理解。(“如果我早看到这个视频…”–你也不一定能学好!) | |
22 | 维度是数学概念而不是物理概念 | |
21 | The primary reason that you don’t understand a book is that you are not aware of what the writers said. You may be aware of what others have told you, and that’s the very great significance of discussion. | |
20 | 当你觉得问题很多,想不清或提不出明确的问题时,你就该做题了(看看stack exchange上的提问就知道什么是“明确”了,与知乎没法比) | |
19 | 数学没有物理是瞎子,物理没有数学是跛子 | |
18 | 数值计算要检验现实性! | |
17 | 一个等式(关系式)正向、逆向都成立! | |
16 | 以有崖随无涯,殆已!(此无涯非彼无涯) | |
15 | 分而治之:化指数复杂度为线性复杂度 | |
14 | “学会”概念的最高境界:Aware of it’s motivation and intuition(sometime loose). | |
13 | 人类永远无法“了解”知识,知识在更高的维度上,我们学到的,只是那个抽象形象的投影(例如用open set来理解topological space),学到的投影越多,就越能体会知识的结构以及其哲学含义。 | |
12 | Images of knowledge in your brain are built through examples.(对一个抽象知识的理解一定建立在已经存在的具体物质之上,比如理解int U, ext U) | |
11 | imagination. –Feynman | |
10 | 知识结构决定学习方法 | |
9 | 若一个概念有多种定义,态度:先确定一个定义,另一个定义用“碰巧成立”(happen to hold). | |
8 | You can learn anything, right now! | |
7 | 对某些方法,思想,定理的重要性非常 aware | |
6 | 数学是优美的,数学是自然的,数学是严格的 | |
5 | 你要知道你在做什么 | |
4 | 我相信如果抛开数学符号,有关数学定理的理解会容易一半 | |
3 | 有目的的学习 | |
2 | intuition and axiom are both important! | |
1 | A good question is far more valuable than it’s Sol. |
3 How to Learn Physics
No. | Date | Ideas |
---|---|---|
34 | 2025-4-22 | Hamiltonian mechanics is the conservation of information entropy. — Assumptions of Physics |
33 | 2025-4-9 | 直觉是需要修正的。 —杨振宁 |
32 | 光子数对应的波动物理量是电场的最大振幅! | |
31 | 你看到的、经历的每一件事都对应一个动力学方程,hence对应一个电路。(这是多么令人振奋的消息!爱情变化-混沌电路,弹簧振子-震荡电路等等) | |
30 | Control Theory: 弹簧-电容,质量-电感,阻尼-电阻;弹性势能-电场能,动能-磁场能,阻力系数(damping coeff)-电阻 | |
29 | 动量是空间平移的生成元。 | |
28 | 我们的科学竟然对想像提出如此可怕的要求。——Feynman | |
27 | No writing, No Reading! | |
26 | No Reading, No learning! | |
25 | 伤其十指不如断其一指!You gotta have taste! –杨振宁,张颢 | |
24 | 不要想在不清楚事实的条件下尝试研究问题! | |
23 | 看视频记得跳着看(无论那个视频讲的多么好,如3b1b的同理),连着看容易obsessed by details or kidnapped locally. 看视频不要一次性看懂,要留一些小的空隙,让大脑主动把他们连接起来。 | |
22 | Think of the same thing in multiple different ways! That’s how you understand things. | |
21 | The electric field is the time version of the magnetic field. Energy is the time version of momentum. | |
20 | 物理是试图建立一套简单符号逻辑体系,解释并预言复杂客观世界的学科。 | |
19 | 要注意你对这个世界的认识有没有实质性的进展(如Physics I中的光的衍射等,只需知道光是波就能推导出所有结论,等于没有实质性进展) | |
18 | Physics works, and I’m still alive. | |
17 | You’d better have a picture of each physical quantities in which you can be free to analyse then, and the picture should be able to contain as much quantities as possible.(Set up relationships) | |
16 | 物理中cos里面的东西是现实量与抽象量的映射,由于cos能够接受的量只能是角度,现实量一般不会带有角度,所以在转化时一般会乘上带有 \(2\pi\) 的量(角空间中的量,如 \(\omega, k\)) | |
15 | Ideal Lab中的规则:只要理论上能测出的物理量都能被精确测出(如1ns通过某微小界面的氧分子数量) | |
14 | 每个物理量都要有“直观的认识”(即要清楚物理意义),这些物理量能够在Ideal Lab里被测出。 | |
13 | 理解的本质是找到了知识本身与现实物体的相似性(映射)。但这种映射一定是片面的,我们之所以感觉到“理解”了某个知识,是因为偏差很小以至于可以忽略。然而对于某些抽象知识(如微观粒子的行为),可能不存在足够精确的映射以至于人类无法“理解”它们,我们能做的,就是针对它们不同的行为应用不同的映射来进行研究和尝试理解,这样做的后果是某些映射相互矛盾(常识层面上)。 | |
12 | 物理建模时要区分哪些是事实,哪些是假设,哪些是简化 | |
11 | 学习的时候可以适当放宽逻辑,比如为什么在研究复摆的时候可以只分析质心?但事后一定要 “complete the story”. | |
10 | It’s nice to do a quick gut check that our formula makes physical sense. (符合: That checks out) | |
9 | 开拓型学习和扩张型学习(开拓型指已有的知识不断扩张,已达到学习新知识的效果;扩张型指每个知识范畴在幼年时期都是较为独立的孤岛,学习过程中不同大小的孤岛不断扩张,从而建立知识之间的联系),事实证明扩张型的学习更为有效和高效。 | |
8 | 某种理论既然能解释某些问题,证明它肯定存在某种真理,我们要读懂的就是这某种真理,而非其他。(波恩看到海森堡的论文能解释一维谐振子问题,他的理论肯定在某种角度来说是对的)(《一个人两个人三个人的文章》) | |
7 | 从可观测的量出发。——海森堡(小方格子力学),原子光谱可观测,而原子轨道无法观测 | |
6 | It takes a lot of imagination. —Feynman | |
5 | 什么叫了解一门学科?我不知道怎么做,但我知道一定存在某种方法可以解决 | |
4 | 世上本没有动能,自从有人类以后,才有的动能 | |
3 | Stand back and look at what you’re doing, that’s when you get an understanding of it. | |
2 | 总是用离散的思想思考,这既具有普适性,又尊重了思维的自然性和牛顿深刻的哲学思想。(离散体模型与连续统模型) | |
1 | 物理学本质上是对宇宙的一种总结和不完全归纳,假设这个宇宙是定域的、物质的(这个假设也使物理学脱离了神学),那么物理学被要求能够自洽地解释人能够感知的外界变化、预测下一个能被人类感知的运动、并追求大道至简。这决定了物理学是一门以实验为基础的科学。 | |