Explore by topic
Mathematics, probability, statistics, formal logic, and algorithmic principles.
Gradient noise in SGD is not a bug to fix but a feature to exploit: it provides implicit regularization that helps the optimizer escape saddle points, traverse flat regions, and prefer wide minima that generalize.
Statistical hypothesis testing provides a framework for distinguishing real patterns from chance ones by measuring how often data with no real pattern would look just as interesting.
Unsupervised models are evaluated by measurable properties of the structure they produce — internal consistency, stability under perturbation, and downstream usefulness — rather than by matching a hidden ground truth.
How unsupervised learning reveals latent structure — clusters, low-dimensional manifolds, and anomalies — hidden inside high-dimensional data.
How an unsupervised algorithm like k-means turns feature vectors into discrete clusters, and why its verdict is real for some shapes and unreliable for others.
Unsupervised learning works by turning raw inputs into compressed, similarity-based representations — clusters, dimensions, and patterns — that the model itself defines as 'consistent.'
有限域上的模乘法本身是双射、结构完全确定，可一旦把多次乘法叠成幂，求逆就退化为遍历整个乘法群，规模随域大小指数膨胀。
Factor pairs always straddle √n, so checking only up to √n — and only at prime candidates — finds every prime factor.
The raised 7th resolves up to tonic because a half step is smaller than a whole step, so upward motion covers the shortest distance to a stable pitch.
A point solves y = -2x + 8 when substituting its x-coordinate yields exactly its y-coordinate; (3, 2) does not, because -2(3)+8 = 2 only by coincidence of arithmetic but equals 2, which equals the y-value, so it actually is a solution.
Borda count collapses the *ordering* of preferences into *points*, so the size of the margin at each rank — not just the ranking itself — decides the winner, and this lets a candidate who is last against every rival still win overall.
Arrow's impossibility becomes tangible when a single set of ballots reveals three different failures: plurality violates Independence of Irrelevant Alternatives through a spoiler, Borda count violates Condorcet/transitivity via a reversal, and instant-runoff voting fails transitivity through a cycle.
Factor pairs always straddle √n, so checking only up to √n — and only at prime candidates — finds every prime factor.
Arrow's theorem proves that no ranked-ballot voting rule with three or more options can simultaneously satisfy every basic fairness condition—at least one must be sacrificed.
A parabola is the only curve where every ray from its focus reflects parallel to the axis, because every point on the curve lies the same total distance from the focus and the exit line.
A segment's length is the straight-line distance between its endpoints, so crossing an axis does not change its length.
A parabola focuses any wave that obeys the law of reflection, including sound, because equal path lengths from the focus to the curve collapse spherical wavefronts into a beam.
ASA 和 AAS 的真正差异是已知元素的位置组合，而不是判定能力的强弱；通过内角和定理，两者可以互相转换。
理解 ASA 全等判定中'夹边'的关键作用，以及如何用 ASA 写出规范的全等证明步骤。
抗碰撞性是指没有任何比穷举更快的方法能故意制造碰撞,哈希的安全性建立在计算不可行而非数学不可能之上。
群论中的对称群为数学结构、分子几何和物理守恒定律提供了一套统一的描述语言——同一套抽象结构在不同领域呈现出平行的具体形态。
A z-score measures how many standard deviations a value sits from the mean, and the standard normal table translates that distance into a cumulative probability.
五次方程不可解的根源在于对称群 S₅（及其交错子群 A₅）的内部结构：A₅ 是一个'简单群'且非可解群，从而破坏了一般五次方程伽罗瓦群的群列终止条件。
伽罗瓦把方程根的置换视为一个群，方程可用根式求解当且仅当这个伽罗瓦群是可解群。