导航几秒算出全城最短路线
导航用从起点逐层向外扩展的贪心搜索，只检查每一个路口最优的一次到达方式，因此几秒内即可锁定全城最短路线。
8 min 4 scenes
Explore by topic
Math & Logic
Mathematics, probability, statistics, formal logic, and algorithmic principles.
Where Entropy Hides Beyond Hot Coffee
Entropy is the tendency of systems to move toward their most probable arrangements, and this single statistical rule governs mixing, information loss, and the arrow of time itself.
18 min 9 scenes
Interpolation vs. Approximation in Neural Networks
Interpolation forces the network to pass exactly through every training point, producing high-frequency oscillations between them; approximation sacrifices exact fit for a smoother function that captures the underlying trend and behaves reasonably on unseen inputs.
16 min 8 scenes
When Queries and Keys Speak Different Languages
Dimension mismatch between queries and keys distorts the attention score distribution — specifically inflating its variance — and the sqrt(d_k) normalization in scaled dot-product attention exists precisely to keep that distribution stable regardless of dimension.
18 min 9 scenes
Screening vs. Diagnostic: What Does a Positive Really Mean?
A positive screening result signals that further testing is warranted, while a positive diagnostic result signals that the condition is sufficiently likely to guide treatment decisions.
16 min 8 scenes
The Edge of a Nearest‑Centroid Cluster
The nearest‑centroid rule defines cluster membership by which centroid is closest, so cluster edges are the straight perpendicular bisectors between centroids — independent of the actual spread, shape, or density of the points.
16 min 8 scenes
When Big-O Lies: The Constant Trap
Big-O describes how running time grows with input size, but it hides three things that can dominate real performance: the multiplicative constant, the cost of each primitive operation, and how the data interacts with memory hierarchy. Algorithm choice should weight all three, not just the asymptotic class.
16 min 8 scenes
Inside the Hash Function
How keys are converted into numeric values, how those values are squeezed into a fixed array range with modulo arithmetic, and why collisions are an expected consequence.
18 min 9 scenes
Why Neural Nets Fit Noise
A network that can fit unseen noise isn't memorizing exceptions — it's because high-capacity models can interpolate the training data exactly, which forces the learned function to be smooth enough to generalize on real patterns.
24 min 12 scenes
Why a Positive Test Often Means You're Healthy
Positive medical tests mean healthy because screening tests are designed to detect a rare, specific condition — so a normal result is flagged as 'positive' for normality, and the meaning of positive is inverted by the test's purpose and the population it screens.
16 min 8 scenes
What Is Attention Actually Comparing?
Attention compares a query vector to a bank of key vectors via a dot product, and the resulting scores weight how much of each value vector is read — a similarity operation over learned representations, not over surface words.
16 min 8 scenes
Where Does a Cluster End?
A cluster's edge is the set of points whose nearest centroid switches, and the centroid itself is the mean of its assigned points.
14 min 7 scenes
Why Big-O Ignores Constants
Big-O ignores constants because it describes how runtime scales with input size, and any fixed multiplier becomes irrelevant once the input grows.
8 min 4 scenes
Why Hash Table Lookup Is O(1)
Hash table lookup is constant time because the hash function maps a key directly to a slot index, turning a search across n entries into a single computed jump — provided the table is sized and the hash is uniform enough that collisions stay rare and amortized constant.
18 min 9 scenes
如何快速生成质数候选
解释筛法通过预先标记倍数来实现批量淘汰的机制，并说明它在候选质数生成中相对于试除法的效率优势。
16 min 8 scenes
勾股数与高次方程
理解勾股数 (m² − n², 2mn, m² + n²) 的参数化如何为费马无穷递降法提供起点，并展示由此推出 x⁴ + y⁴ = z² 无正整数解的逻辑链条。
18 min 9 scenes
Why the Cauchy Distribution Defies the Bell Curve
The Cauchy distribution is preserved by convolution because it is a stable distribution with stability parameter α = 1; its heavy (1/x²) tails ensure that extreme observations keep reappearing, so repeated averaging never concentrates probability around the mean.
18 min 9 scenes
Gradient Noise vs. Parameter Noise
Gradient noise and parameter noise shape optimization through fundamentally different mechanisms: gradient noise produces an implicit regularization that scales with learning rate, while parameter noise produces an explicit perturbation that scales with weight magnitude.
16 min 8 scenes
Why Repeated Convolution Becomes a Bell Curve
Repeated convolution of any probability distribution with itself converges to a Gaussian, and this happens because convolution is the operation of adding independent random variables, whose sums the Central Limit Theorem forces toward the normal distribution.
16 min 8 scenes
From Skewed to Bell-Shaped: How the CLT Works
Averages of skewed data become normal because each sample mean is a sum of independent random pulls that symmetrizes through aggregation, with the spread shrinking predictably as 1/√n.
18 min 9 scenes
When the Central Limit Theorem Breaks
The Central Limit Theorem requires finite variance and independent (or weakly dependent) summands; heavy tails push variance to infinity, and correlations let individual shocks persist across the average.
18 min 9 scenes
Why Self-Convolution Becomes a Bell Curve
Repeated self-convergence toward a Gaussian, because each convolution sums independent copies and the fixed point of this operation is the bell curve.
16 min 8 scenes
How Convolution Erases Edges into a Bell Curve
Convolution of independent distributions is equivalent to summing their random variables, and the Central Limit Theorem guarantees that many such convolutions smooth every feature into a Gaussian.
18 min 9 scenes
Bootstrapping a Bell Curve
The bell shape in bootstrapping comes from the sampling distribution of the mean, not from the underlying dataset — and that distribution becomes bell-shaped by the Central Limit Theorem as resamples grow in size.
16 min 8 scenes