The most beautiful formula not enough people understand

2026-02-27 · Source: 3Blue1Brown · Field: Science & Research — Mathematics & Computational Sciences · Depth: Advanced, extended

Summary

This content explores the counterintuitive nature of high-dimensional geometry, particularly focusing on the volume of n-dimensional spheres. It begins with two puzzles: calculating the probability of random points falling within a unit circle or sphere in 2D and 3D, then generalizing to higher dimensions, highlighting their utility in fields like machine learning. The second puzzle demonstrates the counterintuitive behavior of spheres within cubes in higher dimensions, where an inner sphere can exceed the bounding box, suggesting cubes become "spiky." The core of the discussion derives a general formula for the volume of an n-dimensional unit ball, V_n(R) = (π^(n/2) / Γ(n/2 + 1)) * R^n, using a recursive "knight's move" method based on Archimedes' principle and calculus. This formula reveals that the volume of a unit ball initially increases with dimension, peaks around n=5, and then rapidly approaches zero, becoming infinitesimal in very high dimensions. This phenomenon implies that in high-dimensional spaces, most of a sphere's volume is concentrated near its boundary, and spheres occupy a negligible fraction of their enclosing cubes.

Key takeaway

For Machine Learning Engineers working with high-dimensional data representations, understanding the counterintuitive properties of n-dimensional spheres is crucial. Be aware that in high dimensions, unit balls occupy an infinitesimally small fraction of their enclosing cubes, and their volume is concentrated almost entirely near the boundary. This insight can inform data sampling strategies, distance metric choices, and the interpretation of model behavior in complex feature spaces, preventing misinterpretations based on low-dimensional intuition.

Key insights

High-dimensional geometry reveals counterintuitive properties, with n-dimensional sphere volumes peaking then becoming infinitesimal.

Principles

• Higher-dimensional geometry is useful for data representation.
• Archimedes' method generalizes to higher dimensions.
• Volume of unit balls becomes negligible in high dimensions.

Method

Derive n-dimensional sphere volumes using a recurrence relation: V_n = V_(n-2) * (2π/n), based on Archimedes' area-preserving projection and calculus integration, then generalize with the Gamma function.

In practice

• Interpret long lists of numbers as points in high-dimensional space.
• Recognize that high-dimensional data is mostly near the "surface."
• Use simulation to empirically verify high-dimensional volume formulas.

Topics

• High-Dimensional Geometry
• Sphere Volume Formulas
• Gamma Function
• Machine Learning
• Archimedes' Generalization

Best for: AI Student, Machine Learning Engineer, Data Scientist

Related on AIssential

• Why Transformer Representations Tend to Live on Hyperspheres — Transformer representations cluster on hyperspheres due to normalization and high-dimensional geometry, making direction paramount.
• The Partition Principle Revisited: Non-Equal Volume Designs Achieve Minimal Expected Star Discrepancy — Non-equal volume partitions can achieve lower expected star discrepancy than classical jittered sampling.
• Toy Combinatorial Interpretability Models Reveal Lottery Tickets in Early Feature Space — Winning tickets preserve feature-space geometry, not just weight-space subnetwork identity.
• Shape, Symmetries, and Structure: The Changing Role of Mathematics in Machine Learning Research — Mathematics' role in ML is shifting from theoretical guarantees to post-hoc explanations and high-level architectural guidance, leveraging advanced pure math.
• Non-asymptotic quantisation of spherically symmetric distributions — This paper addresses the challenge of optimal quantisation in high dimensions ($d$), where Zador's theorem requires an astronomically large sample size ($n$) for its asymptotic behavior to be observed.
• You Can Fit a Million Nearly-Perpendicular Arrows in 768 Dimensions. — High-dimensional spaces enable millions of nearly-perpendicular vectors, explaining why embeddings effectively represent numerous distinct ideas.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by 3Blue1Brown.