The packing number P(ε,S,ρ) is the maximum cardinality of a subset of S whose elements are pairwise at distance greater than 2ε (equivalently, the maximum number of disjoint closed balls of radius ε that can be placed inside S); it measures how many well-separated points S can contain at scale ε.

Principle

Quantify capacity by maximal separation: count the largest set of points in S that are mutually ε-separated so that corresponding ε-balls do not overlap, capturing internal diversity at that scale.

Demonstration

Inside the unit ball in R^d, one can place on the order of (1/ε)^d disjoint ε-balls; for example, a lattice packing yields a lower bound on P(ε) showing the dependence on dimension and radius.

Misapplication

Assuming packing and covering numbers are identical; or using packing counts to infer covering without accounting for the factor-of-two radius relationship and metric-specific geometry.

Consequence

High packing number at small ε indicates high capacity and separability, which translates into lower bounds on sample complexity and implies many distinguishable hypotheses at that scale.

Reversal

Consider the covering number as the complementary notion: while packing measures how many disjoint balls fit inside S, covering measures how many balls are needed to cover S; inequalities relate the two (e.g., packing(2ε) ≤ covering(ε) ≤ packing(ε)).

Boundary

Depends on metric, ball shape, and ambient space; packing assumes disjointness of balls, so it is undefined or trivial if S is too small relative to ε, and it does not account for near-overlap or probabilistic overlap tolerances.

Semantic Tension

Competes with covering number and with continuous notions of capacity (entropy, dimension); sets can have similar packing numbers yet differ in covering behaviour or in their amenability to approximation by structured models.

Synthesis

Packing number captures the maximal number of ε-separated points in a set: it is a scale-dependent count of internal separability that complements covering-based measures of compactness.