The Observation Theory EncyclopediaFrom TSKAboutBy kindBy chapterBy Lean fileLedgerProvenance

metric concept

DefinitionA rule for the distance between two rows or two outputs. A distance is a choice of what to ignore, and a local metric on an output is a positive semidefinite matrix, a pseudometric until its kernel is quotiented out, which the book calls a metric by convention. Chapter 3 section 3.1 and chapter 0 section 0.5.
ExampleEuclidean distance between (0, 0) and (3, 4) is 5, and a metric that ignores the second coordinate reads 3.
BookData Mining as Observation, draft 0.2, commit f3914f0; entry id metric, kind concept.
Statusno ledger row names this entry. Corrections: none recorded.
Defining equation

Book equation 3.1.

Assumptions and scope
  • A rule for the distance between two rows, or between two outputs. A distance function is a choice of what to ignore, and the local metric on a consumer’s output is the positive semidefinite matrix that says how a small change is scored. The squared Euclidean distance is symmetric and zero exactly between equal rows.
  • A dataset-level loss, accuracy, F1, or a rank correlation is not a local metric and cannot be inserted into the read-operator formula, and the same two codes get opposite verdicts from two metrics.
  • A positive semidefinite read form is a pseudometric until the directions in its kernel are quotiented out. The book calls it a metric by convention, on the quotient.
Prior artnone recorded
Evidencelean/DataMiningAsObservation/EuclideanDistance.lean, lean/DataMiningAsObservation/OutputMetric.lean
Reviewedsemantic review 2026-09-06; generated 2026-09-10 from records at the commits on the provenance page.
xyθ
A rule for the distance between two rows or two outputs.

Equation

Book equation 3.1.

\[\begin{gathered} d_O(u)=u^{\top}\Sigma\,u, \qquad u=(\cos15^\circ,\ \sin15^\circ), \\ \Sigma_1=\operatorname{diag}(0.3,1.7),\ \Sigma_2=\operatorname{diag}(1.7,0.3), \qquad d_O=0.394\ \text{vs}\ 1.606. \end{gathered}\]

Book equation 0.11.

\[P_C(x)=J(x)^{\top}G\big(C(x)\big)\,J(x),\qquad J(x)=\frac{\partial C}{\partial x}(x),\qquad \bar P_{C,\mu}=\mathbb E_{\mu}\!\left[P_C(x)\right].\]

Conditions

Conditions are curated in entries.toml rather than read from a record.

Ledger

none

First stated

Chapter 3 section 3.1 and chapter 0 section 0.5 of Data Mining as Observation, with the read metric in geometric-observation/chapters/ch05_the_read_metric_and_the_quotient.md:7-48.

Measurements

none

Failures and corrections

none

Invariance envelope

none declared

Machine checked

lean/DataMiningAsObservation/EuclideanDistance.lean, theorems distSq_eq_quad_one, distSq_comm, distSq_eq_zero_iff, ranking_flips, at observation-data-mining f3914f0; what the check covers is stated in the book’s appendix C.

lean/DataMiningAsObservation/OutputMetric.lean, theorems neg_reverses, readOp_of_neg, cost_flips, at observation-data-mining f3914f0; what the check covers is stated in the book’s appendix C.

Used in

Data Mining as Observation chapters 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14.

Related

output metric; Euclidean distance; read operator; quotient; geodesic distance.

See also

Book equations stated beside the entry’s terms, not defining it: 1.1.

Ledger rows that cite the entry’s records without naming it: GO-EC-3.

Sources-table rows that share a record with the entry without naming it: chapter 2 section 2.2, chapter 3 section 3.2.

Status

Generated 2026-09-10 by encyclopedia/generate.py; book at observation-data-mining f3914f0; the commit of every record is listed in the encyclopedia’s provenance.

← mean, medianmin-over-strata →