The Observation Theory EncyclopediaFrom TSKAboutBy kindBy chapterBy Lean fileLedgerProvenance

low-rank approximation concept

DefinitionThe best approximation of a matrix by one of rank k in total squared error keeps the first k terms of its singular value decomposition, and the error is the sum of the squared singular values dropped, the Eckart and Young theorem. It is the identity reader's best approximation, and chapter 4 is about consumers for which it is not the best. Primer L, equation L.18. Also low-rank, Eckart.
ExampleKeeping only the root 3 term of the three by two example leaves squared error 1, one quarter of the total.
BookData Mining as Observation, draft 0.2, commit f3914f0; entry id low-rank-approximation, kind concept.
Statusno ledger row names this entry. Corrections: none recorded.
Defining equation

Book equation L.18.

Assumptions and scopenone
Prior artnone recorded
Evidencenone
Reviewedsemantic review 2026-09-09; generated 2026-09-10 from records at the commits on the provenance page.

Equation

Book equation L.18.

\[\|A\|_F^{2}=\sum_{i,j}A_{ij}^{2}=\sum_i\sigma_i^{2},\qquad \min_{\operatorname{rank}B\le k}\|A-B\|_F^{2}=\sum_{i>k}\sigma_i^{2}.\]

Conditions

none

Ledger

none

First stated

Primer L section L.8 of Data Mining as Observation, added in draft 0.3 (2026-09-09) for the ECE 514 readers whose first courses are far behind. The idea is standard and TSK Appendix A covers it at length.

Measurements

none

Failures and corrections

none

Invariance envelope

none declared

Machine checked

none

Used in

Data Mining as Observation primer L, chapters 6, 11.

Related

singular value decomposition; truncation; explained variance; the flip; identity reader.

See also

none

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.

← logistic regressionMahalanobis distance →