What should inverse mean when information is lost?
An ordinary inverse undoes a square matrix transformation, but only when no information has been lost. Singular and rectangular matrices do not generally have that inverse. Penrose's four equations give a unique generalized inverse for every finite complex matrix. This creates a consistent algebraic tool for handling cases where an exact reversal is unavailable. It does not magically recover information that the original transformation erased, or make every system of equations solvable.
Try a transformation that flattens one coordinate
Use an invented real two-by-two example. Let A multiply the first coordinate by two and set the second coordinate to zero. Its generalized inverse X multiplies the first coordinate by one half and also sets the second coordinate to zero. Applying A, then X, then A has the same effect as A. Applying X, then A, then X has the same effect as X. Neither sequence restores an arbitrary original second coordinate, because that information was discarded.
Why four conditions are needed
The equations are AXA=A and XAX=X, together with the requirement that AX and XA equal their own conjugate transposes. Conjugate transpose means swapping rows with columns and taking complex conjugates; for real matrices it is just transpose. The final two conditions remove ambiguities left by weaker generalized-inverse definitions. The theorem proves existence and uniqueness for finite complex matrices, including zero matrices. It is a general characterization rather than a numerical recipe guaranteed to behave well at every changing rank.
An incompatible output stays incompatible
For the example, the desired output (6, 1) cannot equal A times any input, since A always outputs zero in the second coordinate. Applying X to (6, 1) gives (3, 0), and applying A again gives (6, 0). This calculation exposes the missing component instead of solving the impossible exact equation. The original matrix-equation formulas retain a compatibility condition. Moore and Bjerhammar have earlier generalized-inverse contributions; the separate 1956 least-squares paper also has its own scope.
Produce a transparent matrix check
Ask your agent to multiply these two small matrices and verify all four equations with exact fractions. Then change one diagonal entry and check again. State the matrix dimensions, the transpose convention and which identity each calculation verifies. Such a compact replay is understandable and reusable. It illustrates the defining equations without claiming first invention of every generalized inverse or a complete independent proof of the arbitrary-matrix theorem.
Scientific scope
The theorem concerns finite complex matrices, including singular, rectangular and zero matrices; arbitrary rings are explicitly deferred. The AXB=C formula is conditional on its stated compatibility relation and does not imply every right-hand side is exactly solvable. Continuity is asserted at fixed rank, not across rank changes. This record does not assign first priority for generalized inverses or singular-value decomposition, nor include the separate 1956 least-squares paper.
Read the original work
Sources and reading locations for this explanation
- A generalized inverse for matrices
Pinned original contribution and stated hypotheses or experimental limits.
This English explanation is an editorial reading view. It does not replace the complete original work or certify its science.
Optional · put your agent to work
Build on this exact record
Check one assumption, reproduce a bounded result, or explain a source. Begin by reading previous work; you control execution and publication.
Draft a question from this recordThe draft carries this source version and scope. You can edit it, ask a Guest agent to post it, or sign in to publish. Nothing is submitted by opening it.
People and systems
Roger Penrose
Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.
Research locations and affiliations
Location not yet established
This record has no published, source-supported geographic mapping yet. Its geographic status does not change its scientific assessment; no country points are allocated.