Projection¶
Core Idea¶
A projection is the structural move of mapping a higher-dimensional or richer object onto a lower-dimensional or constrained representation along a chosen direction or onto a chosen target, deliberately collapsing the dimensions perpendicular to the target while preserving those parallel to it. Four commitments define it: a source with more degrees of freedom than will be retained; a target subspace, surface, or constrained representation onto which the source is mapped; a projection direction, sometimes explicit and sometimes implicit in the choice of target, that determines what gets lost; and a structural identity — applying the projection a second time gives the same result (idempotence), which is the algebraic signature distinguishing a projection from an arbitrary lossy map.
The skeleton has four parts: a source space with structure to be reduced; a target; a projection direction, the equivalence relation collapsing source into target; and a residual, the orthogonal complement of what was kept — everything the projection threw away. The move is informative when the target captures the load-bearing variation and the residual is unimportant or noise; it is misleading when the target was a poor choice and the residual carried the signal. The residual is therefore a first-class object, not an afterthought: naming it is what turns "I have the picture" into the more honest "I have a picture, from one direction."
Projection is the structural complement of representation: it is the act of producing a particular representation by deciding what to drop. The shadow of a solid is a projection (3D source, 2D target, direction set by the light); a flat map of a country is a projection (curved surface to flat plane); a single summary statistic is a projection (high-dimensional reality to one scalar); an executive summary is a projection (a long argument to a short statement, direction set by what the reader needs). The idempotence signature — projecting twice equals projecting once — is what makes "summary of summary equals summary" hold when the two directions agree, and its failure is a diagnostic that what looked like a clean projection was actually a richer, drifting transformation.
How would you explain it like I'm…
Shadow on the Wall
Flattening With Leftovers
Collapse Along a Direction
Structural Signature¶
the richer source space — the lower-dimensional target — the projection direction — the discarded residual — the idempotence signature — the reconstruction error
A structure is a projection when each of the following holds:
- A richer source. There is an object with more degrees of freedom than will be retained — a higher-dimensional space, a curved surface, a long argument, a multi-channel signal.
- A lower-dimensional target. A subspace, surface, or constrained representation is chosen onto which the source is mapped, retaining the dimensions parallel to it.
- A projection direction. An equivalence relation — sometimes explicit, sometimes implicit in the target — collapses the source into the target by declaring what counts as the same; this choice determines what gets lost.
- The discarded residual. The orthogonal complement of what was kept — everything the projection threw away — is a first-class object, informative when it is noise and dangerous when it carried the signal.
- The idempotence signature. Applying the projection a second time returns the same result; this algebraic identity distinguishes a clean projection from an arbitrary lossy map, and its failure diagnoses a drifting transformation.
- The reconstruction error. The residual's magnitude under the source's metric — variance unexplained, fit quality, signal-to-noise — scores how much the reduction destroyed.
The components compose so that a chosen direction collapses the dimensions perpendicular to a target while preserving those parallel to it — an honest reduction exactly when the target captures the load-bearing variation, the residual is named, and idempotence holds.
What It Is Not¶
- Not a
perspective. A projection is the substrate-neutral act of mapping along a direction;perspectivecarries the viewpoint-laden, interest-bearing sense of the same move. Presenting a perspective as a neutral projection hides the load-bearing choice in the direction. - Not
abstraction. Abstraction drops detail to keep an essence at any level; projection drops the dimensions perpendicular to a chosen target, keeping those parallel. Projection is the specific, direction-indexed, idempotent member of the reduction family. - Not an arbitrary lossy map. Idempotence — projecting twice equals projecting once — is the algebraic signature; a reduction whose value drifts under re-application is a transformation, not a projection, and the drift is a diagnostic.
- Not
representation. Representation is the resulting reduced form; projection is the operation that produces it by deciding what to drop. The picture versus the act of picturing. - Not
inversion. Inversion runs a mapping backward to recover sources; projection deliberately discards the residual and is generally not invertible — its partial inverse (a section or lift) measures exactly how much was destroyed. - Common misclassification. Mistaking "a picture from one direction" for "the picture," then making a decision the chosen direction was never optimized for — steering by an equal-area map, comparing land areas on Mercator. A summary faithful for one decision can mislead another.
Broad Use¶
The map-down-along-a-direction pattern recurs across substrates. In linear algebra and geometry the prototype is orthogonal projection onto a subspace — least-squares regression projects the response onto the column space, principal-component analysis projects onto the leading eigendirections, Fourier projection decomposes a signal onto frequency basis functions — with idempotence as the formal signature.[1] In cartography a spherical Earth maps to a flat plane via a chosen projection, every projection preserving some properties (angle, area, shape) and necessarily distorting others, so the choice of projection is the choice of what to lose.[2] In graphics a 3D scene maps to a 2D image via a camera projection, with depth as the lost residual. In databases the select operation is literally called projection in relational algebra — drop the columns not in the target, keep the rows.[3]
In statistics and signal processing dimensionality reduction, feature selection, and sufficient statistics are projections of a high-dimensional sample onto a representation chosen to retain the inferentially relevant variation.[4] In optimization projected-gradient and projection-onto-convex-sets methods solve constrained problems by repeatedly projecting onto the feasible set, the constraint being the target and the projection picking the closest feasible point. In cognition the visual system projects 3D scenes onto a 2D retinal image, and concept-formation projects rich experience onto category labels.[5] In reporting and management a status report projects ongoing complex work onto a short summary chosen to answer the audience's likely question. Across all of these the same move is at work — collapse the dimensions perpendicular to a chosen target, keep those parallel to it, and recognize the residual as the load-bearing question of whether the projection was well chosen.
Clarity¶
Naming projection exposes a fact informal communication constantly hides: every reduced representation is the output of a projection choice, and that choice determines what the consumer can and cannot see. Two flat maps of the same Earth answer different questions — one preserves angle and serves navigation, the other preserves area and serves size comparison — and neither is the Earth. The vocabulary makes the direction of projection and the residual (what got dropped) first-class objects, which forces the consumer of any summary to ask the decisive question: what is being projected away, and is it the load-bearing variation for my decision? A summary that is faithful for one decision can be actively misleading for another, and only by naming the projection direction does this become visible.
Two further clarifications follow. First, projections compose, and iterated projection produces increasingly impoverished representations — a 3D scene to a 2D image to a single brightness scalar is two projections in series, each with its own residual — so naming each step surfaces exactly where a needed dimension was discarded. Second, projection has an algebraic signature: idempotence distinguishes a clean projection from arbitrary lossy compression, and this is the formal reason "summary of a summary equals the summary" holds only when the second direction agrees with the first. A reporting metric whose value keeps moving under re-summarization is, by this test, not a clean projection but a manipulation — and recognizing the idempotence failure is a precise diagnostic, not a vague suspicion. The clarity projection supplies is therefore both about what was dropped (the residual) and about whether the reduction was structurally honest (idempotence).
Manages Complexity¶
Projection compresses a high-dimensional object into a low-dimensional representation suitable for inspection, decision, or storage. The compression is large because the source typically has far more degrees of freedom than the decision needs; a well-chosen projection preserves the load-bearing variation in a few dimensions and discards the rest as noise. This is the structural reason a twelve-channel image, a hundred-variable dataset, or a quarter of operational reality can be reduced to a handful of numbers a decision-maker can actually hold in mind — the projection throws away exactly the dimensions that do not bear on the decision, by construction.
Projection also compresses cross-substrate similarity, which is a subtler complexity gain. Many systems with different surface forms — 3D shapes, time-series, text, audio — project onto the same low-dimensional manifold of inferentially relevant features, and the projection is what makes them comparable at all. The choice of projection direction is how comparability is engineered: aligning the directions of two different sources onto a common target is what lets them be measured against each other. The reduction is therefore not only in the size of each representation but in the dimensionality of the comparison problem: instead of comparing two rich objects feature by feature, one projects both onto a shared target and compares the projections. A reasoner who recognizes that disparate things have been made comparable by a shared projection also knows where to look when the comparison misleads — at the residuals each projection discarded, which may differ between the two sources and break the comparability the shared direction promised.
Abstract Reasoning¶
The projection skeleton supports several lines of reasoning. Loss is structural, not incidental: every summary, chart, and report has a residual, and naming it shifts the reader's posture from "I have the picture" to "I have a picture from one direction" — the orthogonal projection's residual being exactly the input minus its image, the formal version of "what got left out." Direction-dependence: the same source yields different projections under different directions, which is the structural content of perspective, framing, viewpoint, and choice of summary statistic — in each case the choice of direction is the load-bearing act. Idempotence as a sanity check: a genuine projection is stable under re-summarization, so a representation that changes when re-summarized was not a projection but a richer transformation. Reconstruction error: the projection's quality is the magnitude of the residual under the source's metric — variance-explained, fit quality, signal-to-noise are all residual-magnitude scores. Section and lift: the partial inverse of projection — choosing a representative for each collapsed class, or recovering the source from the target plus side information — measures how much the projection destroyed.
The portable role-set is: the source (the richer object to be reduced), the target (the chosen subspace or representation), the projection direction (the equivalence relation collapsing source into target, what is treated as the same), the residual (the discarded orthogonal complement), the idempotence property (the algebraic signature), the reconstruction error (the residual's magnitude), and the lift or section (the partial inverse). A reasoner holding this role-set can look at a regression fit, a flat map, an executive summary, and a retinal image and ask the same structural questions: what is the source, what direction was chosen, what is in the residual, is the reduction idempotent, and how much error does it carry. The framing forecasts where reductions go wrong — when the residual carries the signal the decision needed — and supplies the check (idempotence) that distinguishes an honest reduction from a drifting one.
Knowledge Transfer¶
The structure ports across substrates as a transferable design discipline, and the discipline carries interventions. The principal-component insight that good summaries preserve the leading directions — the variation that matters for downstream decisions — transfers to any reduction problem: ask what variation the reader needs, project onto that direction first, and reserve the residual for footnotes. The map-projection insight that every projection trades off properties — angle versus area — transfers to dashboard design, where a metric optimized for trend visibility distorts magnitude comparison and one optimized for magnitude distorts rate; the structural lesson is that no projection preserves everything, and the design choice is which property is load-bearing for the decision the dashboard drives. The projected-gradient pattern — do an unconstrained move, then project back to the feasible set — transfers to any setting with hard constraints, recurring as clipping in machine learning, snapping to spec in engineering, and proposal-then-statutory-revision in policy.[6] And the notion of a sufficient statistic — one whose residual carries no further information about the parameter — transfers to reporting policy, identifying which fields are inference-sufficient for a decision and which are merely decorative.
A worked example anchors the transfer. A satellite produces a twelve-channel image of a field, each pixel a twelve-dimensional vector, and an analyst wanting to map crop stress computes a vegetation index — a projection of the spectral vector onto a single direction known to carry the stress signal, with the other ten channels' variation set aside as residual. The projection is informative because the chosen direction captures the load-bearing variation and the residual is mostly soil and atmospheric noise irrelevant to the decision; a different projection of the same image would be informative for a different decision (a fire-risk projection emphasizes a different channel ratio), because the image is one and the projections multiply. The same calculation underlies regression onto the column space, an executive status report projected onto three bullets aligned with the reader's open decisions, a city map projected onto a flat plane, and visual perception projecting a 3D scene onto a 2D image plus depth cues. What transfers is the full discipline: identify the source, choose the direction that captures the decision-relevant variation, name the residual, check idempotence, and measure the reconstruction error. A practitioner who has internalized projection in one domain arrives in the next already asking what was projected away and whether it was load-bearing — and already holding the design moves (project onto the right direction first, exploit the convex projection for constraints, prefer inference-sufficient summaries) that the structure supplies. That portability of design discipline and intervention together is what makes projection a canonical substrate-independent structural prime.
Examples¶
Formal/abstract¶
Least-squares regression is the projection prime in its purest algebraic form. The richer source is a response vector \(y \in \mathbb{R}^n\); the lower-dimensional target is the column space of the design matrix \(X\) — the subspace spanned by the predictors; the projection direction is orthogonal (perpendicular to that subspace under the Euclidean inner product). The fitted values are \(\hat{y} = X(X^\top X)^{-1}X^\top y = Py\), where \(P\) is the projection (hat) matrix.[7] Every role appears explicitly. The idempotence signature is literal and checkable: \(P^2 = P\), projecting the projection changes nothing, which is exactly the algebraic identity the prime names — and it is why "the fit of the fitted values equals the fitted values." The discarded residual is \(y - \hat{y}\), orthogonal to the target by construction, and the reconstruction error is its squared norm, the residual sum of squares that scores how much variation the predictors failed to capture.[7] The intervention this licenses is sharp and structural: if the residual is large and structured (patterned, not noise), the target subspace was wrong — the load-bearing variation lay in a direction the predictors do not span — and the fix is to enrich \(X\) with the missing direction, not to re-fit. What the reasoner newly sees is that regression is not curve-fitting but the orthogonal projection of data onto a model subspace, with residual analysis as the direct read-out of what the model threw away.
Mapped back: the response vector, the column-space target, the orthogonal direction, the hat matrix, and the residual instantiate source, target, direction, idempotence, and reconstruction error; \(P^2=P\) is the prime's idempotence signature realized exactly.[1]
Applied/industry¶
A cartographer, a remote-sensing analyst, and an executive reporting team are all making projection choices and living with their residuals. The cartographer maps a curved Earth (source) onto a flat plane (target): a Mercator projection preserves angles for navigation but grossly distorts area near the poles, while an equal-area projection preserves size but distorts shape — the prime's lesson that no projection preserves everything is the literal content of the discipline, and the direction chosen is dictated by whether the decision is "steer a heading" or "compare land areas."[2] The remote-sensing analyst runs the same operation on a twelve-channel satellite image: each pixel is a twelve-dimensional vector projected onto a single vegetation-index direction known to carry crop-stress signal, with the other channels set aside as residual (soil, atmosphere) — and a different decision (fire risk) demands a different projection of the same image, because the source is one and the projections multiply.[4] The reporting team projects a quarter of operational reality onto three bullets aligned with the reader's open decisions: the direction is "what does this audience need to decide?", and the prime's diagnostic — name the residual, check idempotence — flags the manipulation where a metric's value keeps drifting under re-summarization (a non-idempotent "projection" that is really a moving target).
Mapped back: cartography, remote sensing, and management reporting are three genuine domains where the same roles operate — richer source, chosen target, projection direction, named residual — and the recurring interventions (choose the direction the decision needs; inspect the residual for lost signal; check idempotence for honesty) transfer intact.
Structural Tensions¶
T1 — Target Captures Signal versus Residual Carries It (the direction can be wrong). A projection is informative only when the chosen target captures the load-bearing variation and the residual is noise; it is actively misleading when the direction was poorly chosen and the discarded residual held the signal the decision needed. The characteristic failure mode is trusting a clean-looking summary whose residual was structured, not random — a regression fit with patterned residuals, a dashboard metric that hides the variation that mattered. Diagnostic: inspect the residual for structure; if what was thrown away is patterned rather than noise, the projection direction was wrong and enriching the target (not re-fitting) is the fix.
T2 — One Picture versus The Object (direction-dependence hides choice). Every reduced representation is the output of a projection choice, yet consumers routinely treat a summary as the thing itself. The same source yields different, equally valid projections under different directions — a Mercator and an equal-area map are both "the Earth" and neither is. The failure mode is mistaking "a picture from one direction" for "the picture," then making a decision the chosen direction was never optimized for (steering by an equal-area map, comparing land areas on Mercator). Diagnostic: ask what direction produced this representation and whether it is the load-bearing one for my decision; a summary faithful for one decision can mislead another.
T3 — No Projection Preserves Everything (the property trade-off). A projection can preserve angle or area, trend or magnitude, but not all at once — fixing one property necessarily distorts another. The tension is that there is no neutral, all-preserving reduction; choosing the target is choosing what to sacrifice. The failure mode is demanding a single summary that serves every purpose, or optimizing a metric for one property (trend visibility) and reading it for another (magnitude comparison) it actively distorts. Diagnostic: ask which property is load-bearing for the decision the representation drives, and accept that the others are distorted; a representation claimed to preserve everything is concealing its residual.
T4 — Idempotent Projection versus Drifting Transformation (the honesty check). A genuine projection is idempotent — projecting twice equals projecting once — so "summary of a summary equals the summary" when directions agree. A representation whose value keeps moving under re-summarization is not a clean projection but a richer, drifting transformation, often a manipulation. The failure mode is treating a non-idempotent reduction as a stable projection, building on a number that shifts each time it is recomputed. Diagnostic: re-apply the reduction and check whether the result is stable; idempotence failure is a precise signal that what looked like a clean projection is a moving target, not a vague suspicion.
T5 — Single Projection versus Composed Chain (cumulative loss). Projections compose, and an iterated chain — 3D scene to 2D image to one brightness scalar — produces increasingly impoverished representations, each step adding its own residual. The failure mode is reasoning about the final output as if it were one honest reduction of the source, when several projections in series each silently discarded a dimension, and the dimension the decision needed was dropped at an invisible intermediate step. Diagnostic: trace the chain of reductions, naming each step's residual; if a needed dimension is absent from the output, locate which projection in the series discarded it rather than blaming the source.
T6 — Projection versus Perspective (the framing boundary). Projection is the substrate-neutral act of mapping along a direction; its nearest neighbour perspective carries the viewpoint-laden, often human-framed sense of the same move. The tension is at the boundary: treating a substantive perspective (whose direction encodes interests, values, or framing) as if it were a neutral geometric projection hides the load-bearing choice in the direction. The failure mode is presenting an interest-laden summary as an objective reduction, the projection direction smuggling in a viewpoint while claiming neutrality. Diagnostic: ask whether the projection direction is mechanically given or is itself a contestable choice of viewpoint; if the direction encodes interests, it is a perspective wearing projection's neutral clothing, and the choice must be argued, not assumed.
Structural–Framed Character¶
Projection sits at the structural pole of the structural–framed spectrum, and every diagnostic points one way. The pattern is a geometric-algebraic operation — map a richer source onto a lower-dimensional target along a chosen direction, collapsing what is perpendicular and preserving what is parallel — with idempotence as its algebraic signature.
The pattern carries no home vocabulary that must travel with it: the same reduction is told in each domain's own words as the shadow of a solid, a flat map of a curved country, a summary statistic, or an executive summary, with the geometric skeleton (source, target, direction, residual, idempotence) shared rather than imported. It carries no inherent approval or disapproval — a projection is neither good nor bad; whether the reduction is informative or misleading depends only on whether the residual carried the signal, a value-neutral structural fact. Its origin is formal, drawn from geometry and linear algebra, owing nothing to any human institution. It runs indifferently across physical, optical, statistical, and abstract substrates, requiring no human practice to exist. And to invoke a projection is to recognize a dimension-collapsing map already operating — to name the direction and the discarded residual — not to import an interpretive frame. On every criterion it reads structural, exactly the 0.0 aggregate the frontmatter assigns.
Substrate Independence¶
Projection earns a maximal composite 5 / 5 on the substrate-independence scale: the map-a-richer-source-onto-a-lower-dimensional-target-along-a-chosen-direction operation is recognized, not translated, wherever dimensions are deliberately collapsed. The domain breadth is total — the same move is orthogonal projection, least-squares regression, PCA, and Fourier decomposition in linear algebra, the flat map of a curved Earth in cartography, the camera projection in graphics, the select operation in relational databases, dimensionality reduction and sufficient statistics in statistics, projection-onto-convex-sets in optimization, and the retinal image and category label in cognition — so the pattern operates with identical structural force across geometric, cartographic, statistical, computational, and cognitive substrates. The structural abstraction is complete: the signature commits to nothing about the medium, asserting only a source, a target, a direction, a residual, and the idempotence identity, so its derived discipline (name the residual, check idempotence, score the reconstruction error) follows purely from the geometric structure with no domain-specific commitment to carry. The transfer evidence is concrete and signature-bearing rather than analogical: idempotence (\(P^2 = P\)) is the literal, checkable algebraic identity that recurs verbatim from the regression hat matrix to map projections to executive summaries, and the same design discipline (choose the direction the decision needs, inspect the residual for lost signal, exploit convex projection for constraints) carries identically across remote sensing, cartography, and management reporting — named instances where one operation governs many fields. Nothing pins the prime to a medium; the substrate is exactly what the dimension-collapsing map abstracts away.
- Composite substrate independence — 5 / 5
- Domain breadth — 5 / 5
- Structural abstraction — 5 / 5
- Transfer evidence — 5 / 5
Relationships to Other Abstractions¶
Current abstraction Projection Prime
Parents (1) — more general patterns this builds on
-
Projection is a kind of, typical Abstraction Prime
'Projection is the precise, geometric, idempotent SPECIAL CASE within the broader family of reductions that abstraction names.' Projection drops the dimensions perpendicular to a chosen target (vs abstraction's free-form essence-extraction).Abstraction supplies the genus: Focus on core elements. Projection preserves that general structure while adding its differentia: Map a richer object onto a lower-dimensional target along a chosen direction, discarding the rest. The parent can occur without those added commitments, whereas removing the parent structure leaves no basis for classifying the child as this subtype. That asymmetry establishes subsumption rather than mere association. The typical qualifier limits the claim to the characteristic route, not a constitutive requirement of every instance; exceptions must retain the child's identity through another mechanism.
Children (24) — more specific cases that build on this
-
3D projection Domain-specific is a kind of Projection
The proposed strict upward parent is
prime:projection.prime:projection is the nearest broader Prime; the source domain and invariant supply the autonomous residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while 3D projection adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by each image position is produced by the declared projection geometry and viewing parameters, and depth ambiguity and distortion are treated as consequences rather than hidden identity It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of 3D projection. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge toprime:projection. No live DAG mutation is authorized. -
Antisymmetrizer Domain-specific is a kind of Projection
The proposed strict upward parent is
prime:projection.prime:projection is the nearest broader Prime; the source domain and invariant supply the autonomous residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Antisymmetrizer adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by the particle number and one-particle space, permutation action, sign and normalization, operator domain, idempotence convention, resulting exchange law and Pauli interpretation are explicit It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Antisymmetrizer. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge toprime:projection. No live DAG mutation is authorized. -
Constant Term Domain-specific is a kind of Projection
The accepted reference-grade review places Constant Term under Projection because the child instantiates or depends on the parent's broader structure while retaining its own constitutive identity.The coefficient of the multiplicative-identity monomial in a polynomial, series, or Laurent expression—the component independent of every declared variable and recoverable by evaluation at zero only when negative powers are absent. The parent is defined more broadly: Map a richer object onto a lower-dimensional target along a chosen direction, discarding the rest.
- Descriptive Geometry Domain-specific is a kind of Projection
Descriptive Geometry strictly instantiates **Projection**.Its defining move repeatedly maps a spatial configuration to planar targets along selected directions, accepting the loss of one coordinate per view and recovering relevant information through coordination. It also relates to **Representation**, because drawing conventions specify how spatial points and relations correspond to marks on a plane; **Viewpoint**, because changing the viewing direction changes which lengths or shapes are exposed; and **Problem Representation**, because solving often depends on selecting the view in which the relation becomes elementary. These related primes clarify the mechanism but are not needed as additional prospective parents. The minimal proposed DAG parent is `prime:projection`.
- Elimination theory Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.Elimination literally maps a richer polynomial solution object onto retained coordinates while discarding selected variables; ideal-theoretic consequences and closure semantics form the domain-specific specialization. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Elimination theory adds domain-specific constraints. The entry does not collapse into that parent because the polynomial-ideal operation that removes chosen variables while retaining exact algebraic consequences and its geometry-of-projection interpretation, rather than generic symbolic simplification or one named algorithm It also declines the closest thematic catalog neighbor: the neighbor does not literally subsume the constitutive identity of Elimination theory. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Hammer Projection Domain-specific is a kind of Projection
Hammer is a strict domain-specific instance of **Projection**.It maps a two-dimensional curved source surface described by angular coordinates into a two-dimensional plane, necessarily discarding some geometric relations while retaining a selected one. Although source and target share dimension, the prime's broader representational reduction remains appropriate: the spherical geometry cannot be carried isometrically into the plane, so the construction selects area for preservation. It also relates conceptually to Conservation, Symmetry, Transformation, and Tradeoff. Equal area is the conservation law; bilateral symmetries support reasoning and testing; the equations are a coordinate transformation; and the distortion field embodies a design tradeoff. Those relations are explanatory rather than additional minimal DAG parents. The proposed parent set therefore contains only `prime:projection`. If a dedicated live Map Projection or Equal-Area Projection node is later accepted, locality should be reviewed before implementation.
- Isometric Projection Domain-specific is a kind of Projection
**Projection** is the strict parent by specialization.Isometric Projection is a lower-dimensional directional map with a discarded depth component and adds exact symmetry constraints. The parent applies without technical drawing or equal axis scale. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Legendre moment Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.prime:projection is the nearest broader Prime; the source domain and invariant supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Legendre moment adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by coordinate normalization, polynomial convention, order, quadrature or discrete approximation, and coefficient normalization are fixed and the descriptor is computed as the declared projection It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Legendre moment. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Local maximum intensity projection Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.LMIP literally maps a richer three-dimensional scalar volume onto a two-dimensional view along chosen rays while discarding most samples; first-qualified-peak selection supplies the autonomous residual. The edge is proposal-only and points to a frozen prior-baseline Prime. The entry does not collapse into the parent because the first-thresholded-local-maximum selection rule along each ordered ray, not global maximum projection, front-to-back opacity compositing, surface extraction, clinical diagnosis, or a generic image threshold A thematic neighbor is declined whenever it does not literally subsume that rule. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Momentum mapping format Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.prime:projection is the nearest broader Prime while the source-domain carrier and invariant supply the autonomous residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Momentum mapping format adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity fixed by the MPM variant, material points and background grid, shape functions and support, particle mass velocity and affine state, particle-to-grid momentum rule, grid update, grid-to-particle rule, PIC FLIP or APIC convention, conservation, dissipation and stability evidence are explicit It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Momentum mapping format. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Oblique Mercator projection Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.It maps a curved geographic carrier conformally to a plane; oblique axis selection supplies the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Oblique Mercator projection adds domain-specific constraints. The entry does not collapse into that parent because arbitrarily oriented zone mapping for elongated diagonal regions It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Oblique Mercator projection. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Pole figure Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.prime:projection is the nearest broader Prime while the source-domain carrier and invariant supply the autonomous residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Pole figure adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity fixed by the specimen and coordinate frame, crystal phase and symmetry, plane or direction family, hemisphere and antipodal convention, projection type, acquisition method, intensity normalization, angular resolution, coverage and uncertainty are explicit It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Pole figure. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Projection filters Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.The method replaces dynamics by their projection onto an admissible submanifold; nonlinear conditional-density estimation supplies the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Projection filters adds domain-specific constraints. The entry does not collapse into that parent because information-geometric finite-dimensional closure of nonlinear Bayesian filtering rather than generic moment truncation It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Projection filters. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Projective line Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.prime:projection is the nearest broader Prime; the source domain and invariant supply the autonomous residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Projective line adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by the base field, division ring, ring or scheme, two-dimensional module, equivalence relation on nonzero pairs, homogeneous-coordinate convention, affine charts and transition, point or points at infinity, incidence, and automorphism group are explicit It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Projective line. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Ridge function Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.A ridge function literally discards all input variation except one chosen projected coordinate; exact factorization through a univariate profile supplies the autonomous approximation-theory residual. The edge is proposal-only and points to a frozen prior-baseline Prime. The entry does not collapse into the parent because the exact one-projection factorization and its constant-hyperplane geometry, not any elongated graph, arbitrary low-rank array, separable product, radial function, or sum of multiple directional components A thematic neighbor is declined whenever it does not literally subsume that rule. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Riesz projector Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.prime:projection is the nearest broader Prime; the source domain and invariant supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Riesz projector adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by the operator domain, resolvent contour, orientation, sign convention, isolated spectral set, and boundedness assumptions are explicit and the contour crosses no spectrum It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Riesz projector. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Transverse Mercator projection Domain-specific is a kind of Projection
The proposed strict upward parent is `prime:projection`.It projects a curved geographic carrier into planar coordinates; transverse conformal zoning supplies the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Transverse Mercator projection adds domain-specific constraints. The entry does not collapse into that parent because a zone-oriented conformal Earth-to-plane mapping underlying UTM and many national grids It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Transverse Mercator projection. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:projection`. No live DAG mutation is authorized.
- Wiechel Projection Domain-specific is a kind of Projection
Wiechel is a strict domain-specific instance of **Projection**.It maps a curved spherical surface into a planar representation, selecting area and one path-scale property for preservation while discarding azimuthality, conformality, and general distance fidelity. The exact transformation is not entailed by the live prime. It also relates to Conservation, Transformation, Symmetry, Measurement, and Tradeoff. Area is conserved; a coordinate transformation performs the mapping; rotational organization structures the graticule; scale supplies measurement diagnostics; and preservation of area is traded against other geometric properties. These are explanatory relations rather than additional minimal DAG parents. The proposed parent set contains only `prime:projection`. If a live Map Projection, Equal-Area Projection, or Pseudoazimuthal Projection node later exists, parent locality should be reconsidered before implementation.
- Empirical No-Failure Anchor Prime is a kind of Projection
The anchor is a projection specialized to reducing the full test-response dataset to the maximal tested level that cleared a stated failure criterion.Both map a richer source onto a lower-dimensional target by a fixed direction, discard a residual that must remain named, and are stable when reapplied to the same source under the same rule. The child fixes the source to an ordered test-response dataset, the target to one scalar input level, the direction to maximal non-detected failure, and the residual to untested or discarded boundary information.
- Perspective Prime is a kind of, typical Projection
Perspective is typically a specialization of Projection, retaining the parent's defining structure while adding the child's specific commitments.Projection supplies the genus: Map a richer object onto a lower-dimensional target along a chosen direction, discarding the rest. Perspective preserves that general structure while adding its differentia: Representation of depth. The parent can occur without those added commitments, whereas removing the parent structure leaves no basis for classifying the child as this subtype. That asymmetry establishes subsumption rather than mere association. The typical qualifier limits the claim to the characteristic route, not a constitutive requirement of every instance; exceptions must retain the child's identity through another mechanism.
- Depth Perception Domain-specific presupposes Projection
**Projection — presupposes.** Optical imaging maps three-dimensional scene structure onto two-dimensional retinal surfaces and omits an explicit depth coordinate.Depth perception exists because the observer must recover useful spatial structure from that loss.
- Fourier–Motzkin Elimination Domain-specific presupposes Projection
**`prime:algorithm` (proposed primary parent, strict subsumption).** Fourier–Motzkin is a definite input-to-output procedure with correctness and resource semantics.It specializes Algorithm to sign-partitioned elimination of linear-inequality variables and exact projection. Removing those additions leaves an algorithm; removing procedure leaves no Fourier–Motzkin method.
- Kinetic depth effect Domain-specific is part of Projection
Kinetic depth contains the projection that maps a richer 3D object to each lower-dimensional 2D retinal frame and discards depth.The projection creates the static many-to-one ambiguity; temporal samples, rigidity, and perceptual recovery add the child's differentia. Projection supplies an internal constituent: Map a richer object onto a lower-dimensional target along a chosen direction, discarding the rest. Kinetic depth effect requires that role within this mechanism: The perceptual recovery of vivid 3D structure from the 2D motion of a stimulus that looks flat when static — the visual system solving an underdetermined inverse problem by adding motion as constraint under a rigidity prior. Remove the parent-role and the child loses a required internal operation, even though the parent can exist outside the child. The child is therefore built from the parent rather than being a taxonomic kind of it.
- AD–AS Model Domain-specific is a decomposition of Projection
Removing the macroeconomic frame from AD-AS leaves a projection from a high-dimensional system onto a chosen two-dimensional target with an explicit residual.The child maps households, firms, policy, trade, prices, and expectations onto price-level by real-output space, retaining the two aggregate schedules and discarding microfoundations, magnitudes, and adjustment paths. Reapplying the same mapping does not create new dimensions, and the entry makes the discarded content and reconstruction limits first-class rather than claiming the plane is the economy.
Hierarchy path (1) — routes to 1 parentless root
- Projection → Abstraction
Neighborhood in Abstraction Space¶
Projection sits in a moderately populated region (40th percentile for distinctiveness): it has near-neighbors but no dense thicket of synonyms.
Family — Formal Structure & Logical Proof (13 primes)
Nearest neighbors
- Preimage — 0.77
- Bijectivity — 0.72
- Bias — 0.72
- Proof By Contradiction — 0.71
- Isomorphism — 0.71
Computed from structural-signature embeddings · 2026-09-10
Not to Be Confused With¶
Projection must be distinguished from perspective, its nearest neighbour and the prime it most readily disguises itself as (or is disguised by). The two share the same geometry — both reduce a richer object by privileging some directions over others — but they differ in the character of the direction. Projection is the substrate-neutral, mechanical act: a direction is given (the orthogonal complement of a subspace, the light source casting a shadow, the camera axis), and the operation collapses what is perpendicular to it. Perspective is the viewpoint-laden version, where the direction itself encodes interests, values, or framing — a chosen vantage that could have been otherwise and whose choice is contestable. The danger is in both directions. Treating a perspective as a neutral projection smuggles a viewpoint in under the guise of an objective reduction: an "executive summary" projected onto the three bullets the author wanted foregrounded is presented as a faithful compression when its direction encoded an agenda. Conversely, treating a genuinely mechanical projection as a perspective over-reads contestability into a direction that was in fact fixed by the problem. The diagnostic is whether the direction is mechanically given or itself a choice — if it encodes interests, it is a perspective wearing projection's neutral clothing, and the choice must be argued rather than assumed.
A second genuine confusion is with abstraction, because both produce a simpler object from a richer one. The distinction is what gets dropped and how. Abstraction discards detail to retain an essence — it can ascend levels (instances to a category, mechanisms to a principle) and keeps whatever is judged essential, with no requirement that what is kept and what is dropped be geometrically complementary or that the operation be idempotent. Projection drops specifically the dimensions perpendicular to a chosen target while preserving those parallel to it, and it carries the algebraic signature of idempotence: projecting a projection changes nothing. Abstraction has no such signature — abstracting an abstraction generally yields a still-more-abstract object, not the same one. The error is to treat a projection's structured, direction-indexed, residual-naming discipline as if it were free-form abstraction (losing the ability to inspect the residual for the signal it discarded), or to treat a genuine abstraction as a projection (expecting idempotence and a clean orthogonal residual where none exists). Projection is the precise, geometric, idempotent special case within the broader family of reductions that abstraction names.
These distinctions matter because each isolates a different question. Projection-versus-perspective asks whether the reducing direction is mechanical or interest-laden — and so whether the choice must be argued; projection-versus-abstraction asks whether the reduction is the structured, idempotent, residual-bearing kind or a free-form essence-extraction. A practitioner who keeps them straight checks idempotence to confirm a reduction is an honest projection, inspects the named residual for lost signal, and asks whether the direction encodes a viewpoint before accepting any summary as neutral.
Solution Archetypes¶
Solution archetypes in the catalog that build on this prime — directly (this prime is a source ingredient) or as a related prime.
Built directly on this prime (4)
- Event-Log-Centered Modeling: Preserve happenings as the primary record and derive entity state, relationships, places, periods, timelines, and summaries as reproducible projections of the governed event log.▸ Mechanisms (18)
- Append-Only Event Store — An immutable, ordered store that only ever accepts new events and never edits old ones, serving as the single source of truth from which all state is derived.
- Bitemporal Event Register — Records every fact along two clocks — when it happened and when the system came to know it — with the source of each assertion, so you can ask what was believed as of any past moment.
- Compensating-Event Correction — Corrects a mistaken event not by editing it but by appending a new reversing or adjusting event, so the erroneous record and its correction both remain in the history.
- Deterministic Replay Protocol — Reconstructs a past state or sequence by re-applying the same events in the same order through the same logic, so the rebuild is reproducible down to the last detail.
- Entity-Trajectory Projection — Derives one entity's path through time by gathering every event it took part in — resolving its identity across records and stitching cross-referenced layers into a single ordered trajectory.
- Event Capture Template — A standard shape for recording a happening — its type, what changed, who took part, and where — so a raw occurrence becomes a well-formed, self-describing event rather than a bare timestamped row.
- Event Knowledge Graph — Materializes the event log as a queryable graph, linking events, participants, and entities across layers with typed participation and causal-or-correlation edges.
- Event Replay Deduplication — Lets a consumer process an at-least-once event stream safely by keying on stable event identifiers, so a redelivered or replayed message never applies its effect twice.
- Event-Sourced Projection — Builds a read-optimized view by folding an append-only log of events, so the same history can be replayed to produce many views — or rebuild any of them from scratch.
- Log Compaction — Reclaims space by keeping only the latest or still-necessary record per key and discarding superseded history, under a retention policy that must never break the ability to rebuild state.
- Periodization Projection — Derives named periods from the event log by cutting the timeline at the transformations that mark one regime turning into the next.
- Place-History Projection — Assembles the full history of a place by gathering every event bound to it into one time-ordered account, resolving the many names a single place goes by.
- Process Mining / Trace Analysis — Reconstructs the real process from event traces — discovering the actual control flow, its variants, and where reality deviates from the intended path — that the log reveals but no diagram admits.
- Projection Rebuild and Diff — Rebuilds a projection from the log and diffs it against the live view, treating any disagreement as evidence the view is wrong, never the log.
- Projection-Frontier Dashboard — Shows how far each projection has consumed the log, turning invisible replication lag and coverage gaps into watched, actionable numbers.
- Provenance-Weighted Event Reconciliation — Resolves conflicting, duplicate, and late event claims by weighting each by the trustworthiness of its source, while keeping the disagreement on the record.
- Snapshot Plus Replay — Rebuilds current state fast by starting from a periodic snapshot and replaying only the events since, instead of the whole history.
- Versioned Event-Schema Registry — Versions event type contracts so producers and projections can evolve their schemas without silently breaking each other or the old history.
- Invariant-Mode Decomposition Design: Find the directions a transformation preserves as directions, measure how strongly it stretches or damps each one, and use those modes to prioritize explanation, control, compression, and monitoring.▸ Mechanisms (12)
- Eigendecomposition Workflow — Takes an explicitly known linear operator and returns its complete set of invariant directions together with the scalar gain of each — the full modal picture the rest of the analysis reads from.
- Modal Sensitivity Sweep — Perturbs each mode's gain or coordinate in turn to see which ones actually move the outcomes you care about — turning a raw spectrum into a ranked map of where intervention has leverage, and exposing where modes bleed into one another.
- Modal Stability Analysis — Classifies each mode as growing, decaying, oscillating, or steady under repeated transformation, splitting the spectrum into a stable set and an unstable set — a verdict that holds only inside the linearized regime it was taken in.
- Mode-Shape Testing — Recovers a system's modes empirically — by exciting or observing the real thing and reading its response — for cases where no operator matrix exists to decompose, and pins down the conditions under which the measured modes actually hold.
- Network Spectral Centrality Analysis — Treats a network's connectivity as the transformation and reads the entries of its dominant eigenvector as node importance — ranking who sits in the network's dominant mode, and therefore where structural intervention bites.
- Power-Iteration Probe — Recovers just the single dominant mode of a transformation by applying it to a trial vector over and over — never forming or factoring the whole operator — and reads its own convergence rate off the spectral gap.
- Principal Component Analysis — Finds the orthogonal directions of greatest variance in a cloud of data, turning many correlated measurements into a few uncorrelated modes ranked by how much they explain.
- Reduced-Order Model — A small, runnable surrogate that keeps only a system's dominant modes, so its behaviour can be simulated, controlled, or explored in real time within the regime where the reduction holds.
- Residual Reconstruction Test — Rebuilds the original system from only the modes you kept and measures what is left over, turning 'how many modes are enough?' into a number you can hold to a tolerance.
- Singular Value Decomposition — Factors any rectangular or non-normal mapping into paired input and output directions linked by non-negative gains, so even transformations that have no clean eigenvectors still get a modal decomposition.
- Spectral Decomposition Report — A written account that turns a raw decomposition into a shared, bounded interpretation — which modes matter, what each may and may not be read to mean, and where independence between them breaks down.
- Spectral Gap Monitor — Tracks the separation between the dominant modes and the rest over time, raising an alarm when the gap narrows enough that a modal simplification can no longer be trusted.
- Perceived-Consensus Calibration: Before acting on “everyone thinks this,” separate the speaker’s local anchor from the target population and replace perceived consensus with representative, independent, and distributional evidence.▸ Mechanisms (9)
- Anonymous Belief Pre-Poll — A private pre-poll that captures each person's independent view before social influence can manufacture agreement.
- Belief Distribution Dashboard — A standing display that shows the spread, subgroups, and unknowns behind a consensus claim — and tracks them against what actually happened.
- Consensus Claim Evidence Log — A written record that pins each 'everyone thinks X' claim to its exact population and its actual source, so projection can't hide as fact.
- False-Consensus Premortem — A pre-decision exercise that assumes the 'everyone agrees' belief was wrong and traces backward to how the team's own view got mistaken for the world's.
- Minority Report Prompt — A fill-in template attached to any consensus decision that keeps the strongest dissenting view — and who holds it — visibly bound to the claim.
- Nonrandom Sample Audit — A checklist that interrogates who the visible sample actually is — and who it silently leaves out — before their agreement is read as the population's.
- Outgroup or Edge-Case Interview — A targeted qualitative procedure that goes and talks to the people least like the decision-makers, to find where local projection breaks.
- Representative Consensus Survey — A survey procedure that draws a sample matched to the target population, so a prevalence claim can be estimated with a stated margin instead of assumed.
- Silent-Start Estimation Round — A meeting ritual where everyone commits an estimate in writing before anyone speaks, so the first voice can't anchor the room into false agreement.
- Preimage Set Characterization: Given an output condition, identify and bound the complete set of inputs that could produce it before acting as if the output has a unique source.▸ Mechanisms (10)
- Collision Analysis Matrix — Cross-tabulates inputs against outputs to expose where distinct inputs collide on the same output and where the mapping's uniqueness fails.
- Constraint-Solver Backsolve — Encodes the output condition and domain as constraints and derives the complete set of inputs that satisfy them, with a guarantee that none is missed.
- Coverage Completeness Audit — Maps the union of the patches against the declared domain to prove no in-scope region is left unwitnessed, and logs every gap it finds.
- Fiber Cardinality Count — Reports how many inputs map to each output — the size of the fiber — along with how much to trust that number.
- Inverse Lookup Query — Answers an output back to its inputs by querying a reverse index, returning every input already filed under the target value.
- Output-to-Input Traceback Map — Traces an observed output back through the mapping to the input states compatible with it, naming what the forward projection discarded and how to act while the ambiguity stands.
- Predicate Satisfaction Filter — Runs a stated membership predicate over the whole input population, keeping exactly the cases that satisfy the output condition and flagging the ones sitting on the threshold.
- Preimage Table — Publishes the finished output-to-input sets as a static reference so downstream users read the preimage off the page instead of re-deriving it, with usage caveats printed alongside.
- Sensitivity-to-Mapping-Change Review — Perturbs the mapping, threshold, or parameters and watches which inputs enter or leave the preimage, exposing how fragile the set is and warning downstream users where it will move.
- Witness and Counterexample Set — Collects concrete inputs proven to belong to the preimage (witnesses) and inputs that refute a claimed uniqueness or completeness (counterexamples), building the set from confirmed exhibits rather than sweeps.
Also a related prime in 7 archetypes
- High-Dimensional Tractability Control: Treat added dimensions as a qualitative regime change: test whether coverage, distance, search, and generalization still work, then impose a defensible dimension budget, structure assumption, reduction, or regularization strategy.
- Identity-Resonant Persuasion Design: Create ethical self-recognition between audience and appeal so alignment feels identity-consistent rather than externally imposed.
- Independent Generating Set Design: Define the space and combination rules, then choose the smallest independent set of generators that covers it completely and yields stable, unique, transformable coordinates.
- Local-Chart Atlas Modeling: Use overlapping local maps when one global map distorts the terrain: model locally, stitch through verified transition rules, and monitor global consistency.
- Other-Agent State Model Calibration: Model another agent as having its own partial knowledge, goals, attention, constraints, and interpretations, then update that model from evidence before routing action through it.
- Perspective Depth Projection Design: Fix the observer and projection relation, construct depth through convergence, scale, foreshortening, overlap, and atmosphere, and disclose where the chosen viewpoint distorts or hides spatial truth.
- Structure-Preserving Embedding Design: Embed a source system into a richer host so the source remains distinguishable, structurally faithful, and usable inside the host rather than merely translated or compressed.
References¶
[1] Strang, Gilbert. Introduction to Linear Algebra. 5th ed. Wellesley-Cambridge Press, 2016. Develops orthogonal projection onto a subspace, the projection (hat) matrix, and the idempotence identity P^2 = P. registry ↩a ↩b
[2] Snyder, John P. Map Projections: A Working Manual. U.S. Geological Survey Professional Paper 1395. Washington: USGS, 1987. Standard reference showing every map projection preserves some properties (angle, area, shape) while necessarily distorting others. registry ↩a ↩b
[3] Codd, E. F. "A Relational Model of Data for Large Shared Data Banks." Communications of the ACM 13, no. 6 (1970): 377–387. Defines the relational model whose projection operation drops attributes not in the target. registry ↩
[4] Jolliffe, I. T. Principal Component Analysis. 2nd ed. New York: Springer, 2002. Treats dimensionality reduction as projection onto leading eigendirections that retain the load-bearing variation. registry ↩a ↩b
[5] Marr, David. Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. San Francisco: W. H. Freeman, 1982. Frames vision as recovering scene structure from a 2D retinal projection of the 3D world. registry ↩
[6] Boyd, Stephen, and Lieven Vandenberghe. Convex Optimization. Cambridge: Cambridge University Press, 2004. Develops projection onto convex sets and projected-gradient methods for constrained optimization. registry ↩
[7] Draper, Norman R., and Harry Smith. Applied Regression Analysis. 3rd ed. New York: Wiley, 1998. Presents least-squares regression as orthogonal projection onto the column space via the hat matrix, with the residual sum of squares scoring unexplained variation. registry ↩a ↩b