Quasi-arithmetic mean¶

Average values by mapping them through a continuous strictly monotone generator, taking an arithmetic mean in transformed coordinates, and mapping the result back.

Version: v1 · 2026-08-30 · History
Domain-specific #: 2599
Origin domain: mathematics
Subdomain: theory of means
Aliases: Kolmogorov–Nagumo mean, Generalized f-mean, Quasi-linear mean

Core Idea¶

For a nondegenerate real interval \(I\), a continuous strictly monotone generator \(f:I\to\mathbb R\), and inputs \(x_1,\ldots,x_n\in I\), the quasi-arithmetic mean generated by \(f\) is \(M_f(x_1,\ldots,x_n)=f^{-1}\!\left(n^{-1}\sum_{i=1}^{n}f(x_i)\right)\). Strict monotonicity makes the inverse well defined on \(f(I)\), while continuity and the interval hypothesis ensure that averaging transformed values stays in the range. Affine changes \(g=af+b\), with \(a\ne0\), generate the same mean, so the generator is a coordinate choice rather than unique data.^[1]

The construction first expresses every input on one monotone transformed scale, collapses those transformed values to their arithmetic average, and returns that summary to the original scale. The transformation changes how gaps are weighted without changing the order of inputs. Choosing the identity generator gives the arithmetic mean; logarithm on positive inputs gives the geometric mean; reciprocal-type generators give the harmonic mean; and power generators recover power means. The common mechanism is conjugation of arithmetic averaging, not a list of unrelated formulas.^[2]

The family is not every generalized mean. Weighted quasi-arithmetic means alter the equal weights; multivariate or vector-valued versions need extra invertibility structure; Bajraktarević and Cauchy means use additional functions; and medians or order statistics do not generally arise by averaging a transformed scale. Domain restrictions are load-bearing: logarithmic and reciprocal examples require positive inputs. A formula that uses a noninjective generator or averages outside its inverse's range does not define this abstraction as stated.^[3]

Structural Signature¶

Input interval. A common ordered domain makes transformed averaging and inversion meaningful.
Generator. A continuous strictly monotone function selects the coordinate in which arithmetic averaging occurs.
Transformed observations. Each input is mapped through the same generator before any collapse.
Equal-weight average. The arithmetic mean of transformed values supplies the aggregation step.
Inverse transformation. The summary returns to the original input scale.
Affine equivalence. Generators differing by a nonzero affine change define the same mean.
Internality. The result remains between the smallest and largest input under the stated hypotheses.
Characterization axioms. Symmetry, continuity, strict monotonicity, idempotency, and decomposability identify the family under standard formulations.

What It Is Not¶

Not an arbitrary aggregation function. Many lawful aggregators cannot be represented by transformed arithmetic averaging.
Not a weighted mean by default. The defining displayed form assigns equal weight unless weights are explicitly added.
Not a power mean only. Power generators form an important subfamily but do not exhaust all monotone generators.
Not a uniquely parameterized object. Affine-equivalent generators name the same mean.
Not a mean on unrestricted inputs. Generator domains exclude some values, as logarithm excludes nonpositive inputs.
Not a numerical algorithm. The abstraction specifies a family of aggregation maps, not one evaluation procedure.

Scope of Application¶

The abstraction is literal wherever practitioners can identify the same constitutive roles, apply the same boundary tests, and obtain the same kind of output. The following habitats are uses of Quasi-arithmetic mean itself, not metaphors based only on resemblance.

Theory of means. Comparing mean families by invariance, monotonicity, and functional equations.
Measurement scales. Choosing an averaging coordinate compatible with how a quantity is represented.
Statistics. Expressing transformed-location summaries while retaining the original reporting scale.
Decision aggregation. Combining commensurable values under declared separability and transformation assumptions.
Information aggregation. Studying decomposable summaries generated by a scalar coordinate change.
Inequality theory. Ordering different generators and locating equality cases under adequate smoothness.

Clarity¶

A clear account of Quasi-arithmetic mean must preserve the recognition invariant stated in the Core Idea rather than rely on the title alone. Declare the interval, generator, monotonic direction, and whether weights are equal. Check the generator's inverse on the entire transformed average range. Treat affine-equivalent generators as the same mean rather than different catalog identities. Name domain exclusions before using logarithmic, reciprocal, or power special cases. These declarations are not editorial extras: each changes what observations count, which transformations are licensed, and what conclusion can be drawn. A reader should be able to reconstruct the input, the operative rule, the output, and at least one defeater from the account without consulting an implementation or guessing an unstated convention.

Manages Complexity¶

Quasi-arithmetic mean manages complexity by replacing a diffuse field of observations or possible operations with a bounded role structure: input interval supplies a common ordered domain makes transformed averaging and inversion meaningful.; generator supplies a continuous strictly monotone function selects the coordinate in which arithmetic averaging occurs.; transformed observations supplies each input is mapped through the same generator before any collapse.; equal-weight average supplies the arithmetic mean of transformed values supplies the aggregation step.; inverse transformation supplies the summary returns to the original input scale.. The compression is useful because it localizes disagreement. One can ask whether the input was properly formed, whether a constitutive relation held, whether an alternative explanation defeats the inference, or whether the output was overinterpreted. The same compression can mislead when its discarded detail is exactly what the decision requires. A reference-grade use therefore reports both the invariant retained and the information intentionally lost.

Abstract Reasoning¶

Verify that all inputs lie in one nondegenerate interval on which the proposed generator is continuous and strictly monotone.
Map every input through the same generator and preserve the input multiplicity or declared weights.
Compute the arithmetic average in transformed coordinates.
Apply the inverse generator and verify that the result returns to the original interval.
Test idempotency by setting every input equal and symmetry by permuting the inputs.
Compare a second representation through affine equivalence before claiming two different means.
Separate a formula-level example from an axiomatic characterization of the whole family.
Test the candidate interpretation against the nearest named confusable rather than accepting a shared surface feature.
State the conclusion at the same scope as the source conditions, and retain uncertainty or nonuniqueness where the construct does not remove it.

Knowledge Transfer¶

The strict upward abstraction is Aggregation. Quasi-arithmetic Mean instantiates Aggregation because it deliberately collapses several scalar inputs to one internal summary, specialized by arithmetic averaging in a monotone transformed coordinate. Within theory of means, the full mechanism transfers literally when the same roles and boundary tests recur. Beyond that domain, only the parent-level skeleton should travel. Reusing the label Quasi-arithmetic mean after removing its constitutive vocabulary would hide a change of mechanism behind an analogy. The honest transfer rule is therefore two-stage: recognize the domain-specific pattern first, then lift only the parent relation that remains invariant under a substrate change.

Examples¶

Canonical¶

For positive \(x_1,\ldots,x_n\), choose \(f(x)=\log x\). Then \(M_f=\exp(n^{-1}\sum_i\log x_i)=(\prod_i x_i)^{1/n}\), the geometric mean. The example is not a coincidental formula match: logarithm converts multiplication into addition, arithmetic averaging occurs on the log scale, and exponentiation restores the original positive scale. Replacing \(f\) by \(2\log x+7\) leaves the result unchanged because the added constant and nonzero scale cancel through averaging and inversion.

Mapped back: input and conventions → constitutive role test → bounded output → explicit interpretation and defeater check.

Applied / In Practice¶

Suppose positive ratios are judged on a multiplicative scale. A transformed average with \(f=\log\) can summarize proportional change without giving the arithmetic mean's weight to large absolute ratios. If zero appears, the expression ceases to be defined; silently replacing it by a small positive constant would create a new modeling rule, not an intrinsic property of the quasi-arithmetic mean. The report therefore states the input domain, generator, treatment of missing values, and whether equal weighting is defensible.

Mapped back: field observation or problem → candidate recognition → confusable and limit checks → appropriately scoped conclusion.

Structural Tensions¶

T1: Generator freedom versus mean identity. Infinitely many formulas can encode the same aggregation through affine generator changes. Diagnostic: Solve for affine equivalence before declaring distinct means.
T2: Order preservation versus distance change. A monotone generator preserves ranking while altering the relative size of gaps. Diagnostic: Compare transformed gaps, not only input ranks.
T3: Family generality versus domain restrictions. A flexible generator does not authorize arbitrary inputs. Diagnostic: Check the generator domain and inverse range for every dataset.
T4: Equal weights versus application importance. The canonical form treats positions symmetrically even when an application does not. Diagnostic: State whether weights are part of the model or absent.
T5: Formula example versus characterization. Recognizing geometric or harmonic means does not prove the Kolmogorov–Nagumo axioms. Diagnostic: Audit the full axiom set separately from computing examples.
T6: Autonomy versus generic aggregation. Aggregation supplies many-to-one collapse; quasi-arithmetic means add transformed-coordinate averaging and affine generator equivalence. Diagnostic: Remove the generator and inverse and test whether only generic aggregation remains.

Structural–Framed Character¶

Quasi-arithmetic mean is strongly structural: interval, generator, transformed average, inverse, and affine-equivalence roles determine recognition, while application choices select the defensible scale. The five framing criteria point in a consistent direction. Evaluative weight is limited to whether the defining conditions are met, not whether the outcome is desirable. Human practice matters to the extent that experts choose conventions, instruments, or reporting thresholds, but those choices do not make every verdict arbitrary. Institutional history explains the name and standard use; it does not replace the recognition rule. The operative vocabulary travels within the home field and closely adjacent subfields, while transfer farther away requires translation to the parent prime. Thus recognition remains disciplined even where interpretation is defeasible.

Structural Core vs. Domain Accent¶

What is skeletal. Quasi-arithmetic Mean instantiates Aggregation because it deliberately collapses several scalar inputs to one internal summary, specialized by arithmetic averaging in a monotone transformed coordinate. This is the part that can be expressed without the candidate's specialist nouns.

What is domain-bound. The irreducible accent is real-interval functional equations, strictly monotone generators, internal means, replacement or decomposability axioms, and named arithmetic, geometric, harmonic, and power special cases. Remove those elements and the result is no longer Quasi-arithmetic mean; it is only the parent relation or a loose analogy.

Why this does not clear the prime bar. The name does not recur with unchanged diagnostics across three independent domains. What transfers is already represented by prime:aggregation. The candidate remains autonomous because its in-domain recognition rule, failure modes, and consequences are stable, but its vocabulary and interventions do not float free of the home substrate.

Quasi-arithmetic Mean instantiates Aggregation because it deliberately collapses several scalar inputs to one internal summary, specialized by arithmetic averaging in a monotone transformed coordinate.

The prospective workspace queue contains one strict upward edge to prime:aggregation. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Quasi-arithmetic mean Domain-specific

Parents (1) — more general patterns this builds on

Quasi-arithmetic mean is a kind of Aggregation Prime

Quasi-arithmetic Mean instantiates Aggregation because it deliberately collapses several scalar inputs to one internal summary, specialized by arithmetic averaging in a monotone transformed coordinate.

Hierarchy path (1) — routes to 1 parentless root

Quasi-arithmetic mean → Aggregation → Micro Macro Linkage

Neighborhood in Abstraction Space¶

Quasi-arithmetic mean sits in a sparse region of the domain-specific corpus (94^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Neuman–Sándor Mean — 0.78
Monotonic Function — 0.78
Carleman's equation — 0.77
Condition Number — 0.77
Bayesian Interpretation of Kernel Regularization — 0.76

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Power mean. A one-parameter subfamily generated by powers or logarithm rather than the entire generator class.
Weighted quasi-arithmetic mean. Adds declared nonuniform weights to the transformed average.
Bajraktarević mean. Uses a ratio of weighted function sums and a broader two-function representation.
Median. An order statistic that is not generally obtained by transformed arithmetic averaging.
Arithmetic mean. The identity-generator member rather than a synonym for the family.
Expected utility. May use an affine-unique utility transform but also requires probabilities and decision semantics.

References¶

[1] Nagumo, M. (1930). ‘Über eine Klasse der Mittelwerte.’ Japanese Journal of Mathematics 7, 71–79. https://doi.org/10.4099/jjm1924.7.0_71 registry ↩

[2] Aczél, J., and Dhombres, J. (1989). Functional Equations in Several Variables, chapters 17–19. Cambridge University Press. https://doi.org/10.1017/CBO9781139086578 registry ↩

[3] Bullen, P. S. (2003). Handbook of Means and Their Inequalities. Kluwer Academic Publishers. https://doi.org/10.1007/978-94-017-0399-4 registry ↩