Fact-checked by Grok 2 weeks ago

Absolute continuity

In , absolute continuity is a property of real-valued defined on a closed [a, b] that strengthens the notion of by controlling the variation of the function over small disjoint subintervals in a uniform manner. Formally, a f: [a, b] \to \mathbb{R} is absolutely continuous if for every \epsilon > 0, there exists \delta > 0 such that for any finite collection of pairwise disjoint subintervals [a_i, b_i] \subset [a, b] satisfying \sum_{i=1}^n (b_i - a_i) < \delta, it follows that \sum_{i=1}^n |f(b_i) - f(a_i)| < \epsilon. This condition ensures that the function does not exhibit pathological oscillations or jumps that would prevent it from being recoverable from its derivative through integration. Absolute continuity plays a central role in real analysis because it bridges classical calculus with Lebesgue integration. Every absolutely continuous function is uniformly continuous and of bounded variation, and conversely, continuous functions of bounded variation can be decomposed into an absolutely continuous part and a singular part. Moreover, such functions are differentiable almost everywhere with respect to Lebesgue measure, and their derivative f' belongs to the Lebesgue space L^1[a, b]. A fundamental characterization states that f is absolutely continuous on [a, b] if and only if there exists an integrable function g \in L^1[a, b] such that f(x) = f(a) + \int_a^x g(t) \, dt for all x \in [a, b], with g = f' almost everywhere. This equivalence establishes the validity of the Fundamental Theorem of Calculus in the Lebesgue setting and highlights absolute continuity as the precise condition under which indefinite integrals behave like antiderivatives. Examples include all Lipschitz continuous functions, which form a proper subset, as well as non-Lipschitz cases like f(x) = \sqrt{x} on [0, 1]. In contrast, the Cantor function provides a continuous but singular (non-absolutely continuous) example, illustrating the distinction. Beyond one dimension, the concept extends to measures, where a measure \mu is absolutely continuous with respect to Lebesgue measure if \mu(E) = 0 whenever the Lebesgue measure of E is zero, leading to the Radon-Nikodym theorem for representing such measures via densities. This framework is essential in probability theory, functional analysis, and partial differential equations for studying regularity and integrability properties.

Absolute continuity of functions

Definition

A function f: [a, b] \to \mathbb{R} is absolutely continuous on the closed interval [a, b] if, for every \epsilon > 0, there exists a \delta > 0 such that for any finite collection of pairwise disjoint subintervals [a_i, b_i] \subset [a, b] (i.e., b_i \leq a_{i+1} for each i) satisfying \sum (b_i - a_i) < \delta, it holds that \sum |f(b_i) - f(a_i)| < \epsilon. This condition strengthens uniform continuity by uniformly controlling the function's variation over small disjoint intervals. It parallels the notion for measures, where one measure vanishes on null sets of another, but here it applies directly to functions via the Lebesgue measure on intervals.

Equivalent characterizations

A function f: [a, b] \to \mathbb{R} is absolutely continuous if and only if there exists a Lebesgue integrable function g \in L^1[a, b] such that f(x) = f(a) + \int_a^x g(t) \, dt for all x \in [a, b]. Moreover, f is differentiable Lebesgue-almost everywhere, f' = g almost everywhere, and f' is integrable, so f is the indefinite Lebesgue integral of its derivative. This equivalence follows from the fundamental theorem of calculus for Lebesgue integrals, which links the \varepsilon-\delta definition of absolute continuity to recovery via integration. Absolute continuity also connects to bounded variation: a function f of bounded variation on [a, b] is absolutely continuous if and only if its total variation function V_f is absolutely continuous. Since absolute continuity implies bounded variation (with total variation bounded by the \varepsilon-\delta condition applied uniformly), this provides a refinement within the class of bounded variation functions. The link between the \varepsilon-\delta condition and differentiability almost everywhere relies on the Vitali covering lemma, which allows control of oscillations over fine covers to establish the existence of the derivative and its integrability. Specifically, for an absolutely continuous f, the lemma helps show that the derivative exists almost everywhere by selecting disjoint intervals that approximate the behavior near points of differentiability. Absolute continuity implies differentiability almost everywhere, but the converse fails: there exist functions differentiable almost everywhere whose derivatives are integrable yet fail absolute continuity. The Cantor function provides such a counterexample, as it is continuous and increasing with derivative zero almost everywhere, but it maps a set of Lebesgue measure zero (the Cantor set) to a set of positive measure, violating the null-set preservation required for absolute continuity. The Banach–Zarecki theorem offers another characterization: a function f: [a, b] \to \mathbb{R} is absolutely continuous if and only if it is continuous, of bounded variation, and maps Lebesgue-null sets to Lebesgue-null sets (Lusin's condition (N)). Equivalently, f is continuous, differentiable almost everywhere with f' \in L^1[a, b], and satisfies condition (N). This theorem interconnects the measure-theoretic, variational, and analytic perspectives on absolute continuity.

Basic properties

Absolutely continuous functions possess several key properties that distinguish them within the class of continuous functions. First, every absolutely continuous function is uniformly continuous on [a, b], as the \varepsilon-\delta condition implies the standard uniform continuity criterion by considering single intervals. Second, absolutely continuous functions are of bounded variation. The total variation is controlled by the \varepsilon-\delta property, ensuring that the supremum of sums of absolute differences over partitions is finite. Moreover, any function of bounded variation can be decomposed into an absolutely continuous part and a singular part via the . Third, an absolutely continuous function f is differentiable almost everywhere with respect to Lebesgue measure, and its derivative f' belongs to L^1[a, b]. Furthermore, f(x) = f(a) + \int_a^x f'(t) \, dt for all x \in [a, b], validating the fundamental theorem of calculus in the Lebesgue sense. Additionally, absolutely continuous functions satisfy Lusin's condition (N): they map Lebesgue-null sets to Lebesgue-null sets. This property ensures that the function preserves the notion of measure zero in its range.

Examples and counterexamples

A canonical class of absolutely continuous functions consists of on a closed interval [a, b]. For such a function f satisfying |f(x) - f(y)| \leq K |x - y| for some constant K > 0 and all x, y \in [a, b], absolute continuity follows directly: given \epsilon > 0, choose \delta = \epsilon / K, so that for any finite collection of disjoint subintervals with total length less than \delta, the sum of |f(x_i) - f(y_i)| is less than \epsilon. Another fundamental class comprises indefinite integrals of integrable functions. Specifically, if g \in L^1[a, b], then F(x) = \int_a^x g(t) \, dt is absolutely continuous on [a, b], as it satisfies the - condition via the absolute continuity of the Lebesgue . A concrete example is f(x) = \sqrt{x} on [0, 1], which can be expressed as f(x) = \int_0^x \frac{1}{2\sqrt{t}} \, dt; here, \frac{1}{2\sqrt{t}} belongs to L^1[0, 1] since \int_0^1 t^{-1/2} \, dt = 2, confirming absolute continuity. This function is not Lipschitz continuous, however, because its \frac{1}{2\sqrt{x}} is unbounded near x = 0. Counterexamples illustrate the distinction from mere continuity. The (or ), defined on [0, 1], is continuous and non-decreasing, mapping the (of measure zero) onto an interval of positive length, but it is constant almost everywhere and has derivative zero almost everywhere while increasing overall from 0 to 1; thus, it fails absolute continuity. The H(x), defined as H(x) = 0 for x < 0 and H(x) = 1 for x \geq 0, is discontinuous at x = 0. Since absolute continuity implies uniform continuity (and hence continuity), H cannot be absolutely continuous on any interval containing 0. The Weierstrass function w(x) = \sum_{n=0}^\infty a^n \cos(b^n \pi x) (with $0 < a < 1, ab > 1 + \frac{3\pi}{2}) is continuous on \mathbb{R} but differentiable nowhere. Absolutely continuous functions are differentiable almost everywhere, so w is not absolutely continuous. The following table compares these properties for representative functions on [0, 1]:
FunctionUniformly ContinuousLipschitz ContinuousAbsolutely Continuous
Constant function f(x) = cYesYesYes
f(x) = xYesYesYes
f(x) = \sqrt{x}YesNoYes
YesNoNo
Heaviside function H(x) (adjusted to [0,1])NoNoNo
YesNoNo

Generalizations

Absolute continuity extends beyond scalar-valued functions on intervals. For vector-valued functions f: [a, b] \to \mathbb{R}^m, the notion is defined componentwise: f is absolutely continuous if each component f_j is absolutely continuous. Equivalently, f(x) = f(a) + \int_a^x f'(t) \, dt almost everywhere, with f' Bochner integrable. This preserves properties like differentiability almost everywhere and the fundamental theorem of calculus. In higher dimensions, absolute continuity generalizes to mappings f: \Omega \subset \mathbb{R}^n \to \mathbb{R}^m. A key extension is n-absolute continuity (or absolute continuity on n-dimensional measure), introduced by Gehring and developed by Malý and others: for every \epsilon > 0, there exists \delta > 0 such that for any finite collection of disjoint n-cubes with total n-measure less than \delta, the sum of the n-dimensional Hausdorff measures of the images is less than \epsilon. This condition ensures the mapping is approximately differentiable and is crucial for regularity theory in PDEs and . Further generalizations include \alpha-absolute continuity on rectangles in \mathbb{R}^n, where the control is over products of intervals with small \alpha-dimensional content, facilitating change-of-variable formulas in multiple integrals. These notions align with the one-dimensional case but account for the of higher-dimensional domains.

Absolute continuity of measures

Definition

In measure theory, absolute continuity describes a relationship between two measures on the same , where one measure vanishes on the null sets of the other. Let (X, \mathcal{M}) be a equipped with two positive \sigma-finite measures \mu: \mathcal{M} \to [0, \infty] and \nu: \mathcal{M} \to [0, \infty]. The measure \mu is said to be absolutely continuous with respect to \nu, denoted \mu \ll \nu, if for every measurable set E \in \mathcal{M}, \nu(E) = 0 implies \mu(E) = 0. This condition establishes that \nu dominates \mu with respect to null sets, meaning the collection of \nu-null sets contains all \mu-null sets, or equivalently, the preimage under \mu of \{0\} is a subset of the preimage under \nu of \{0\}. In applications, \nu is frequently the Lebesgue measure on \mathbb{R}^n, ensuring that \mu assigns zero measure to sets of Lebesgue measure zero, such as sets of topological dimension less than n. The \sigma-finiteness assumption on both measures guarantees that the space can be covered by countably many sets of finite measure, facilitating extensions to theorems like the Radon-Nikodym theorem without additional pathologies. This set-theoretic notion parallels absolute continuity for functions, where small intervals under correspond to small changes in function values.

Equivalent conditions

A measure \mu is absolutely continuous with respect to another measure \nu, denoted \mu \ll \nu, if \mu(E) = 0 whenever \nu(E) = 0 for every measurable set E. For finite signed measures \mu and nonnegative \nu, absolute continuity \mu \ll \nu is equivalent to the uniform absolute continuity condition: for every \varepsilon > 0, there exists \delta > 0 such that \nu(E) < \delta implies |\mu(E)| < \varepsilon for all measurable E. This \varepsilon-\delta characterization captures the intuitive notion that \mu is controlled by small sets under \nu. Under the assumption that \mu is \sigma-finite and nonnegative, \mu \ll \nu if and only if there exists a nonnegative integrable function f \in L^1(\nu) (the Radon-Nikodym derivative) such that \mu(E) = \int_E f \, d\nu for every measurable E. This representation links absolute continuity directly to integration, providing a density function that expresses \mu in terms of \nu. For finite measures, another equivalent formulation involves uniform integrability of simple function approximations: if simple functions \phi_n approximate \mu from below such that \int \phi_n \, d\nu \leq \mu(X) and \sup_n \int_E \phi_n \, d\nu \to 0 as \nu(E) \to 0, then \mu \ll \nu. This condition ensures that the approximations behave uniformly well on small \nu-measure sets. These equivalences can be established using standard techniques in measure theory. For instance, the \varepsilon-\delta condition follows from the continuity of measures, while the existence of the density relies on a Hahn decomposition of the space and constructing the derivative as the pointwise supremum over simple functions via monotone convergence; uniqueness holds \nu-almost everywhere. Hahn-Banach separation arguments can also underpin the construction in more abstract settings. Absolute continuity \mu \ll \nu is a one-directional relation, differing from mutual (or bi-) absolute continuity, which requires both \mu \ll \nu and \nu \ll \mu. The latter implies that \mu and \nu share the same null sets but does not necessarily mean they are equal.

Basic properties

One fundamental property of absolute continuity for measures is transitivity: if \lambda \ll \mu and \mu \ll \nu, then \lambda \ll \nu. If \mu \ll \nu and \lambda \ll \mu, where \mu, \nu, and \lambda are \sigma-finite positive measures on the same measurable space, then \lambda \ll \nu, and the Radon--Nikodym derivative satisfies the chain rule \frac{d\lambda}{d\nu} = \frac{d\lambda}{d\mu} \cdot \frac{d\mu}{d\nu} almost everywhere with respect to \nu. For probability measures on \mathbb{R}^d, absolute continuity is preserved under convolution: if \mu is absolutely continuous with respect to Lebesgue measure (i.e., \mu admits a density in L^1(\mathbb{R}^d)) and \rho is any probability measure, then the convolution \mu * \rho is also absolutely continuous with respect to Lebesgue measure. Similarly, if \rho is also absolutely continuous with respect to Lebesgue measure, then the product measure \mu \times \rho on \mathbb{R}^d \times \mathbb{R}^d is absolutely continuous with respect to Lebesgue measure on \mathbb{R}^{2d}. Absolute continuity also implies continuity from below and from above: if \mu \ll \nu and \{E_n\} is an increasing sequence of measurable sets with \bigcup_n E_n = E, then \mu(E) = \lim_{n \to \infty} \mu(E_n); likewise, for a decreasing sequence \{F_n\} with \bigcap_n F_n = F and \nu(F_n) < \infty for all n, \mu(F) = \lim_{n \to \infty} \mu(F_n). In the \sigma-finite case, \mu \ll \nu if and only if \mu belongs to the closure in L^1(\nu) of the simple measures (i.e., finite linear combinations of characteristic functions of measurable sets with finite \nu-measure).

Decomposition into singular and absolutely continuous parts

In measure theory, the Lebesgue decomposition theorem asserts that given a σ-finite measure μ and another σ-finite measure ν on the same measurable space (X, \mathcal{M}), there exist unique measures μ_{ac} and μ_s such that μ = μ_{ac} + μ_s, where μ_{ac} is absolutely continuous with respect to ν (μ_{ac} \ll ν) and μ_s is singular with respect to ν (μ_s \perp ν). This decomposition uniquely separates the "smooth" component of μ that aligns with ν from the "concentrated" component that avoids ν-null sets in a precise sense. Two positive measures μ and ν on (X, \mathcal{M}) are said to be singular (μ \perp ν) if there exist disjoint measurable sets E, F \in \mathcal{M} with E \cup F = X such that μ(F) = 0 and ν(E) = 0. Equivalently, μ is singular with respect to ν if μ is concentrated on a ν-null set, meaning there exists A \in \mathcal{M} with ν(A) = 0 such that μ(X \setminus A) = 0. In the context of the decomposition, μ_s satisfies this condition relative to ν, ensuring no overlap in their "supports" beyond null sets. The uniqueness of the decomposition follows from the fact that if μ = μ_{ac} + μ_s = μ_{ac}' + μ_s' with the same properties, then μ_{ac} - μ_{ac}' and μ_s - μ_s' must both be zero by mutual singularity and absolute continuity arguments. The construction typically proceeds via the : for a suitable signed measure derived from μ and ν, a Hahn decomposition yields sets separating the positive and negative parts, from which the singular and absolutely continuous components are extracted by projection onto L^1(ν) and the singular part in L^\infty-like behaviors orthogonal to it. This yields an orthogonal sum in the measure algebra, where μ_{ac} and μ_s are mutually singular, analogous to orthogonal direct sums in but respecting the lattice structure of measures. For example, the Lebesgue measure λ on \mathbb{R} decomposes with respect to itself as λ = λ + 0, where the absolutely continuous part is λ itself and the singular part is the zero measure. In contrast, the Dirac measure δ_0 at the origin is singular with respect to λ, as it concentrates entirely on {0}, a set of Lebesgue measure zero, so its decomposition is δ_0 = 0 + δ_0.

Generalizations

For signed measures, absolute continuity with respect to a positive measure \mu is defined componentwise via the Jordan decomposition \nu = \nu^+ - \nu^-, where \nu^+ and \nu^- are the positive and negative parts, respectively; specifically, \nu \ll \mu if and only if both \nu^+ \ll \mu and \nu^- \ll \mu. This extension preserves the core property that \mu(E) = 0 implies \nu(E) = 0 for all measurable E, leveraging the uniqueness of the . In the case of non-\sigma-finite measures, the standard definition of absolute continuity still applies directly—\nu \ll \mu if \mu(E) = 0 implies \nu(E) = 0—but the Radon-Nikodym theorem may fail without \sigma-finiteness of \mu, necessitating local versions restricted to \sigma-finite subclasses or sets of finite \mu-measure. Local absolute continuity requires that for every \varepsilon > 0, there exists \delta > 0 such that if \mu(F) < \delta for a set F of finite \mu-measure, then |\nu(F)| < \varepsilon, allowing decomposition on saturated classes of sets where \mu is locally finite. In potential theory, absolute continuity extends to capacities, such as Riesz capacities C_{\alpha,p}, where a measure \mu is absolutely continuous with respect to a capacity C if C(E) = 0 implies \mu(E) = 0 for relevant sets E; this notion underpins Choquet integrals and balayage for Riesz kernels I_\alpha(x) = |x|^{\alpha - d} in \mathbb{R}^d. For example, measures absolutely continuous with respect to Riesz capacities \dot{C}_{\alpha,p} admit representations via L^q-potentials when $1 < p < q < \infty, facilitating applications in nonlinear potential theory. The concept also generalizes to quasi-measures or contents, which are finitely additive set functions on algebras; here, absolute continuity of a finitely additive \alpha with respect to another \beta means that for every \varepsilon > 0, there exists \delta > 0 such that \beta(E) < \delta implies |\alpha(E)| < \varepsilon, enabling Radon-Nikodym-type theorems under additional uniformity conditions. This \varepsilon-\delta formulation aligns with order-continuity and supports extensions to non-additive functionals while avoiding \sigma-additivity. Recent connections to descriptive set theory link absolute continuity to the , where sets defined via absolute continuity relations (e.g., between Borel measures) occupy levels like \Pi^0_3-complete classes, with minor refinements post-2020 emphasizing invariance under Borel reducibility in spaces. No fundamental changes have emerged since the mid-20th century foundations, but these ties aid in classifying measure-theoretic structures within descriptive hierarchies.

Interconnections and applications

Relationship between the two notions

The notions of absolute continuity for functions and for measures share a fundamental : both capture a form of dependence in which negligible inputs produce negligible outputs, particularly by preserving sets of measure zero. For functions defined on the real line, absolute continuity ensures that the image of any has zero, reflecting a strong form of regularity beyond mere . Similarly, in the measure-theoretic setting, one measure ν is absolutely continuous with respect to another μ if every μ-null set is also a ν-null set, meaning ν assigns zero mass precisely where μ does. This shared structure manifests concretely through the indefinite integral, which bridges the two concepts. Consider a measure μ on the real line given informally by dμ = f dx, where dx denotes ; such a μ is absolutely continuous with respect to if and only if the density f belongs to L^1, the space of integrable functions. In this case, the associated F(x) = μ([0, x]) qualifies as an absolutely continuous function, illustrating how measure-theoretic absolute continuity induces functional absolute continuity via . Conversely, the absolute continuity of F as a function guarantees that the corresponding measure is absolutely continuous with respect to . Historically, the concept emerged in the context of functions before its generalization to measures, with early developments by mathematicians like Giuseppe Vitali, who in provided a key characterization of absolutely continuous functions in relation to measurable functions and . further advanced the idea around the same period, linking it to the foundations of theory. This functional origin directly informs the measure-theoretic version, as the absolute continuity of a function like the indefinite of a implies the absolute continuity of the induced measure. Despite these parallels, the two notions differ in their formulation and applicability. Absolute continuity for functions operates locally, being defined relative to finite partitions of intervals and allowing verification on compact subintervals independently. In contrast, absolute continuity for measures is inherently global, depending on the behavior across the entire underlying of sets.

Role in the Radon–Nikodym theorem

The Radon–Nikodym theorem establishes that absolute continuity is the key condition under which one measure can be represented as an integral with respect to another. Specifically, let (\Omega, \mathcal{F}, \nu) be a measure space where \nu is \sigma-finite, and let \mu be another measure on \mathcal{F} such that \mu \ll \nu. Then there exists a unique (up to \nu-almost everywhere equivalence) non-negative measurable function f \in L^1(\nu) satisfying \mu(E) = \int_E f \, d\nu for every E \in \mathcal{F}. This function f, called the Radon–Nikodym derivative and denoted \frac{d\mu}{d\nu}, provides a density for \mu with respect to \nu, generalizing the classical notion of a derivative for indefinite integrals of functions, where the fundamental theorem of calculus recovers the integrand almost everywhere. Absolute continuity \mu \ll \nu is both necessary and sufficient for the existence of such an f, as the vanishing of \mu on \nu-null sets ensures the integral representation aligns precisely with \mu. A standard proof of the theorem for \sigma-finite measures proceeds by first extending to signed measures via the Jordan \mu = \mu^+ - \mu^-, where each part admits a non-negative , yielding f = f^+ - f^- by . For the non-negative case, one constructs f as the limit of an increasing of functions f_n derived from optimizing constants in approximations like \nu(E) - \int_E c \, d\mu \geq 0 for suitable c, leveraging the \sigma-finiteness to control approximations on finite-measure subsets and pass to the limit. Alternatively, in a functional-analytic approach, the map g \mapsto \int g \, d\mu defines a continuous linear functional on L^\infty(\nu), which by the (or direct projection onto the closure of functions in L^1(\nu)) corresponds to integration against some f \in L^1(\nu). The Yosida–Hewitt further underscores the role of absolute continuity by partitioning any into absolutely continuous and singular parts with respect to \nu, with \mu \ll \nu implying the singular part vanishes. The theorem extends naturally to signed measures by applying the result to positive and negative parts, and to complex measures \mu = \mu_r + i \mu_i by representing each real and imaginary component via real-valued densities f_r, f_i \in L^1(\nu), yielding a complex f = f_r + i f_i. For vector-valued measures taking values in a X, a Radon–Nikodym derivative exists (as an X-valued Bochner integrable function) X possesses the Radon–Nikodym property, ensuring the integral representation holds under absolute continuity. In the non-\sigma-finite setting, where standard proofs fail due to potential non-localizability, Dieudonné provided a generalization in the by replacing absolute continuity with the stronger condition of "truly continuous" additive set functions—those vanishing uniformly on sets of arbitrarily small \nu-measure—allowing the density representation without \sigma-finiteness assumptions, a framework still in use for localizable measure spaces.

Applications in real analysis

One key application of absolute continuity in real analysis is the change of variables formula for integrals. If f: [a, b] \to \mathbb{R} is integrable and g: [c, d] \to [a, b] is strictly increasing and absolutely continuous, then \int_c^d f(g(x)) g'(x) \, dx = \int_a^b f(y) \, dy, where the right-hand side is taken over the image g([c, d]). This holds because absolute continuity ensures g is differentiable almost everywhere with an integrable derivative, allowing the substitution via the fundamental theorem of calculus for Lebesgue integrals. Absolute continuity also plays a crucial role in Fubini's theorem for product measures. For measures \mu_1 and \mu_2 on spaces X_1 and X_2, if \nu_1 \ll \mu_1 and \nu_2 \ll \mu_2 (absolutely continuous with to \mu_1 and \mu_2), then the product \nu_1 \times \nu_2 \ll \mu_1 \times \mu_2. This guarantees that iterated integrals equal the for integrable functions over the product space, ensuring the validity of multiple integrals in higher dimensions without singularities on null sets. In the study of weak convergence of measures, absolute continuity facilitates convergence criteria via densities. If \{\mu_n\} is a sequence of probability measures on \mathbb{R} absolutely continuous with respect to with densities f_n \in L^1(\mathbb{R}), and f_n \to f in L^1 norm where f \geq 0 and \int f = 1, then \mu_n converges weakly to the measure with density f. This follows from Scheffé's theorem, which equates L^1 convergence of densities to convergence, implying . Sobolev embedding theorems leverage absolute continuity to relate function spaces. In one dimension, functions in the W^{1,p}(a,b) for $1 \leq p < \infty coincide almost everywhere with absolutely continuous functions, and the embedding W^{1,p}(a,b) \hookrightarrow C[a,b] holds, mapping into continuous functions with a bound on the depending on the W^{1,p} norm. This provides regularity results essential for partial differential equations, where weak solutions gain higher smoothness. In modern applications, such as , absolute continuity of probability measures underpins normalizing flows for . Normalizing flows construct complex distributions by applying sequences of invertible, differentiable transformations to a simple base measure (e.g., Gaussian), preserving absolute continuity with respect to and enabling exact likelihood computation via the change-of-variables formula. This approach, detailed in foundational reviews, supports generative modeling tasks like variational inference and sampling from multimodal distributions.

References

  1. [1]
    [PDF] Real Analysis MAA 6616 Lecture 22 Absolutely Continuous Functions
    Page 2. A function f : [a, b] −→ R is absolutely continuous on [a, b] if for any > 0 there exists. δ > 0 such that for any countable (finite or infinite) ...
  2. [2]
    [PDF] Appendix: Functions of one variable - UC Davis Math
    Absolute continuity is a strength- ening of uniform continuity that provides a necessary and sufficient condition for the fundamental theorem of calculus to ...
  3. [3]
    [PDF] Absolutely continuous functions, Radon-Nikodym Derivative APPM ...
    Apr 22, 2016 · A real-valued function f on a closed, bounded interval [a, b] is. said to be absolutely continuous on [a, b] provided for each > 0, there is δ ...
  4. [4]
    [PDF] Measure Theory Princeton University MAT425 Lecture Notes
    Jan 10, 2025 · ... measures on it. Definition 5.35 (Absolute continuity). We say that ν is absolutely continuous w.r.t. µ, and write ν ◁ µ, iff for any E ∈ M ...
  5. [5]
  6. [6]
    [PDF] Absolute Continuity and the Banach–Zaretsky Theorem
    We use the standard notations of real analysis. We mention some specific ... G. (x) = g(x) for almost every x ∈ [a,b]. 3.2 Absolute Continuity. To motivate the ...
  7. [7]
    Absolutely Continuous Functions
    If Ω ⊂ R is an open set, then we define the notion of absolute continuity for a function u : Ω → R as in Definition 3.1, with the only change that we now ...
  8. [8]
    [PDF] Fall 2001 Real Variables I Dr. R. Boyer Bounded Variation, Absolute ...
    Proposition: A function that is absolutely continuous is of bounded variation. Proof: We let:9 = 1 in the condition for absolute continuity and letC= Ш. ; 0 ...
  9. [9]
    [PDF] 3.5 absolutely continuous and singular functions - Christopher Heil
    Now we turn to absolutely continuous functions on the real line. A collection of intervals in R are called nonoverlapping if their interiors are disjoint.
  10. [10]
    [PDF] Vitali's covering lemma - mimuw
    Vitali's covering lemma. A Vitali covering of a set ECRY is a family I ... रह. If(p;) - f(x;) |. Absolute continuity of the integral. F(x) = ff(t)at is ...
  11. [11]
    The Cantor function - ScienceDirect
    G is continuous and increasing but not absolutely continuous. 2.1.2. G is constant on each interval from I ∘ . 2.1.3. G is a singular function. 2.1.4. G ...<|control11|><|separator|>
  12. [12]
    AN ELEMENTARY PROOF OF THE BANACH–ZARECKI THEOREM
    This theorem asserts that if a continuous and V B function satisfies Lusin's condition (N) on an interval then it is also AC on that interval. The proofs in [5] ...
  13. [13]
    [PDF] Lectures on Measure Theory - IMPA
    Nov 21, 2022 · Show that the relation ≪ of absolute continuity between measure is an order relation (that is: prove reflexivity and transitivity). c ...
  14. [14]
  15. [15]
    [PDF] An Introduction to Measure Theory - Terry Tao
    Most of the material here is self-contained, assuming only an undergraduate knowledge in real analysis (and in particular, on the. Heine-Borel theorem, which we ...
  16. [16]
    [PDF] MIRA.pdf - Measure, Integration & Real Analysis
    ... measure, integration, and real analysis. This book aims to guide you to the wonders of this subject. You cannot read mathematics the way you read a novel ...
  17. [17]
    [PDF] Differentiation and - Integration - UCSD Math
    Theorem: A function F is an indefinite integral if and only if it is absolutely continuous. Proof: If F is an indefinite integral, then F is absolutely contin-.<|separator|>
  18. [18]
    [PDF] The Cantor function
    Proposition 1.1. 1.1.1. G is continuous and increasing but not absolutely continuous. 1.1.2. G is constant on each interval from I. ◦ . Page 4. The Cantor ...
  19. [19]
    [PDF] Chapter 3. Absolutely Continuous Functions
    In this section we show that absolutely continuous functions are precisely those func- tions for which the fundamental theorem of calculus is valid. Theorem 2.1 ...
  20. [20]
    Continuous, Nowhere Differentiable Functions - Penn Math
    Jan 23, 2024 · The Weierstrass M-test implies continuity of f whenever . The proof shows directly that the difference quotient does not have a limit as h → 0 at any point x.
  21. [21]
    [PDF] Signed Measures and the Radon-Nikodym Theorem
    The Jordan decomposition of ν on X is the representation of ν as the difference ν+ −ν− of two (non-negative) measures ν+ and ν− defined on A, where, for each S ...
  22. [22]
    [PDF] 9. Signed Measures
    Let µ be a signed measure with Jordan decomposition µ = µ. +. − µ−. Define ... Absolute Continuity. Definition 9.13. Let µ, ν be two measures. We say ν ...
  23. [23]
    Radon-Nikodym theorem for non-sigma finite measures
    Jan 6, 2017 · The Radon-Nikodym theorem provides a decomposition of λ in a sum of an absolutely continuous measure wrt μ plus a singular measure wrt μ.What is the "continuity" in "absolute continuity", in general?Fubini's theorem without completeness or σ-finiteness conditionsMore results from mathoverflow.net
  24. [24]
    [PDF] absolute continuity and singularity of two probability measures on a ...
    Apr 6, 2011 · ... σ-algebra B which is generated by an increasing sequence of σ- algebras Bn (under the condition of their local absolute continuity). A. R. ...
  25. [25]
    [PDF] CHOQUET INTEGRALS IN POTENTIAL THEORY - DMLe
    is absolutely continuous with respect to the capacity ˙Cα,p (i.e. ˙Cα,p(e)=0 implies µ(e)=0 for e Borel: µ ¿ ˙Cα,p). If 0 < p − 1 <q< ∞, then there is a ...
  26. [26]
    [PDF] arXiv:2404.06810v3 [math.AP] 12 Jun 2024
    Jun 12, 2024 · We develop a theory of capacities associated with local Mucken- houpt weights. Fundamental properties of local Muckenhoupt weights will be ...
  27. [27]
    On a Radon-Nikodym theorem for finitely additive set functions - MSP
    ( i ) a is absolutely continuous with respect to β(a< β) means: given ε > 0, there exists δ > 0 such that v(β, E) < δ (EeΣ) implies v(a, E) < ε. (ii) a is ...
  28. [28]
    ON THE ABSOLUTE CONTINUITY OF ADDITIVE AND NON ...
    Sep 11, 2013 · Abstract: We are concerned with the notions of absolute continuity in terms of 0-continuity, and of ( , δ)-continuity, with respect to a ...
  29. [29]
    (PDF) Normal Numbers and the Borel Hierarchy - ResearchGate
    Dec 30, 2024 · PDF | We show that the set of absolutely normal numbers is $\mathbf \Pi^0_3$-complete in the Borel hierarchy of subsets of real numbers.
  30. [30]
    [PDF] The complexity of the Structure and Classification of Dynamical ...
    Mar 20, 2022 · with descriptive set theory, we consider probability measures. We ... absolute continuity of measures on the unit circle. The latter ...
  31. [31]
    [PDF] Real Analysis
    Folland, Gerald B. Real analysis : modern techniques and their applications I Gerald B. Folland. - 2nd ed. p. em. - (Pure and applied mathematics).
  32. [32]
    [PDF] REAL AND COMPLEX ANALYSIS
    Rudin, Walter, 1921 -. Real and complex analysis. Bibliography: p ... uniform continuity, and uniform convergence). The first seven chapters of my ...
  33. [33]
  34. [34]
    Giuseppe Vitali (1875 - 1932) - Biography - MacTutor
    ... continuity of measurable functions (1905), the first example of a nonmeasurable set for Lebesgue measure (1905), the characterisation of absolutely continuous ...
  35. [35]
    NOTES ON ABSOLUTELY CONTINUOUS FUNCTIONS OF ...
    We give an example of an absolutely continuous function of two variables, whose derivative is not in L2,1. The boundary behavior of n-absolutely continuous ...
  36. [36]
    [PDF] Differentiation - UC Davis Math
    The Radon-Nikodym theorem provides a converse to Example 6.24 for abso- lutely continuous, σ-finite measures. As part of the proof, from [4], we also show.
  37. [37]
    [PDF] Version of 17.11.04 Chapter 23 The Radon-Nikodým Theorem
    Then there is a µ-integrable function f on X such that νE = RE f for every E ∈ Σ iff ν is finitely additive and absolutely continuous with respect to µ.
  38. [38]
    Change of variables for absolutely continuous functions
    March 1969 Change of variables for absolutely continuous functions. William P. Ziemer · DOWNLOAD PDF + SAVE TO MY LIBRARY. Duke Math.Missing: theorem | Show results with:theorem
  39. [39]
    [PDF] Lecture 5 Theorems of Fubini-Tonelli and Radon-Nikodym
    (S, S, µ)=(T, T , ν) = ([0, 1], B([0, 1]), λ). The existence of the product measure gives us an easy access to the Lebesgue measure on higher-dimensional ...
  40. [40]
    [PDF] Convergence of Probability Measures - CERMICS
    Billingsley, Patrick. Convergence of probability measures / Patrick Billingsley. - 2nd ed. Probability and statistics) p. cm. - (Wiley series in probability ...Missing: transitivity | Show results with:transitivity
  41. [41]
    Sobolev spaces
    A function u 2 Lp( ) belongs to the space W1;p( ) if and only if it has a representative u that is absolutely continuous on LN1-a.e. line segments of that are ...
  42. [42]
    [PDF] Normalizing Flows for Probabilistic Modeling and Inference
    Normalizing flows provide a general mechanism for defining expressive probability distribu- tions, only requiring the specification of a (usually simple) base ...