Let be Lebesgue measurable functions on a Lebesgue measurable set , converging almost everywhere to , with . The limit has a measurable representative: use off the exceptional null set and zero on it. The completeness of Lebesgue measure makes any alternative values on that null set measurable as well. Fatou's lemma applied to gives the displayed bound, so is finite almost everywhere and Lebesgue integrable. A uniform L1 norm bound alone does not imply convergence in L1 space; the strict inequality in Fatou lemma supplies a shrinking-spike example.
Approximate identity Created 2026-09-24 Updated 2026-11-05
A standard approximate identity on is a family withThe absolute tail condition matters for signed or complex kernels: cancellation of the signed mass alone is insufficient. For and , these conditions imply .
Indeed, express the difference as . Translation continuity in Lp makes the integrand norm small near zero, the uniform L1 norm controls that part, and the vanishing absolute tail controls the remainder. The same proof gives uniform convergence for bounded uniformly continuous functions. It does not give norm convergence for every essentially bounded function.
For nonnegative normalized mollifiers with smooth compactly supported , convergence also holds at every Lebesgue point of a locally integrable function. Such pointwise conclusions use additional size control on the kernels; they should not be attributed to the three norm-approximation conditions alone. The Cauchy approximate identity and summability kernel on the circle are related constructions on their respective domains.
Bounded errors with prescribed correlation 2026-11-05
A covariance matrix describes second moments, but does not specify a complete probability distribution. A bounded model can realize a prescribed covariance matrix exactly while keeping every possible error finite. This is useful for checking a nonlinear measurement model without accidentally generating physically impossible inputs; it is a constructed model, not evidence that an actual instrument has that distribution.
For two errors with desired standard deviations and correlation coefficient , let be independent random variables with the Rademacher distribution: each takes and with probability . SetBoth expected values vanish. Since and , direct expansion givesThere are at most four distinct states, each obtained from an equally likely sign pair. Also and . At the second sign has no effect, leaving only two distinct states. If either standard deviation is zero, the corresponding error is zero and its correlation coefficient is undefined; the covariance construction still makes sense.
More generally, a finite-support realization of a covariance matrix exists for every real symmetric positive semidefinite matrix . By the spectral theorem, write using its positive eigenvalues and corresponding orthonormal eigenvectors, and put . Let have independent Rademacher distribution coordinates. The random vector then satisfiesThe identity matrix appears because the signs have unit variance and zero pairwise covariance. There are at most distinct states, with component bounds by the Cauchy-Schwarz inequality. When , take . Thus singular covariance matrices are allowed; a nonsingular Cholesky decomposition is not required.
This full sign model also makes robust linear constraints for bounded covariance models exact. To require for every sign state, choose each sign to minimize its contribution. The result is the necessary and sufficient conditionwhere is the L1 norm. This is a small robust optimization calculation. A physical domain must be checked in the chosen units; clipping inadmissible samples afterwards generally changes both the probability distribution and its covariance matrix.
For example, in correlated threshold and amplitude errors in RC timing, take nominal amplitude , threshold , , and . Use and . Every state has , while the minimum amplitude-minus-threshold isAll four states therefore remain within the RC threshold-crossing time domain. Enumerating them supplies an exact check for this particular bounded error model; it does not establish the instrument's actual tail probabilities.
The nonuniqueness of bounded uncertainty laws with a fixed covariance is visible even in two dimensions. LetFor independent fair signs , both and have covariance matrix because is an orthogonal matrix. Yet the first coordinate of is always , whereas that of is with probability and each of with probability . Hence is respectively zero and one half, despite identical expected values and covariance matrices. A nonlinear risk calculation needs more than those moments.
Two bounded four-state error models with identical covariance and different tail probabilities. All points are constructed, not measured data.
Details. Bounded-variation contraction under clipping Created 2026-10-05 Updated 2026-11-05
Let , , and let . The variation contractionfollows from the one-Lipschitz scalar composition rule, by smooth approximation and lower semicontinuity. To conclude global membership , one must also retain its L1 norm.
That integrability is automatic when has finite Lebesgue measure. It also holds on an arbitrary domain when , equivalently , since . Without either condition it can fail: on belongs to BV space, but clipping it to gives the nonintegrable constant one. Its derivative variation is zero, which does not repair the missing global integrability.
For total variation denoising on a bounded domain with data in , clipping cannot increase either the variation penalty or the pointwise squared error. Uniqueness of the minimizer therefore forces range preservation. The variation inequality and this integrability qualification describe different parts of the argument.
Box-constrained TV-L1 denoising 2026-10-07
This finite-image convex optimization model combines robust L1 norm fidelity with discrete isotropic total variation and pixel-range constraints. An exact second-order cone program uses two orthant fidelity slacks per pixel, two box slacks per pixel and a three-dimensional Lorentz cone for each gradient magnitude. The product canonical barrier has parameter for pixels.
Central-difference noise-bias bound 2026-10-06
For the mixed forward, central finite difference, and backward derivative regularizer on , with dividing points and , the overlap multiplicity bound for piecewise difference operators has . On the middle interval, the error kernel is for , with L1 norm . Young's convolution inequality bounds that part of the bias by . If the two outer intervals together have squared error at most , combining the three pieces gives the displayed estimate for and . Interpreting as the inverse of the Volterra integration operator requires the range of the Volterra integration operator boundary condition .
Continuous functions are meagre in L1 2026-10-25
The image of in the L1 space is a meagre set. The canonical map is a bounded linear operator from the Banach space with its supremum norm, since . It is not onto: the indicator of has no continuous function representative. Continuity would force values zero throughout and one throughout , contradicting continuity at . The operator range category dichotomy therefore proves the claim. Continuous ramps also show directly that this image is not closed, since their L1 norm distance to that indicator is for .
Debiased-Lasso asymptotic normality Created 2026-09-24 Updated 2026-11-05
In , put and choose a matrix from alone. The corrected Lasso estimatehas the exact decompositionIf , every remainder coordinate is bounded by . Thus a sufficiently accurate approximate inverse of a Gram matrix and a small coefficient L1 norm error make the remainder negligible.[ref]
Given , Gaussian noise makes the first term exactly normal with covariance matrix . For coordinate , let . If the remainder divided by tends to zero in probability, then converges to a standard normal distribution. Replacing requires a consistent estimator; non-Gaussian errors require a corresponding central limit theorem.
Dirichlet kernel Created 2026-09-24 Updated 2026-11-05
The Dirichlet kernel is the trigonometric polynomialAt multiples of , the quotient is interpreted by its removable value . Summing a finite geometric series proves the formula. In the convention , the Fourier partial sum is .
Its normalized integral is one, but it changes sign. Its growing L1 norm explains why sharper truncation need not improve convergence for every continuous input; the positive Fejér kernel behaves differently.
Dual norm Created 2026-10-06 Updated 2026-10-18
A dual norm measures how strongly a bounded linear functional can respond to a unit-sized input. If is a bounded linear functional on a normed vector space , write and defineIt is the operator norm of when its scalar output is measured by absolute value. On , represent using a coefficient vector . The unit ball is symmetric, so maximizing gives the same value as maximizing : negate an input whenever its output is negative.[ref]
Convex Optimization by Stephen Boyd and Lieven Vandenberghe. §A.1.6, printed p. 636: dual norm and scalar operator-norm interpretation
For an original two-coordinate example, fix , so . The input norm changes the allowable unit ball and hence the dual norm:Under the L1 norm, , and attains the bound. Under the Euclidean norm, the Cauchy-Schwarz inequality gives , with equality at the displayed normalized vector. Under the maximum norm, each coordinate can reach unit absolute value, so ; attains it. Thus the dual norms of these three vector norms are respectively the maximum norm, the Euclidean norm, and the L1 norm. For , Lp norms on a finite coordinate space pair with their Holder conjugate exponents.[ref]
Convex Optimization by Stephen Boyd and Lieven Vandenberghe. §A.1.6, printed p. 637: Euclidean self-duality and one-, infinity- and p-norm duality; the numerical vector and maximizing points here are independent examples
This geometry explains the usefulness of a dual norm: it translates a bound on inputs into the sharp bound on a linear output. The support function of the unit ball gives its value in the coefficient direction . Changing the input budget changes the set of possible inputs; the three different answers do not conflict. In finite dimensions a closed unit ball is compact, so a continuous linear functional achieves its maximum absolute value. In an infinite-dimensional normed vector space, the supremum remains the definition and need not be achieved.
Dual norm as a worst-case linear error bound 2026-10-18
Suppose a perturbation satisfies , with . Absolute homogeneity of a norm and the dual norm definition giveFor nonzero , divide it by its norm to obtain an input in the unit ball; for zero perturbation the inequality is immediate. This is an exact error bound for a linear functional, for every perturbation in the stated input set.
For and a budget , the sharp output-error bounds are for , for , and for . Scale the maximizing vectors from the preceding example by to attain the positive endpoints; their negatives attain the negative endpoints. The L1 norm budget constrains a total absolute error; the maximum norm budget permits each coordinate to have that error separately. These budgets define nested sets and carry different information.
Dual norms give exact worst-case errors for the same linear functional
. Original bounds for . Lines show the sharp error magnitude as the stated input norm budget increases. The right panel gives the full possible error intervals at ; these are the full feasible intervals for the stated deterministic bounds. Details.The coordinates must use the stated scale and units. Rescaling one coordinate changes both its coefficient in and the meaning of its input budget. For a nonlinear differentiable function with scalar values, its Fréchet derivative at the base point is a linear functional governing the first-order response. An exact finite-perturbation error bound additionally needs control of the Taylor remainder or of the Fréchet derivative along the path.
For an integrable function on , average on each dyadic interval of length . The resulting step function converges to almost everywhere and in the L1 norm. If has the uniform distribution on , then , where rounds down to a dyadic endpoint. Refinement makes these conditional expectations a uniformly integrable martingale. The uniformly integrable martingale convergence theorem gives almost-sure and convergence in L1 to a limit .
Since , the generated limiting sigma-algebra contains up to null sets. For every event in any finite-stage sigma-algebra, convergence in the L1 norm gives . The pi-lambda theorem extends this equality to the limiting sigma-algebra, identifying . The uniform distribution of translates the two convergence conclusions into the assertions for . Choosing a Borel representative of and assigning arbitrary values at the endpoint one makes all random variables well defined without changing either conclusion.
Entrywise matrix L1 norm 2026-10-06
Fused Lasso 2026-10-05
The fused Lasso adds an L1 norm penalty on adjacent coefficient differences to a squared-error objective, optionally together with an L1 norm penalty on the coefficients themselves. The differences penalty encourages neighbouring coefficients to coincide.
Hamming cube as an L1 and L2 metric Created 2026-09-28 Updated 2026-10-14
Use the convention for the Hamming cube . Its elements are length- binary strings, and their Hamming distance counts the differing coordinates:Send each binary string to the vector with the same coordinates, by . Every differing coordinate contributes both to the sum of absolute differences and to the sum of squared differences. Therefore the L1 norm and L2 norm give, respectively,This short coordinate calculation proves that the L1 norm gives an isometric embedding, while Euclidean distance grows like the square root of the number of changed binary digits. For example, and differ twice: their Hamming distance and Manhattan distance are , but their Euclidean distance is . Comparing binary strings in these two geometric representations answers different questions even though the points have the same coordinates.
For distinct pairs ranges over every integer from to . The metric embedding into the L2 norm thus has sharp ratiosNeighbors attain and opposite corners attain . This computes the bi-Lipschitz distortion of this coordinate map; it does not prove that every metric embedding into an L2 space has at least this bi-Lipschitz distortion. Uniformly multiplying the output coordinates changes and together and leaves unchanged.
The important family distinction is that these coordinate maps have no dimension-independent bound on their bi-Lipschitz distortion into the L2 norm, yet they do give a uniform coarse embedding of a family of metric spaces. Pad each output vector with zero coordinates to view all maps in the single l2 sequence space, an L2 space for counting measure. Then the same functionswork for every , and . Similarly, padding into the l1 sequence space gives common controls and exact preservation of Hamming distance. These claims use unnormalized Hamming distance; replacing it by normalized Hamming distance changes the family of source metrics.
The geometry has a practical origin: Hamming's 1950 paper uses the number of differing positions between binary strings to study error-correcting codes. Changing one binary digit moves to a neighboring corner, so a question about corrupted messages becomes a question about separation in the Hamming cube. [ref]
https://ineffectivetheory.com/edu/papers/hamming-codes-1950.pdf. R. W. Hamming, Error Detecting and Error Correcting Codes, 1950, Section 5
Suppose . By summable L1 norms imply almost-everywhere absolute convergence, let where the series converges absolutely, and choose zero on the exceptional null set. Then is a Lebesgue measurable function and almost everywhere. Since is Lebesgue integrable, so is . The signed partial sums are bounded in absolute value by , so the dominated convergence theorem proves termwise integration. It also gives the quantitative L1 norm estimateThe numerical series of integrals is absolutely convergent because .
For example, on take . Its L1 norm is , and these norms sum to two. Every positive lies in only finitely many supports, so the series is pointwise finite, although its terms are not uniformly small. Termwise integration gives , by differentiating the geometric series. This illustrates why a small L1 norm can permit tall terms on short intervals.
For nonnegative integrable , Tonelli's theorem applied on the triangle gives the exact identityFor signed or complex integrable , first apply the same calculation to and use the triangle inequality for an integral. The defining integral then exists absolutely for almost every , and the same L1 norm bound holds. The constant is sharp: use , whose L1 norm is one and whose weighted integrals above tend to .
Finiteness at every point is stronger than this L1 space assertion. For , take , , and on , zero otherwise. It is integrable because , but because the power near is . This is compatible with an integrable output after changing values on a null set.
Let be a Borel measurable function in L1 space on , and let be Lebesgue integrable. The function is measurable. Tonelli's theorem for and translation invariance of Lebesgue measure giveFiniteness justifies Fubini's theorem for the signed product, yielding . In particular an integrable odd function makes this signed total zero while the joint L1 norm can remain positive. The triangle inequality for an integral also gives the usual convolution estimate .
L1 space Created 2026-10-14 Updated 2026-10-25
The Lp space for consists of equivalence classes of integrable functions, with L1 norm . Conditional expectation is a contraction in this norm. Convergence in this space is L1 convergence, which controls the change in every integral over a measurable subset.
Laplace prior for sparse images Created 2026-10-05 Updated 2026-10-18
Independent centered Laplace distributions for pixel intensities give the prior distributionThis prior favors small intensities through its sharp peak at zero, while its tails permit isolated appreciable pixels. Its negative log density is an L1 norm penalty, so with a Gaussian likelihood a maximum a posteriori estimate minimizes a sum of squared residuals plus that penalty. A draw from this continuous prior has no exactly zero coordinate with positive probability; sparsity of a maximizing estimate and concentration of prior mass near zero are different statements. The product prior also does not encode spatial contiguity. Inverse transform sampling produces each coordinate from an independent uniform distribution by for and otherwise, ignoring the probability-zero endpoints.



