Error Bound Moduli: Theory and Applications

• Error bound moduli are quantitative measures that link approximation errors to underlying problem data, playing a critical role in optimization, interpolation, and analytic number theory.
• They are computed using techniques from variational analysis, including subdifferentials and slope operators, which yield precise estimates on stability and convergence rates.
• Their applications range from algorithmic robustness in convex programming and polynomial interpolation to establishing explicit arithmetic bounds in number theory.

Error bound moduli provide a quantitative and structural measure of the relationship between approximation errors and underlying problem data in optimization, interpolation, coding, and analytic number theory. The characterization of these moduli is crucial for establishing stability, convergence rates, robustness of algorithms, and effective computation in a range of settings, from variational analysis and convex programming to polynomial interpolation and explicit analytic estimates. This entry synthesizes the key definitions, geometric and analytic principles, main theorems, calculation methodologies, and applied significance across prominent domains.

1. Notions and Definitions of Error Bound Moduli

The error-bound modulus quantifies, typically via a scalar τ or γ, the best constant in inequalities linking a functional error (residual, violation, or deviation) to a natural notion of distance from ideality (distance to feasible set, uniqueness, or algebraic norm difference). For a lower semicontinuous function $f:X \rightarrow \overline{\mathbb{R}}$ on a metric or normed space $X$, the error-bound modulus at $\bar x$ respects

$$\tau \cdot d(x, [f \le 0]) \le [f(x)]_+ \quad \text{locally near } \bar x,$$

with the exact modulus given by

$$\tau(f; \bar x) = \liminf_{x \to \bar x, f(x)>0} \frac{f(x)}{d(x,[f \le 0])}.$$

Similar quantities arise in more specialized contexts, such as the modulus of divided difference in interpolation, the plus-function modulus in LCP, strong unicity constants in approximation, and explicit algebraic norms in arithmetic geometry. Each context tailors the error quantity and geometric or analytic notion of distance to the relevant structure (Cuong et al., 2020, Li et al., 2016, Cai, 2015, Kumar et al., 2022, Sipos, 2019).

2. Structural Characterizations and Theoretical Frameworks

Comprehensive frameworks classify error-bound moduli via primal (distance, slope), dual (subdifferential), and combined conditions. In metric, Banach, or Asplund spaces, equivalences relate the modulus to slope operators:

• Nonlocal slope: $$|\nabla f|^{\diamond}(x) = \sup_{u \ne x} \frac{[f(x) - f_+(u)]_+}{d(u, x)}$$
• Strong (local) slope: $$|\nabla f|(x) = \limsup_{u \to x, u \ne x} \frac{[f(x) - f(u)]_+}{d(u, x)}$$
• Clarke/Fréchet subdifferentials: error moduli lower-bounded by minimal norm of subgradients, with exactness in convex/Asplund settings (Cuong et al., 2020).

For locally Lipschitz regular $f$, (Li et al., 2016) expresses the modulus using the support function and outer limiting subdifferential:

$$\mathrm{ebm}(f;\bar x) \le \mathrm{dist}\, \big(0,\, \partial^> \sigma_{\partial f(\bar x)}(0)\big)$$

with equality in convex and lower-$\mathcal{C}^1$ circumstances.

In polynomial interpolation on Jordan arcs, divided-difference moduli admit explicit upper bounds:

$$\big|d_n\big(f|z_1,\ldots,z_{n+1}\big)\big| \le \frac{C_\gamma (1+\mathrm{diam}(\gamma))^{n-1}}{n!},\quad C_\gamma \text { independent of parametrization}$$

leading directly to geometric error estimates for complex interpolation (Cai, 2015).

In LCPs, an explicit modulus $\beta(A)$,

$$\beta(A) = \min_{\|x\|_\infty = 1} \max_i x_i (A x)_i,$$

completely governs absolute and relative error bounds in plus-function residual reformulation (Kumar et al., 2022).

3. Explicit Computation and Practical Estimation

Error-bound moduli often admit representations via infimum/liminf formulas over slopes or (sub-)gradients, enabling direct or indirectly computable estimation:

• In Banach/Asplund spaces,

$$\tau(f;\bar x) = \inf \{\|s\|: x \to \bar x,\, f(x)>0,\, s\in\partial f(x)\}$$

• For locally Lipschitz regular functions, $\mathrm{ebm}$ estimation is operationally tied to the distance from 0 to the end set of the Clarke subdifferential (Li et al., 2016).
• For AVEs $A x - B |x| = b$, the modulus $\gamma = \max_D \| (A - B D)^{-1} \|$ is fundamental, with tractable sufficient upper bounds derived from spectral or norm inequalities (e.g., $\rho(|A^{-1}B|)<1$ or $\sigma_{\min}(A) > \sigma_{\max}(B)$) (Wu et al., 2022).
• In polynomial remainder codes, the maximal allowable error-degree $\tau_{\max}$ derives from the structure of the moduli and their gcds, code distances, and systematic combinatorial recipes ensure computable determination of error resilience (Xiao et al., 2014).
• For Chebyshev approximation with bounded coefficients, explicit constants in moduli of uniqueness and strong unicity are constructed using proof-mined alternation and Schur polynomial interpolation machinery (Sipos, 2019).

4. Geometric and Parametric Invariance

Error-bound moduli inherit or promote invariance principles reflecting the underlying geometry and do not depend on extrinsic representations or parametrizations:

• The modulus in interpolation on Jordan arcs is invariant under reparametrization of the curve; only geometric quantities (chord lengths, diameter) and intrinsic function derivatives appear in the bound (Cai, 2015).
• For convex or lower-$\mathcal{C}^1$ functions, the modulus is invariant under local coordinate changes and can be extracted from the geometry of the convex subdifferential set (Li et al., 2016).
• In algebraic settings, lower bounds for the norm of differences of singular moduli depend only on discriminant size and the arithmetic properties of the points, not on specific parametrizations (Cai, 2020).

5. Applications and Impact Across Domains

The error-bound modulus serves as a sharp quantitative constant in error analysis, convergence, and stability theorems:

• In optimization and variational inequalities, it controls feasibility-violation decay, sensitivity analysis, and convergence rates (Cuong et al., 2020, Li et al., 2016).
• In polynomial interpolation, explicit modulus bounds extend classical real-variable error estimates to complex arcs and general contours, impacting numerical analysis and approximation theory (Cai, 2015).
• In absolute value equations and linear complementarity problems, moduli translate residuals to solution errors directly, with two-sided error bounds central to algorithmic stopping criteria and perturbation analysis (Wu et al., 2022, Kumar et al., 2022).
• In coding theory, modulus characterizations distinguish the maximal error degree tolerable for robust recovery of polynomials under arbitrary and structured residue errors (Xiao et al., 2014).
• In best approximation and uniqueness theory, effective moduli provide explicit constants for stability and uniqueness of Chebyshev-type projections under coefficient constraints (Sipos, 2019).
• In analytic number theory, explicit error factors governed by arithmetic moduli (e.g., divisor functions, number of distinct prime factors) provide sharp uniform control on exponential and character sums, enabling effective bounds in $L$-function theory and Diophantine analysis (Jain-Sharma et al., 2020, González et al., 2023, Cai, 2020).

6. Illustrative Examples

Domain	Error-Bound Modulus Description	Representative Bound/Formulation
Variational Analysis/Convex Programs	Slope/subdifferential-based liminf/infimum	$\tau(f;\bar x)$ formulas (Cuong et al., 2020)
Polynomial Interpolation (Complex Arc)	Divided difference modulus; geometric constants	$C_\gamma,\, (1+\mathrm{diam}(\gamma))$ (Cai, 2015)
LCP/AVE	Spectral/plus-function modulus (matrix norm)	$\beta(A)$, $\gamma = \max_D \| (A-BD)^{-1}\|$ (Kumar et al., 2022, Wu et al., 2022)
Coding Theory (Polynomial Remainder)	GCD/code distance-based degree threshold	$\tau_{\max}$ via gcd/code structure (Xiao et al., 2014)
Chebyshev/Strong Unicity	Combinatorial/proof-mined constant	$\gamma = 1/C$, $C$ explicit (Sipos, 2019)
Analytic Number Theory	Arithmetic invariants of modulus (divisors, factors)	$(2^{\omega(q)} d(q))^{9/4}$ scaling (Jain-Sharma et al., 2020)
Modular/CM Points (Diophantine)	Discriminant-root-based algebraic lower bound	$\exp(c|\Delta|^{1/2})$ (Cai, 2020)

7. Significance, Unification, and Future Directions

Systematic characterization of error-bound moduli has unified disparate quantitative stability and error control results under common variational, geometric, and algebraic frameworks. The equivalence of primal and dual representations, and the emergence of computable, invariant moduli, have transformed both theoretical analysis and practical algorithm design in optimization, numerical analysis, coding theory, and arithmetic geometry. Ongoing work focuses on further sharpening these moduli in high-dimensional, nonconvex, or non-smooth settings, and on extending explicit, effective computations to more intricate algebraic or analytic contexts (Cuong et al., 2020, Li et al., 2016, Cai, 2015, Xiao et al., 2014, Wu et al., 2022, Kumar et al., 2022, Jain-Sharma et al., 2020, González et al., 2023, Cai, 2020, Sipos, 2019).