--- title: Random Wavelet Series url: https://www.emergentmind.com/topics/random-wavelet-series type: topic --- # Random Wavelet Series Random wavelet series are random functions, fields, or distributions represented through wavelet expansions whose coefficients are random, randomized, or induced by an underlying stochastic process. In the narrow sense used in multifractal analysis, a Random Wavelet Series is a one-periodic random function with scale-dependent random wavelet coefficients [2510.00622]. In a broader probabilistic sense, the term also covers deterministic wavelet series randomized by i.i.d. multipliers, stable and multistable wavelet expansions, and dependent wavelet coefficient arrays arising from long-memory processes [2304.00811], [1901.07153], [1006.2282]. Across these formulations, wavelet localization makes coefficient moduli unusually informative for regularity, so boundedness, continuity, Besov membership, Hölder behavior, and multifractal spectra can often be read directly from scale-by-scale coefficient statistics [2304.00811], [2411.18155]. ## 1. Foundational constructions and coefficient models A standard deterministic wavelet expansion on \(\mathbb T^d\) or \(\mathbb R^d\) is written in the non-\(L^2\)-normalized form \[ f=\sum_j f_j,\qquad f_j=\sum_{i,k} c^i_{j,k}\,\psi^{(i)}_{j,k}, \] with the scale envelope \[ \omega_j=\sup_{i,k}|c^i_{j,k}|. \] This parameter \(\omega_j\) is central in the regularity analysis of randomized wavelet expansions [2304.00811]. In the one-periodic independent-coefficient model, one instead writes \[ f=\sum_{j\in\mathbb N_0}\sum_{k=0}^{2^j-1} c_{j,k}\psi_{j,k}, \] and studies the scale law of \[ X_{j,k}=\frac{-\log_2|c_{j,k}|}{j}, \] whose distribution prescribes the abundance of coefficients of a given size at each scale [2510.00622]. On \(\mathbb R^d\), a widely used Besov-prior model takes coefficients of the form \[ a_{j,t,m} = 2^{j\alpha}(j+1)^\theta \Big(1+\frac{\|m\|_\infty}{2^j}\Big)^\gamma \,\xi_{j,t,m} \quad\text{for }j\ge 1, \] with a separate coarse-scale exponent at \(j=0\), and with i.i.d. template variables \(\xi_{j,t,m}\sim X\) [2411.18155]. Sparse variants replace \(\xi_{j,t,m}\) by \(\lambda_{j,t,m}\xi_{j,t,m}\), where \(\lambda_{j,t,m}\) are Bernoulli variables with scale-location dependent success probabilities [2411.18155]. Three recurrent paradigms organize the subject. | Paradigm | Coefficient mechanism | Representative result | |---|---|---| | Randomization of a fixed series | \(c^i_{j,k}\mapsto c^i_{j,k}\chi^i_{j,k}\) | Unbounded multipliers can destroy continuity and local boundedness [2304.00811] | | Independent random coefficients | \(c_{j,k}\) i.i.d. across positions at fixed scale | Exact \(p\)-spectrum from wavelet coefficient statistics [2510.00622] | | Process-induced coefficient field | \(W_{j,k}\) obtained by wavelet transform of a stochastic process | Limits in Wiener chaos and generalized self-similar processes [1006.2282] | This variety matters because the phrase “random wavelet series” is not tied to a single probabilistic architecture. Some theories emphasize synthesis from prescribed random coefficients, whereas others analyze wavelet coefficients generated by an ambient random process. A plausible implication is that the most robust structural distinction is not constructive versus analytic, but whether regularity is controlled directly by coefficient moduli or by more global interactions between coefficients. ## 2. Regularity, continuity, and Besov behavior For deterministic wavelet series, a sharp continuity threshold is expressed through \(\omega_j\). If \[ \sum_{j\ge 0}\omega_j<\infty, \] then the wavelet series converges normally, \[ |f_j(x)|\le C\,\omega_j \quad \forall x, \] hence \(\sum_j f_j\) converges uniformly to a bounded function, and if the wavelets are continuous then \(f\) is continuous [2304.00811]. Conversely, if \((\omega_j)\) is nonnegative and \(\sum_j\omega_j=\infty\), there exists a wavelet series with \(\sup_{i,k}|c^i_{j,k}|=\omega_j\) for all \(j\) that is nowhere locally bounded [2304.00811]. In this deterministic framework, \(\ell^1\)-summability of \((\omega_j)\) is therefore the exact threshold for continuity. Randomization changes this picture in a way that is opposite to the familiar Fourier phenomenon. If \[ X=\sum_{i,j,k} c^i_{j,k}\,\chi^i_{j,k}\,\psi^{(i)}_{j,k}, \] with i.i.d. multipliers \(\chi^i_{j,k}\) having unbounded support in the literal sense \[ \mathbb P(|\chi|\ge n^3)>0\qquad \forall n, \] then unbounded randomization can destroy boundedness and continuity “very violently” [2304.00811]. The paper proves that for almost every \(f\in C(\mathbb T^d)\) in the prevalent sense, the randomized wavelet series associated with \(f\) by unbounded i.i.d. multipliers is almost surely nowhere locally bounded [2304.00811]. The decisive distinction is bounded versus unbounded multipliers, not Gaussian versus Rademacher. Bounded multipliers preserve the coefficient-modulus criteria behind \(L^p\), Hölder, Sobolev, and Besov regularity, whereas unbounded multipliers create arbitrarily large local coefficients that wavelet localization turns into genuine local singularities [2304.00811]. Gaussian randomization yields a finer threshold. Since there are \(\sim 2^{dj}\) coefficients at scale \(j\), one has almost surely \[ \sup_{i,k}|\chi^i_{j,k}|\le \sqrt{2dj} \quad\text{for all sufficiently large }j, \] so \[ \sum_{j\ge 0}\sqrt{j}\,\omega_j<\infty \] is sufficient for continuity of the Gaussian-randomized series [2304.00811]. For \(f\in C^\alpha(\mathbb T^d)\), if the multipliers have exponential tail of order \(\gamma\), then almost surely \[ |X_f(x)-X_f(y)| \le C |x-y|^\alpha |\log |x-y||^{1/\gamma}, \] so the Hölder exponent is preserved but the modulus acquires a logarithmic correction [2304.00811]. In the language of the uniform Hölder exponent, \[ H_{\min}(X_f)=H_{\min}(f)\qquad \text{a.s.} \] This rules out any general smoothing effect of unbounded randomization in wavelet coordinates [2304.00811]. Besov regularity admits a complementary coefficient-envelope characterization. For the non-sparse Besov prior on \(\mathbb R^d\), almost sure membership in \(B^s_{p,q}(\mathbb R^d)\) is governed by Property A: \[ \gamma<-\frac dp,\qquad \beta<-\frac dp,\qquad s+\frac d2+\alpha<0, \] or \[ \gamma<-\frac dp,\qquad \beta<-\frac dp,\qquad s+\frac d2+\alpha=0,\qquad \theta<-\frac1q, \] with endpoint modifications for \(p=\infty\) and \(q=\infty\) [2411.18155]. Under mild moment assumptions on the template variable \(X\), this criterion is sufficient; under the nondegeneracy condition \(P(X\neq 0)>0\), it is also necessary at the coefficient level, and essentially necessary at the function level once the wavelet characterization assumptions are imposed [2411.18155]. The same deterministic threshold also controls finiteness of moments and, under stronger assumptions, finiteness of exponential moments of the Besov norm [2411.18155]. This suggests that in many random wavelet models the probabilistic input primarily affects integrability requirements, while the location of the regularity threshold remains deterministic. ## 3. Multifractal analysis and \(p\)-spectrum theory Classical multifractal analysis of wavelet series is often formulated through Hölder exponents and iso-Hölder sets. For random wavelet series built from Gibbs measures, with coefficients \[ |d_w|=2^{-|w|(s_0-1/p_0)}\mu([w])^{1/p_0} \] and independent random perturbations of these coefficients, the ordinary singularity spectrum is inherited from the Gibbs measure through an affine change of variables [1001.4337]. The main geometric refinement concerns the graph and range restricted to iso-Hölder sets. If \[ d(h)=\xi_{F_\mu}^*(h), \] then almost surely, for \(h\in(0,1)\), \[ d^G(h)=\frac{d(h)}{h}\wedge(d(h)+1-h),\qquad d^R(h)=\frac{d(h)}{h}\wedge 1 \] for the graph and range singularity spectra [1001.4337]. The proof uses Gibbs measures on subshifts that avoid the zero set of the mother wavelet, a technical device needed to make the potential-theoretic lower bounds work [1001.4337]. A major extension replaces Hölder exponents by \(p\)-exponents, which remain meaningful for functions that are only locally in \(L^p\). For \(p\in[1,\infty)\), the \(p\)-exponent is \[ h_f^{(p)}(x_0)=\sup\left\{\alpha\ge -\frac1p: f\in T_\alpha^p(x_0)\right\}, \] and the corresponding \(p\)-spectrum is \[ \mathscr D_f^{(p)}(h)=\dim_{\mathcal H}\{x_0:h_f^{(p)}(x_0)=h\} \] [2510.00622]. The key control is through the distribution of wavelet coefficients across scales, encoded by the wavelet density \(\rho_{\vec c}\) or profile \(\nu_{\vec c}\). If \(p_0(f)>0\), then for every \(02\) [1201.4831]. This is a sharp failure of any naive reduction principle based only on Hermite rank. A related but analysis-oriented perspective studies statistics of random wavelet coefficients and their moduli rather than synthesis formulas. For processes with stationary increments, the wavelet transform \(W x(t,j)=x\star\psi_j(t)\) yields coefficient fields whose ordinary cross-scale covariance is nearly diagonal, whereas nonlinear statistics such as \(|W X|\) and \(W|W X|\) reveal non-Gaussian dependencies across scales [2204.10177]. The scattering cross-spectrum \[ C_S (j,a,b) = \frac{\E\{|X\star\psi_{j}|\star\psi_{j-b}(t) \, |X\star\psi_{j-a}|\star\psi_{j-b}(t)^*\}} {\sigma_W (j)\, \sigma_W(j-a)} \] is scale-invariant for self-similar processes in the paper’s wide-sense formulation [2204.10177]. This suggests that, for dependent coefficient fields, cross-scale envelope statistics may be as fundamental as the coefficient law itself. ## 5. Stable, multistable, anisotropic, and multiplicative series Random wavelet series with heavy-tailed coefficients require different convergence mechanisms. For i.i.d. symmetric \(p\)-stable variables \(\eta_{j,k,e}\), the fractional wavelet series \[ X_\gamma=\sum_{j,k,e}\eta_{j,k,e}\,\mathcal I_\gamma \psi_{j,k}^e \] converges almost surely in \(\mathcal D'(\mathbb R^d)\) under explicit conditions linking \(p\), \(\gamma\), the dimension \(d\), and the wavelet regularity [1901.07153]. A modified series, \[ Y_\gamma(x)=\sum_{j,k,e}\eta_{j,k,e}\,\mathcal K_\gamma \psi_{j,k}^e(x), \] is pointwise defined and has a measurable version in a higher-regularity regime [1901.07153]. In the Gaussian case \(p=2\), the field is self-similar with exponent \(d/2+\gamma\), but for \(p<2\) exact self-similarity is broken because the law depends on an \(\ell^p\)-norm of wavelet coefficients rather than a basis-invariant \(L^p\)-norm [1901.07153]. For harmonizable fractional stable sheets, the anisotropic expansion is indexed by multi-scales \(J=(j_1,\dots,j_N)\) and locations \(K=(k_1,\dots,k_N)\), \[ Z^H(t)= \Re \sum_{(J,K)} 2^{-\langle J,H\rangle}\, \epsilon_{J,K} \prod_{l=1}^N \Big\{\psi^{H_l}(2^{j_l}t_l-k_l)-\psi^{H_l}(-k_l)\Big\}, \] with stable coefficients \(\epsilon_{J,K}\) defined by integration against a rotationally invariant stable random measure [1903.04397]. The series converges almost surely in every Hölder space \(C^\gamma([-M,M]^N)\) with \(\gamma<\min\{H_1,\dots,H_N\}\), and this regularity feeds into a uniform Hausdorff-dimension theorem for inverse images of the associated \(\mathbb R^d\)-valued sheet [1903.04397]. The construction relies on a Fourier-side wavelet expansion and a LePage representation to control the stable coefficients [1903.04397]. Multistable models lead to yet another type of wavelet random series. For the random field \[ X(u,v)=\int_{\mathbb R} K_{u,v}(s)\,M_\alpha(ds), \] the Haar-wavelet expansion of the kernel yields \[ X(u,v)=\|K_{u,v}\|_1\,\eta+\sum_{j=0}^{+\infty}\sum_{k=0}^{2^j-1} w_{j,k}(u,v)\,\varepsilon_{j,k}, \] where \(\varepsilon_{j,k}\) are Haar-wavelet integrals of the multistable random measure [2004.05874]. The truncated series converges almost surely and uniformly on \([0,1]\times[a,b]\subset [0,1]\times(1/\underline\alpha,1)\), with rate \[ O\!\left(J^\zeta\,2^{-J\min\{\rho_\alpha,\;a-\frac1{\underline\alpha}\}}\right) \quad\text{a.s.} \] [2004.05874]. This strong convergence is used to define and simulate the multifractional multistable Riemann–Liouville process \(Y(t)=\widetilde X(t,H(t))\) [2004.05874]. A multiplicative variant arises for positive processes \(Y(t)=\exp\{X(t)\}\), where \(X\) itself has an additive random wavelet expansion \[ X(t)=\sum_{k\in\mathbb Z}\xi_{0k}a_{0k}(t)+\sum_{j=0}^\infty\sum_{l\in\mathbb Z}\eta_{jl}b_{jl}(t). \] Exponentiation yields the multiplicative wavelet representation \[ Y(t)=\prod_{k\in\mathbb Z}\exp\{\xi_{0k}a_{0k}(t)\} \prod_{j=0}^\infty\prod_{l\in\mathbb Z}\exp\{\eta_{jl}b_{jl}(t)\}, \] with explicit truncation conditions guaranteeing simulation with given accuracy and reliability in both \(C([0,T])\) and \(L_p([0,T])\) when \(X\) is strictly sub-Gaussian [1408.4253]. This suggests that the additive wavelet series paradigm can be transported to nonlinear positive models without abandoning explicit probabilistic error control. ## 6. Simulation, inference, and the boundaries of the concept Several works exploit random wavelet series as constructive simulation devices. For stationary strictly sub-Gaussian \(X\) with spectral density \(f\), one has a random wavelet expansion \[ X(t)=\sum_{k \in \mathbb{Z}}\xi_{0k}a_{0k}(t)+ \sum_{j=0}^{\infty}\sum_{k \in \mathbb{Z}}\eta_{jk}b_{jk}(t), \] with spectral formulas for \(a_{0k}(t)\) and \(b_{jk}(t)\) [1904.13384]. Truncating this series yields a finite simulator \(\hat X\), which in turn gives plug-in simulators for nonlinear processes \[ \hat Y(t)=(\hat X(t))^s,\qquad \hat Z(t)=\hat X_1(t)\hat X_2(t), \] together with explicit lower bounds on truncation depths ensuring prescribed accuracy \(\varepsilon\) and reliability \(\delta\) in \(L_p([0,T])\) [1904.13384]. This is a direct example of wavelet-based stochastic synthesis guided by rigorous error analysis. Wavelet-domain stochastic modeling can, however, extend beyond the formal class of random wavelet series. In the superstatistical interpolation model, each Gaussian component \(u_\xi(t)\) has a multiwavelet expansion with correlated Gaussian coefficients and sparse covariance in the transformed basis, but the full non-Gaussian process is assembled pointwise as \(u(t)=u(t,\xi(t))\), using a slowly varying log-normal latent process \(\xi(t)\) [2208.01486]. The authors explicitly emphasize that this is not a standard random wavelet series with independent random coefficients; rather, it is a hierarchical random field over \((t,\xi)\) whose Gaussian components are synthesized in wavelet space [2208.01486]. The same distinction appears on the deterministic side. A threshold autoregressive model with time-varying threshold represented by \[ \gamma_{\boldsymbol{\theta}}\!\left(\frac{t}{T}\right) = c_{00}\varphi_{00}\!\left(\frac{t}{T}\right) + \sum_{j=0}^{J-1}\sum_{k=0}^{2^j-1} d_{jk}\psi_{jk}\!\left(\frac{t}{T}\right) \] uses a wavelet series inside a stochastic time-series model, but the coefficients \(c_{00},d_{jk}\) are deterministic unknown parameters estimated by conditional least squares rather than random variables [2605.17864]. The paper explicitly states that it does not study random wavelet series in the formal probabilistic sense [2605.17864]. This is a useful boundary case: wavelet series can be central to stochastic modeling without the model itself being a random wavelet series. Taken together, these developments show that “random wavelet series” names both a concrete class of probabilistic expansions and a broader methodological zone in which wavelet coefficients, wavelet moduli, and wavelet-domain envelopes become the natural coordinates for regularity theory, multifractal analysis, and simulation. The strongest unifying principle is that wavelet localization converts coefficient statistics into spatially localized probabilistic geometry. In the independent-coefficient setting this yields sharp \(p\)-spectrum and Besov criteria [2510.00622], [2411.18155]; in randomized deterministic series it makes unbounded multipliers destructive rather than smoothing [2304.00811]; and in stable, long-memory, or multistable settings it produces random coefficient fields whose dependence and heavy tails remain visible at every scale [1901.07153], [1006.2282]. Source: https://www.emergentmind.com/topics/random-wavelet-series