RVMD Reduced-Order Variational Mode Decomposition
Algorithms AIMech:RVMDReducedOrderVariationalModeDecomposition DRAFT
Aliases: Reduced-Order Variational Mode Decomposition, RVMD Spatiotemporal Modal Decomposition, Wiener-Filtered Block Coordinate Descent Reduced-Order VMD
A spatiotemporal modal decomposition and dimensionality reduction algorithm that represents a high-dimensional space-time field q(x, t) as a rank-K expansion sum_{k=1}^K phi_k(x) c_k(t) with unit-norm spatial modes ||phi_k||_x = 1 and minimizes the sum of the space-time reconstruction error and the analytic-signal spectral bandwidth penalty sum_{k=1}^K int_0^infty 2*alpha*(omega - omega_k)^2 |c_hat_k(omega)|^2 d_omega via Block Coordinate Descent alternating residual spatial projection, second-order Wiener filtering g_hat_k(omega) = 1 / [1 + 2*alpha*(omega - omega_k)^2], and power-spectrum center-of-gravity frequency updates.
Identity Notes & Concept Scope
No AIO class represents Reduced-Order Variational Mode Decomposition (RVMD); DimensionalityReduction (AIO:DimensionalityReduction) and PrincipalComponentAnalysis (AIO:PrincipalComponentAnalysis) are the closest broad matches covering low-rank orthogonal/non-orthogonal modal reduction of high-dimensional data.
Claims & Grounded Evidence
4 Claims
Mechanism
#block-coordinate-descent-wiener-filtered-spatiotemporal-alternating-updates
By applying Parseval's theorem to transform the time-domain analytic-signal Hilbert bandwidth penalty into the positive-frequency domain (Equations 9, 17-18), RVMD minimizes F({phi_k, c_hat_k^+, omega_k}) via Block Coordinate Descent over the residual R_hat_k^+ = Q_hat^+ - sum_{i != k} phi_i (c_hat_i^+)^T (Equation 36), alternately updating the normalized spatial mode phi_k^{n+1} = Re{R_hat_k^+ (c_hat_k^+)^*} / ||Re{R_hat_k^+ (c_hat_k^+)^*}|| (Equation 39), the second-order Wiener-filtered temporal spectrum c_hat_k^{+,n+1} = G_hat_k^+ (R_hat_k^+)^T phi_k with g_hat_i^+ = 1 / [1 + 2*alpha*(omega_i^+ - omega_k)^2] (Equations 38, 40), and the center-of-gravity central frequency omega_k^{n+1} = ((c_hat_k^+)^H W^+ c_hat_k^+) / ((c_hat_k^+)^H c_hat_k^+) (Equation 41).
Mechanism
#filtered-two-point-correlation-eigenvalue-problem-and-pod-dft-limits
At BCD convergence, each RVMD spatial mode satisfies the Wiener-filtered two-point spatial correlation eigenvalue equation R_hat_k G_hat_k R_hat_k^H phi_k = lambda_k phi_k (Equation 47) with exact half-power filter bandwidth Delta = sqrt((2*sqrt(2) - 2)/alpha) (Equation 48), continuously bridging Proper Orthogonal Decomposition (POD, as alpha -> 0 and G_hat_k -> I, Equations 49-51) and single-frequency Discrete Fourier Transform modes (as alpha -> infty and Delta -> 0, Section 3.2).
Application
#nonstationary-transient-wake-and-supersonic-screeching-jet-application
Coupled with Hilbert spectral analysis (Equations 52-55), RVMD resolves non-stationary transient growth, shift-modes, and limit-cycle harmonics in a two-dimensional cylinder wake at Re = 100 and isolates intermittent screech-tone feedback modes in a three-dimensional turbulent supersonic rectangular jet at M_j = 1.55.
Limitation
#nonconvex-initialization-sensitivity-and-nonorthogonal-mode-energy-sharing-limitation
Because the joint optimization over {phi_k, c_hat_k, omega_k} is non-convex and does not enforce strict spatial or temporal orthogonality across modes for general alpha > 0, converged RVMD modes depend on the initial central-frequency distribution {omega_k^0} and penalty factor alpha and can share overlapping spectral energy when bandwidths intersect (Section 2.2, p. 11; Section 3.1, p. 14; Appendix A, pp. 29-31).
⚠️ Curated Limitations & Negative Results
- Because the joint optimization over {phi_k, c_hat_k, omega_k} is non-convex and does not enforce strict spatial or temporal orthogonality across modes for general alpha > 0, converged RVMD modes depend on the initial central-frequency distribution {omega_k^0} and penalty factor alpha and can share overlapping spectral energy when bandwidths intersect (Section 2.2, p. 11; Section 3.1, p. 14; Appendix A, pp. 29-31).
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:DimensionalityReduction
Dimensionality Reduction
|
broad_match | RVMD extracts a low-rank rank-K spatiotemporal representation from high-dimensional space-time datasets with tunable spectral bandwidth compression. |
AIO:PrincipalComponentAnalysis
Principal Component Analysis
|
broad_match | In the zero-penalty broadband limit alpha -> 0 under spatial orthogonality, RVMD reduces exactly to Proper Orthogonal Decomposition (POD / PCA) on the two-point spatial correlation tensor. |
Evidence Sources & Primary Literature
3 Sources| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
r836-primary-pdf
|
Zi-Mo Liao, Zhiye Zhao, Liang-Bing Chen, Zhen-Hua Wan, Nan-Sheng Liu, Xi-Yun Lu
|
primary paper
full_text
|
ID: https://arxiv.org/abs/2209.12437
sha256:87b55cfa0ecf…
|
r836-arxiv-abs
|
Zi-Mo Liao, Zhiye Zhao, Liang-Bing Chen, Zhen-Hua Wan, Nan-Sheng Liu, Xi-Yun Lu
|
primary paper
abstract_only
|
ID: https://arxiv.org/abs/2209.12437
sha256:f51db16477ae…
|
r836-metadata-cr
|
Zi-Mo Liao, Zhiye Zhao, Liang-Bing Chen, Zhen-Hua Wan, Nan-Sheng Liu, Xi-Yun Lu
|
primary paper
abstract_only
|
ID: https://arxiv.org/abs/2209.12437
sha256:1102a9e39296…
|
Ontology & Grounding Topology
Open in interactive network graph →Curation Provenance
Curator: Antigravity
Method: ai_assisted
Model: Gemini
Created: 2026-10-19T15:19:00Z
Updated: 2026-10-19T15:19:00Z
Source File:
data/curated/RVMDReducedOrderVariationalModeDecomposition.yaml
· Schema:
curated_concept.yaml
Explore across X-Mech
Related computational & biological Mechs:
Embedding Map
Network Graph
ProteinTraitsMech (ESM-2 pLM)
DUFMech (Domain pLM)
TraitMech
CultureMech
All 14 X-Mech Repositories