CS Paper Radar

Memory in Behavioral Models as Motion on a Slow Invariant Manifold

Nicholas B. Tufillaro

eess.SP math.DS ·submitted 2026-09-30 01:57 UTC · v2
unusual

Abstract

A single-tone large-signal operating point of a nonlinear two-port is a periodic orbit of a periodically forced circuit. When the device has memory, the Floquet exponents of that orbit separate into fast (electrical) and slow (thermal and trapping) modes, and long-term memory is motion on the invariant manifold attached to the slow modes. Existence and uniqueness of that manifold follow from the parameterization method of Cabré, Fontich and de la Llave, applied to the stroboscopic map at the orbit; the manifold is the spectral submanifold of Haller and Ponsioen, without a small-forcing parameter. The manifold is a bundle over the circle of drive phase, its fiber dimension the number of slow exponents, and the dynamic X-parameter kernel of Verspecht et al. identifies its reduced dynamics from the transient after a step in drive amplitude. An exact reduced model has one memory state per slow exponent, the memoryless X-parameter surface is the fixed-point family of the reduced dynamics, and the envelope-domain model is the reduced dynamics driven by the envelope. The hypotheses are verified and the manifold constructed for a GaN HEMT compact model with thermal and trap memory: the trap's time constant at the operating point, $14 μ$s, is set by the linearization, not by its $6$ ms emission time, and the trap's nonlinearity confines a polynomial representation of the manifold to a few millivolts of drive, so trap memory must be represented over the operating range, by tables or a fitted network, not by an expansion about the operating point.

Comment
10 pages, 2 figures, 3 tables. v2: prior work attributed (the hidden-variable formulation of Root et al.; the dynamic gain model of Verspecht et al.); the trap's emission time and its linearized rate distinguished explicitly; wording revised throughout; results unchanged. Companion papers: arXiv:2609.35828 and arXiv:2610.01031. Toolkit v1.1.0: doi:10.5281/zenodo.23107935

Closest papers

Nearest neighbours by meaning across the whole index.

Why the Swing Parameters Are Sloppy

eess.SY ·2026-10-21 ·Sunny Kumar, Madhavi Parimi, Shashank Varma, Sushama Wagh, Alexsandar Stanković +1 more
theoryunusual similarity 0.52
Abstract

When a multi-parameter model is fitted to measured data, only a few parameter combinations are often identifiable, as Fisher Information Matrix eigenvalues span several orders of magnitude. Sensitivity-based tools, including the Fisher Information Matrix and MBAM, reveal this structure analytically, while manifold learning can recover it from sampled trajectories. For an infinite-bus synchronous generator, both approaches show that inertia H and damping D are among the least identifiable parameters on a late post-disturbance window, t in [3,5] s, where six of the eleven parameters suffice to describe the response. We explain this result from the geometry of the dynamics rather than from that of the parameter-to-data map. The rotor swing is a fast fibre that has collapsed onto a two-dimensional normally attracting invariant manifold (NAIM) of the transient fluxes before the window opens. The zeroth-order NAIM is exactly independent of $H$, $D$, $T_{d0}''$ and $T_{q0}''$, four of the five MBAM limits. We prove, under stated smoothness and rank conditions, that each of the four limits is a boundary of the model manifold at a distance equal, to first order in the parameter, to its current profiled speed. The same analysis predicts how identifiability changes with the experiment: neither a larger initial speed deviation nor a load change restores H or D in the late window, whereas the early window [0, 2]s makes H the best-determined parameter. Evaluated in a physical metric fixed by synchrophasor accuracy limits, the late window identifies five parameters rather than six and the early window nine. The spiked-covariance structure of the linearised trajectory map explains why the principal components of whitened, sampled trajectories reproduce the prior-scaled Fisher spectrum, and regression on those components attains the Fisher-predicted recovery error in 113 of 114 parameter scenario pairs.

0.24

Time Invariance, Circle Symmetry, and the Completeness of the Cardiff Behavioral Model

eess.SP ·2026-09-22 ·Nicholas B. Tufillaro
theoryunusual similarity 0.51
Abstract · 10 pages, 4 figures, 5 tables. v3: one sentence in Sec. VIII corrected (the multi-harmonic extension uses the single circle, not a torus); r…

Frequency-domain behavioral models of nonlinear microwave devices (the Cardiff model, X-parameters, higher-order sinusoidal describing functions, and the baseband models of power-amplifier predistortion) share one functional form, and each modeling framework justifies the form by its own argument, time invariance among them. Here we give the mathematical argument by which time invariance alone determines the model form, and state the result as a theorem for two-port devices, with its extension to any number of ports. Under a shift of the time origin each harmonic phasor rotates by a multiple of the fundamental's angle, so the spectral map of a time-invariant device is equivariant under a weighted circle action, and classical invariant theory gives its general form. The result is a completeness theorem for the form already in use: every time-invariant two-port response in single-tone periodic steady state is a phase factor attached to the fundamental multiplying a function of the wave magnitudes and their relative phase. The Cardiff exponents become counters: $m$ is the order of the load-side nonlinearity, $n$ is set by the drive-side harmonic, and $r$ is the number of conjugate pairs. The bound $r_{\max}=\lfloor K/2\rfloor$ for a load-side nonlinearity of degree $K$ makes the familiar restriction $r\le1$ a hypothesis about the device, exact for a cubic nonlinearity; for a loaded device the bound becomes a measurable decay in $r$. The results are illustrated with a published measurement and two simulations.

0.51

Floquet Fibre Geometry and Higher-Order Reduced Coordinates for Off-Manifold Transients near Nonlinear Aeroelastic Flutter

cs.CV ·2026-09-13 ·Puxue Tan
benchmarktheoryunusual similarity 0.49
Abstract

Assigning reduced coordinates to states near an attracting limit cycle requires the correct invariant-fibre geometry. The classical first-order phase-isostable chart obtained from adjoint Floquet modes projects along the strong-stable quotient fibre, whereas a metric-orthogonal complement of the retained slow bundle generally does not. We prove locally that a chart satisfying the linearised semiconjugacy relation leaves an O(delta^2) invariance residual, while projection along a non-invariant complement generically leaves an O(delta) term. For a nonlinear aeroelastic limit cycle, the metric-normal and strong-stable directions differ by 48.5 to 71.7 degrees, and metric-normal perturbations contain first-order retained phase and slow-amplitude components. Replacing the metric normal by the strong-stable fibre changes the measured residual scaling from delta^1.01 to delta^1.87 without fitted parameters. We then test learned higher-order corrections whose linearisation is pinned to the adjoint-Floquet chart, whose symmetry is exact, and whose reduced flow is fixed. Although they reduce the registered fixed-normalisation latent residual, post-hoc amplitude recalibration and adjoint-Floquet-targeted future consistency move or reverse the ranking. Because the learned maps already share the baseline's first-order gauge and the future target is supplied by the baseline chart, these diagnostics establish neither an independent positive nor negative higher-order result. Correct first-order Floquet geometry is therefore necessary in this benchmark, while the additional predictive value of the learned correction remains unidentified by the available representation-dependent diagnostics.

0.30

Intrinsic Associative Memory on Riemannian Manifolds: Curvature, Capacity, and Emergent Modes

cs.LG cs.AIstat.ML ·2026-09-28 ·Krishnakumar Balasubramanian, Zhaoyang Shi
theory similarity 0.48
Abstract

Geometry does more than constrain an associative memory: curvature determines what it remembers and which states it creates. We develop intrinsic dense associative memories on Riemannian manifolds by casting memory as Epanechnikov kernel-density mode seeking. We compare geodesic and volume-corrected energies and show that curvature separates their behavior. We prove that geodesic memory always retains an isolated pattern, while corrected memory obeys a sharp Ricci-curvature threshold: positive curvature can erase memories in high dimensions, while negative curvature reinforces them. We derive geodesic capacity scalings of $q_β^{-1/2}$ for retaining every pattern and $q_β^{-1}$ for a typical one, where $q_β$ is the pairwise kernel-overlap probability. We show how overlap \emph{creates} novel memories: designed $N$-pattern configurations realize all $2^N-1$ subset modes, but random data at the storage threshold yield only a Poisson number. We establish exact one-step recall using Riemannian mean shift. In simulations, we recover the predicted curvature transition and every designed mode. On WordNet's full noun hierarchy, we demonstrate that volume correction improves low-capacity retrieval. Together, our work shows that curvature is a design variable for associative memory, not merely a property of the data.

0.22

Bellman--Shoreline Search in Arbitrary Dimension: Exponential Vector Oscillators, Active Memory, Precession, and Effective Computability

cs.CG math.MGmath.OC ·2026-08-29 ·Florentin Koch
theory similarity 0.48
Abstract · 26 pages, 5 figures. Zenodo DOI: 10.5281/zenodo.22153950

We study online search for an unknown affine hyperplane in $\mathbb{R}^D$, for arbitrary fixed finite dimension. Building on a companion self-similar cell reduction and support-function formulation, we ask how the mechanism changes as the normal space grows from $\mathbb{S}^0$ to $\mathbb{S}^{D-1}$. In $D=1$, alternation and productivity yield an equal-ripple principle and the exact stationary constant $9$. In $D=2$, the analogous relative equilibrium is a logarithmic spiral whose bottleneck chord imposes tangency and selects the pitch. For exponential orbits $Γ(σ)=e^{κσ}ω(σ)$, we develop log-directional geometry, exponentially discounted memory, gauges, and recursive hyperspherical parametrizations. Without a shape ansatz, the bottleneck admits a certificate supported by at most $D$ historical suppliers, and at globally worst phases the current point lies on the active face. Within regular chambers we derive exact variation, tangency, pitch, age, and, in $D=3$, delay-system identities. Odd-dimensional obstructions, antipodal subclasses, and harmonic towers provide constraints and explicit candidate families but are not claimed globally optimal. Finally, the N-COMP theorem shows that $C_D^*$ is a computable real for every fixed finite $D$ and that algebraic polygonal $\varepsilon$-optimal cells can in principle be synthesized. Numerical screening through $D=10$ is kept separate from the proved results.

0.38

Beyond density snapshots: Dynamical memory in Chern-Simons reservoirs

cs.ET cs.AIcs.LGcs.NE ·2026-09-10 ·Jyotiranjan Beuria, Venkatesh H. Chembrolu
similarity 0.47
Abstract · 6 figures, 3 tables

How a reservoir encodes input history jointly in density and geometric degrees of freedom remains insufficiently understood. We address this mechanistic gap with a continuous density on an internal manifold coupled to an independently evolving Chern-Simons (CS) connection through a conserved current. On the same internal manifold, we compare reciprocal transport, instantaneous transverse reconstruction, local nonlinear feedback, and coupled CS evolution. These controls isolate how transport and independent connection evolution contribute to memory. We identify a current-generated exact connection component that the tested instantaneous reconstruction and scalar local closure miss. Resetting this component at fixed density and Gauss-law curl changes the future density response within the chosen gauge and drift law and the small effect persists under numerical refinement. We then distinguish dynamical memory from its readout accessibility. Although the four models have similar scalar-memory scores and density traces of pulse order, their circulating and longitudinal traces differ. With all measured features, coupled CS improves pulse-order decoding by \(6.5\pm1.9\) percentage points over the instantaneous-transverse control in paired seedwise comparisons, while remaining comparable to local feedback. Coupled CS shows less lag than the instantaneous and local-feedback controls on the first cycle, but more on subsequent cycles, revealing a dependence on preparation. We thus connect a field-theoretic mechanism of history retention to controlled tests of its dynamical role and computational use. The four-model framework provides a way to determine what an evolving transport field contributes when density snapshots and aggregate performance alone leave the memory mechanism unresolved.

0.23

Reduced-Order Model Characterization of Nonlinear Sustained Oscillations

eess.SY math.DS ·2026-09-27 ·Kaiyang Huang, Dan Wilson, Kai Sun
similarity 0.45
Abstract

Nonlinear sustained oscillations involving inverter-based resources can happen when device and network interactions produce an attracting limit cycle around an unstable equilibrium. A damping controller must then describe nonlinear dynamics over the region between the two invariant sets, where a local equilibrium linearization can be insufficient. In this paper, a controlled phase--isostable model is first developed on a two-dimensional invariant manifold connecting the equilibrium and the limit cycle. Dynamics on this surface are represented by phase and isostable coordinates, while physical states and control inputs are related to these coordinates through nonlinear reconstruction and response functions. The required functions are computed from connecting trajectories and periodic-orbit data. A control strategy is then designed to provide extra damping while limiting phase distortion. Case studies on a two-area system and an IEEE 39-bus system demonstrate more effective oscillation damping than the compared equilibrium-linearized designs under the same conditions.

0.27

Is Grokking a Loss of Normal Hyperbolicity of the Interpolation Manifold?

cs.LG ·2026-08-14 ·Suvinava Basak
similarity 0.45
Abstract

A recent line of work recasts the post-memorization phase of grokking as constrained optimization: once a network interpolates the training set, weight decay drives a slow drift along the zero-loss manifold toward lower norm. In the language of dynamical systems, this is a fast-slow system in which the interpolation manifold plays the role of a slow manifold. We ask a question that this framing makes natural but the existing literature does not address: is the sharp generalization transition a loss of normal hyperbolicity of that manifold: a fold- or bifurcation-like event in which a normal restoring direction goes flat? Or does the manifold stay uniformly attracting while generalization happens by smooth drift? We propose a simple, optimizer-agnostic diagnostic: the smallest nonzero singular value $σ_{\min}^{+}(\mathbf J)$ of the residual Jacobian, which, for the squared loss, equals the slowest normal restoring rate of the manifold. On a two-layer ReLU network trained to grok modular addition under squared loss, $σ_{\min}^{+}(\mathbf J)$ does not collapse at the transition; it is near zero only before memorization and attains its largest values during the transition. The result holds across five seeds, and the six smallest singular values behave identically; there is no subspace-local collapse either. This is preliminary evidence against the bifurcation hypothesis and in favor of the smooth-contraction picture. We are explicit that a single-setting, gradual-transition experiment under Adam optimizer does not prove the absence of a bifurcation; it constrains where one could hide.

0.04