arXiv subjects

Joshua L. Pughe-Sanford

Publications and source records attributed to Joshua L. Pughe-Sanford.

7 recordsLinked to original sources

arXiv2026

Degree-of-freedom and optimization-dynamic effects on the observability of Kuramoto-Sivashinsky systems

Simulations of chaotic systems can only produce high-fidelity trajectories if the initial and boundary conditions are well specified. When these conditions are unknown but measurements are available, variational state estimation can reconstruct a trajectory that is consistent with both the data and the governing equations. A key open question is how many measurements are required for accurate reconstruction, making the full system trajectory observable from sparse data. We establish observability criteria for variational state estimation applied to the Kuramoto-Sivashinsky equation by linking its observability to embedding theory for dissipative dynamical systems. For a system whose attractor lies on an inertial manifold of dimension $d_M$, we show that $m \geq d_M$ measurements ensures local observability from an arbitrarily good initial guess, and $m \geq 2d_M + 1$ implies global observability using a gradient-based smoother since the only critical point on $M$ is the global minimum. We also analyze optimization-dynamic limitations that persist even when these topological conditions are met, including drift off the manifold, degeneracy of the Hessian, negative curvature, and vanishing gradients. To address these issues, we introduce a robust reconstruction strategy that combines non-convex Newton updates with a novel pseudo-projection step. Numerical simulations of the Kuramoto-Sivashinsky equation validate our analysis and show practical limits of observability for chaotic systems with low-dimensional inertial manifolds.

math.DS↗

arXiv2025

A Network of Biologically Inspired Rectified Spectral Units (ReSUs) Learns Hierarchical Features Without Error Backpropagation

We introduce a biologically inspired, multilayer neural architecture composed of Rectified Spectral Units (ReSUs). Each ReSU projects a recent window of its input history onto a canonical direction obtained via canonical correlation analysis (CCA) of previously observed past-future input pairs, and then rectifies either its positive or negative component. By encoding canonical directions in synaptic weights and temporal filters, ReSUs implement a local, self-supervised algorithm for progressively constructing increasingly complex features. To evaluate both computational power and biological fidelity, we trained a two-layer ReSU network in a self-supervised regime on translating natural scenes. First-layer units, each driven by a single pixel, developed temporal filters resembling those of Drosophila post-photoreceptor neurons (L1/L2 and L3), including their empirically observed adaptation to signal-to-noise ratio (SNR). Second-layer units, which pooled spatially over the first layer, became direction-selective -- analogous to T4 motion-detecting cells -- with learned synaptic weight patterns approximating those derived from connectomic reconstructions. Together, these results suggest that ReSUs offer (i) a principled framework for modeling sensory circuits and (ii) a biologically grounded, backpropagation-free paradigm for constructing deep self-supervised neural networks.

q-bio.NC↗

arXiv2025

Neurons as Detectors of Coherent Sets in Sensory Dynamics

We model sensory streams as observations from high-dimensional stochastic dynamical systems and conceptualize sensory neurons as self-supervised learners of compact representations of such dynamics. From prior experience, neurons learn coherent sets-regions of stimulus state space whose trajectories evolve cohesively over finite times-and assign membership indices to new stimuli. Coherent sets are identified via spectral clustering of the stochastic Koopman operator (SKO), where the sign pattern of a subdominant singular function partitions the state space into minimally coupled regions. For multivariate Ornstein-Uhlenbeck processes, this singular function reduces to a linear projection onto the dominant singular vector of the whitened state-transition matrix. Encoding this singular vector as a receptive field enables neurons to compute membership indices via the projection sign in a biologically plausible manner. Each neuron detects either a predictive coherent set (stimuli with common futures) or a retrospective coherent set (stimuli with common pasts), suggesting a functional dichotomy among neurons. Since neurons lack access to explicit dynamical equations, the requisite singular vectors must be estimated directly from data, for example, via past-future canonical correlation analysis on lag-vector representations-an approach that naturally extends to nonlinear dynamics. This framework provides a novel account of neuronal temporal filtering, the ubiquity of rectification in neural responses, and known functional dichotomies. Coherent-set clustering thus emerges as a fundamental computation underlying sensory processing and transferable to bio-inspired artificial systems.

q-bio.NC↗

arXiv2025

Computing Chaotic Time-Averages from Few Periodic or Non-Periodic Orbits

For appropriately chosen weights, temporal averages in chaotic systems can be approximated as a weighted sum of averages over reference states, such as unstable periodic orbits. Under strict assumptions, such as completeness of the orbit library, these weights can be formally derived using periodic orbit theory. When these assumptions are violated, weights can be obtained empirically using a Markov partition of the chaotic set. Here, we describe an alternative, data-driven approach to computing weights that allows for an accurate approximation of temporal averages from a variety of reference states, including both periodic orbits and non-periodic trajectory segments embedded within the chaotic set. For a broad class of observables, we demonstrate that weights computed with the proposed method significantly outperform those derived from periodic orbit theory or Markov models, achieving superior accuracy while requiring far fewer states -- two critical properties for applications to high-dimensional chaotic systems.

math.DS↗

arXiv2022

Observing a Dynamical Skeleton of Turbulence in Taylor-Couette Flow Experiments

Recent work suggests unstable recurrent solutions of the equations governing fluid flow can play an important role in structuring the dynamics of turbulence. Here we present a method for detecting intervals of time where turbulence "shadows" (spatially and temporally mimics) recurrent solutions. We find that shadowing occurs frequently and repeatedly in both numerical and experimental observations of counter-rotating Taylor-Couette flow, despite the relatively small number of known recurrent solutions in this system. Our results set the stage for experimentally-grounded dynamical descriptions of turbulence in a variety of wall-bounded shear flows, enabling applications to forecasting and control.

physics.flu-dyn↗

arXiv2021

Exact coherent structures and shadowing in turbulent Taylor-Couette flow

We investigate a theoretical framework for modeling fluid turbulence based on the formalism of exact coherent structures (ECSs). Although highly promising, existing evidence for the role of ECSs in turbulent flows is largely circumstantial and comes primarily from idealized numerical simulations. In particular, it remains unclear whether three-dimensional turbulent flows in experiment shadow any ECSs. In order to conclusively answer this question, a hierarchy of ECSs should be computed on a domain and with boundary conditions exactly matching experiment. The present study makes the first step in this direction by investigating a small-aspect-ratio Taylor-Couette flow with naturally periodic boundary condition in the azimuthal direction. We describe the structure of the chaotic set underlying turbulent flow driven by counter-rotating cylinders and present direct numerical evidence for shadowing of a collection of unstable relative periodic orbits and a traveling wave, setting the stage for further experimental tests of the framework.

physics.flu-dyn↗

arXiv2017

Renormalization of Discrete-Time Quantum Walks with non-Grover Coins

We present an in-depth analytic study of discrete-time quantum walks driven by a non-reflective coin. Specifically, we compare the properties of the widely-used Grover coin ${\cal C}_{G}$ that is unitary and reflective (${\cal C}_{G}^{2}=\mathbb{I}$) with those of a $3\times3$ "rotational" coin ${\cal C}_{60}$ that is unitary but non-reflective (${\cal C}_{60}^{2}\not=\mathbb{I}$) and satisfies instead ${\cal C}_{60}^{6}=\mathbb{I}$, which corresponds to a rotation by $60^{\circ}$. While such a modification apparently changes the real-space renormalization group (RG) treatment, we show that nonetheless this non-reflective quantum walk remains in the same universality class as the Grover walk. We first demonstrate the procedure with ${\cal C}_{60}$ for a 3-state quantum walk on a one-dimensional (\emph{1d}) line, where we can solve the RG-recursions in closed form, in the process providing exact solutions for some difficult non-linear recursions. Then, we apply the procedure to a quantum walk on a dual Sierpinski gasket (DSG), for which we reproduce ultimately the same results found for ${\cal C}_{G}$, further demonstrating the robustness of the universality class.

quant-ph↗