Can we learn the differential equations governing the evolution of a temporal network? We investigate this within Random Dot Product Graphs (RDPGs), where each network snapshot is generated from latent positions evolving under unknown dynamics. We identify three fundamental obstructions: gauge freedom from rotational ambiguity in latent positions, realizability constraints from the manifold structure of the probability matrix, and trajectory recovery artifacts from spectral embedding. We develop a geometric framework based on principal fiber bundles that formalizes these obstructions. We characterize invisible dynamics as exactly the skew-symmetric generators, and show the realizable tangent space has dimension $nd - d(d-1)/2$. An holonomy dichotomy emerges: polynomial dynamics have commuting generators, stationary eigenvectors, and trivial holonomy, making gauge alignment purely statistical; Laplacian dynamics satisfy a non-commutativity criterion producing nontrivial holonomy, with curvature weighted by $1/(λ_ι+ λ_γ)$ linking gauge sensitivity to the spectral gap. In $d=2$ this yields full restricted holonomy $\mathrm{SO}(2)$; for $d \ge 3$ generic full $\mathrm{SO}(d)$ remains conjectural. Cram'er--Rao lower bounds reveal that the same spectral gap controlling curvature and injectivity simultaneously controls Fisher information, so geometric and statistical difficulty are inextricable. We prove an identifiability principle: symmetric dynamics cannot absorb skew-symmetric gauge contamination, so dynamics structure can resolve gauge ambiguity. We demonstrate this constructively with anchor-based alignment and a UDE pipeline recovering vector fields from noisy graph sequences. Yet finite-sample interactions between noise, gauge, and dynamics expressiveness remain beyond the asymptotic theory. We frame this gap as an open challenge.
Alex: Welcome to another episode of ResearchPod.
Sam: Today we're looking at a paper titled "Random Dot Product Graphs as Dynamical Systems: Limitations and Opportunities." It studies networks that change over time, like food webs where species interactions shift. The key question is: can we recover the math rules driving those changes?
Alex: So it's about finding the hidden math behind how networks evolve, like in animal food chains?
Sam: Yes. We observe snapshots of these networks over time. But recovering the smooth rules hits three roadblocks: hidden rotations in the model, limits on possible changes, and glitches in estimating hidden parts. The paper uses geometry to explain why.
Alex: Food webs are a clear real-world example. What's the simplest way they model the evolution?
Sam: Each node—like a species—is a hidden point in a low-dimensional space. The chance of a link, say predator-prey, comes from how close those points are—like vectors overlapping to give a probability. That's a Random Dot Product Graph, or RDPG. Over time, points move, changing the network.
Alex: Those points move, but you mentioned roadblocks, starting with rotations?
Sam: Right. The hidden points aren't unique. You can rotate the whole set around the origin—like spinning a mobile—and link chances stay the same, since dot products ignore overall spin. This rotation freedom hides some changes from the network. The paper calls it gauge freedom—the first big block to spotting true motion rules.
Alex: So even smooth point motion looks scrambled because we can't tell real paths from spinning ones?
Sam: Exactly. Estimating points from noisy snapshots—using adjacency spectral embedding, which pulls leading patterns from the data—makes rotations jump between time steps. Smooth true motion appears erratic in estimates.
Alex: Those jumps make smooth motion look choppy. How do they untangle real paths from the rotation mess?
Sam: They use a geometric setup like layers of space. One layer holds observed link chances. Above it, a stack of possible hidden point sets all project to the same observations—like tracking a shark's wavy surface path from a boat, while its full swim twists unseen below. This stack is a principal fiber bundle, with "fibers" as rotation choices at each point.
Alex: The bundle organizes the gauge mess. But how do you pick the right path without arbitrary spins?
Sam: They split motion into two parts: one changing observations, and pure spin that doesn't. For any tiny push, they solve for the spin part—like finding the rotation rate explaining sideways wiggles. This defines "horizontal" motion: no spin, just observable change. They call the tool an Ehresmann connection.
Alex: Horizontal strips out gauge for a clean path lift. Does it always work perfectly?
Sam: It works locally: from any starting point, there's a unique horizontal lift of an observed path, staying spin-free. But over loops or long paths, curvature twists things—like parallel train tracks curving and meeting rotated. This mismatch, called holonomy, builds up. The paper shows it blocks rule learning, especially when curvature spikes from close eigenvalues in sparse nets.
Alex: Curvature from eigenvalue closeness causes drift. Do different rule types trigger this differently?
Sam: Yes. Polynomial rules on link chances—like sums of powers—commute, keeping holonomy zero: lifts close perfectly. But Laplacian flows, mixing degrees and links, don't commute, building holonomy that scrambles recovery. In food webs, this makes evolving interactions hard to learn without gauge fixes.
Alex: Polynomials avoid the trap because they commute, but Laplacians tangle up. What about fixes like embedding all time points together to smooth gauge jumps?
Sam: Joint embedding stacks snapshots and finds patterns across time, assuming shared directions for hidden points. But in evolving RDPGs, directions rotate as positions change, so it distorts true motion into fake tweaks.
Alex: It warps paths. What about Bayesian smoothing for continuous trajectories?
Sam: Those favor smooth position changes, like steady velocities. It helps choppy estimates but misses dynamical consistency: real rules tie velocity strictly to current position. Smoothing creates fluid paths ignoring that—ones that wander without matching equations.
Alex: You get pretty paths not ruled by equations. Geometry matters because holonomy blocks clean recovery. Changes in link chances give partial clues without full gauges?
Sam: Yes. For linear rules, changes follow a structured equation from latent motion. Noise amplifies errors where spectral gaps are small, linking to curvature.
Alex: Small gaps fuzz estimation. Even with identifiability, practical fixes hit walls?
Sam: Yes. Aligning rotations while fitting rules blurs signal in noise. Holonomy adds unfixable global drift. One fix: anchor nodes that barely move—like stable species—align everything to their fixed spots.
Alex: Anchors pin the gauge. How do tests bear out the theory?
Sam: On simulated food web data, anchors keep alignment steady over long paths, while step-by-step matching drifts. For rule learning—like spirals around centers—anchors recover equations far better than alternatives.
Alex: The geometry spots hurdles like holonomy and guides fixes like anchors for evolving networks. Thanks for joining ResearchPod.