How to Fish the Laws of Nature Out of Noise express
A voice recorder captures a city square. You hear the hum of a crowd, footsteps, snatches of conversation, the rumble of traffic. Somewhere in that sonic stew, a street musician plays a simple tune on a flute. Hearing it isn't easy: there's just too much mixed together.
That's roughly the problem a new algorithm called Edwin (maximum-entropy compression with discovery of interpretable nonlinear models) tackles. It takes high-dimensional data (say, hundreds of sensor readings or thousands of video pixels) and looks for hidden simplicity. First, the algorithm compresses the data into a few key variables, as if isolating the pure flute sound from a noisy recording. Then it derives equations that describe how those variables change over time. And all this without losing touch with reality: each variable stays tied to measurable quantities, not becoming an abstract number.
The authors tested the method on six different systems. Among them: random particle walks, molecular self-assembly, the pulsing of a group of neurons, and even a real experiment with RNA-liposome complexes where the data were noisy. In each case, Edwin produced a short equation that wasn't just a fit to the numbers but had physical meaning. For diffusion, for instance, it derived a law resembling the heat equation; for neurons, a model reminiscent of a pendulum with friction.
The big surprise is predictive power. Models found under one set of conditions worked under others the algorithm hadn't been trained on. It's like recognizing a melody in a different tempo or played by a different instrument. That means Edwin truly captures the essence of a phenomenon, not just memorizes data.
This approach could change how scientists seek laws in complex systems — from climate to biological networks. Instead of guessing which variables matter and manually fitting equations, you can hand that job to an algorithm that acts like a seasoned listener, picking the melody out of the noise.
Original abstract
as written by the authors arXiv:2605.16724
Models (i.e., governing equations) are fundamental to science and engineering. Advances in data acquisition now make it possible to extract interpretable, low dimensional descriptions from high dimensional observations. However, existing approaches sacrifice either interpretability for reconstruction accuracy or infer symbolic dynamics without relating latent coordinates to physically meaningful observables. Here we present Edwin (maximum entropy driven compression with interpretable nonlinear model discovery), a unified framework that simultaneously performs dimensionality reduction using the dynamic maximum entropy (DME) principle and discovers sparse symbolic models governing latent dynamics, as well as the coupling between learned features and external metadata. We validate Edwin on diverse simulated systems, including stochastic diffusion, the Ornstein-Uhlenbeck process, self assembling particles, spiking neural populations, and low rank recurrent neural networks, as well as on a noisy experimental time series of aggregating RNA-liposome complexes. Across all systems, Edwin recovers low dimensional symbolic models that are physically interpretable and generalize to unseen conditions. Together, these results establish Edwin as a powerful framework for inferring interpretable, low dimensional dynamics directly from high dimensional data.
The name Edwin isn't an acronym but a nod to astronomer Edwin Hubble, who discovered the expansion of the universe by spotting a simple pattern in seemingly chaotic data on galaxy motion.
In Carl Sagan's novel Contact, aliens transmit blueprints for a machine hidden in the noise of a radio signal — much as Edwin extracts meaningful structure from chaos.