Rhythm Inside Out: How an Algorithm Heard the Reverse Wave of the Universe's Birth ✛ ✎ ✦

Original: "Inverse-k Primordial Oscillations from a Symbolic Regression Search"
· Ze-Yu Peng, Qing-Yu Lan, Yun-Song Piao
arXiv:2607.04925 · 2026-07-06 · arXiv non-exclusive · independent analysis · 3 min
Symbolic regression, analyzing the cosmic microwave background, revealed reversed periodicity — a possible sign of an exotic expansion stage after the Big Bang.

The cosmic microwave background is a frozen portrait of the universe at the age of 380,000 years, when it first became transparent. Having stretched a thousandfold since then, this relic radiation stores in its minuscule temperature variations the imprint of events that unfolded in the first fractions of a second after the Big Bang. According to the theory of Alan Guth and Andrei Linde, inflation, the rapid stretching of space, smoothed the primordial ripples to near-perfect uniformity. But if the process wasn't flawlessly smooth, rhythmic features should have survived. Capturing this hidden rhythm means peering beyond the horizon of standard models.

Imagine a stone thrown into a pond doesn't produce the usual expanding circles, but waves that contract toward the center — ever faster and finer. That's the central metaphor of the discovery: the primordial power spectrum shows inverse oscillations that crowd together as the wave number k increases, like ripples receding into the depths of the microcosm. To catch this reverse wave, researchers used symbolic regression — an algorithm that blindly sifts through combinations of mathematical operators and independently finds the formula that best fits the radio astronomy data from the Planck satellite and the ACT and SPT-3G telescopes. No ready-made templates: the program listened to the cosmic background and composed its own scores.

Symbolic regression is Darwinian evolution of equations: over generations, the analytical expressions that most accurately predict the observed cosmic background survive. Once, while analyzing pendulum motion, it rediscovered the law of energy conservation, astonishing even its creators. Now it has set its sights on the mysteries of the Big Bang.

The result exceeded expectations. Both Planck and the combined SPA dataset independently pointed to the same form: f(k) ∝ cos(B/k) with a characteristic frequency B ≈ 4 Mpc⁻¹. In inverse-scale coordinates this is a pure harmonic, but in the familiar language of wave numbers it's a pattern where peaks follow ever closer, like a sound rapidly soaring upward. Spectroscopy of angular fluctuations and precise numerical simulations in the multipole range ℓ = 800–2000 showed that the inverse template yields a Δχ² ≈ −12.6 gain compared to the unperturbed spectrum — noticeably better than standard linear and logarithmic alternatives. Physically, this could mean that in the early universe there was a phase of expansion with an exotic equation of state resembling phantom energy: the scale factor grew as a(t) ∼ $t^p$ with p = −1, which requires violation of the null energy condition and fields beyond the Standard Model.

The amplitude of the inverse oscillations came out as A = 0.016 ± 0.01—right at the edge of statistical significance, but the very fact that an unbiased algorithm selected this exact form rather than more studied alternatives hints that we may have missed something important by relying only on ready-made recipes.

If the signal is real, it could become a direct test of quantum-gravity effects. But even as a statistical clue, it changes the rules of the game: an algorithm devoid of physical intuition pointed to a structure that theorists might never have thought to test: this machine "curiosity" expands our notions of what's possible. Future sky surveys of the cosmic microwave background by the Simons Observatory and CMB-S4, with their sensitivity and resolution, will either confirm the reverse ripple or tightly constrain its parameters. Meanwhile, symbolic regression, this "evolution of equations," is becoming a universal tool for unbiased pattern discovery, bypassing theorists' preconceptions. A stone thrown inside-out might have been the reality of the infant universe; now we're learning to hear this reverse rhythm.

Symbolic regression once independently rediscovered the law of conservation of mechanical energy — the first time an algorithm derived a physical law rather than just fitting parameters. Now it's aiming for the Big Bang.

P_R(k) = P_{R,0}(k) [1 + f(k)], \quad P_{R,0}(k) = A_s \left( \frac{k}{k_*} \right)^{n_s-1}
The standard power-law spectrum is multiplied by a factor (1+f), where f describes the sought oscillations, allowing the capture of minuscule deviations from the predictions of the simplest inflation model.
f_X(k) = A_X \cos\left( \omega_X \, \frac{k_*}{k} + \phi_X \right)
Amplitude A_X, frequency ω_X, and phase φ_X for argument k_*/k—this exact form, with a cosine of the inverse wave number, was picked out by symbolic regression as the best description of the data.
Scientists
Alan GuthAlexander FriedmannAndrei LindeGeorges LemaîtreJames PeeblesAdam Riess
Concepts
symbolic regression fundamental tone autoregressive model template first principles calculations