Searcharxiv⌕ Search

arXiv subjects

Harry Desmond

Publications and source records attributed to Harry Desmond.

At least 19 recordsLinked to original sources

Galactic warps constrain equivalence-principle violation in the dark sector

Current measurements of the expansion of the Universe allow for the presence of evolving dark energy, distinct from a cosmological constant, with an equation of state that seems to violate the null energy condition. One notable explanation for this behaviour is that the dark energy interacts with the dark matter, violating the equivalence principle yet evading conventional laboratory and astrophysical constraints. Such models produce fifth forces acting only on dark matter. We show how such forces separate halo and stellar disk centres, creating a potential gradient across the disk that warps it into a cup shape. Measuring the warps of $4\times10^3$ edge-on disks in Sloan Digital Sky Survey images and modelling the fifth-force signal together with competing baryonic effects, we find no sign of the predicted correlation between the measured warps and the fifth-force field. This constrains the fifth-force strength relative to gravity to $ΔG/G_N<7.7\times10^{-4}$ at 95% credibility (or $<1.7\times10^{-3}$ once systematic uncertainties are included) for force ranges of 8 Mpc and above. In the infinite-range limit of a scalar rolling today, even the conservative bound is three times stronger than the tightest cosmological bound (which, unlike ours, depends on the scalar's cosmological evolution) and about two orders of magnitude below other astrophysical bounds. It disfavours at 2-3$σ$ the couplings preferred by recent fits of interacting dark energy to cosmological data unless gravitational screening is invoked.

astro-ph.CO↗

Does the inverse distance ladder rule out late-time solutions to the Hubble tension? A model-independent test using DES-Dovekie supernovae and DESI DR2 BAO

It has been argued that the inverse distance ladder disfavours late-time solutions to the Hubble tension. We test this with a cosmographic expansion where $H$ is a fourth-order polynomial in $1 - a$. We combine the uncalibrated DES-Dovekie SNIa with DESI~DR2 BAO, assuming a standard sound horizon $r_{\rm d} = 147.05 \pm 0.30$~Mpc. Using only DES SNIa yields $H_0 = 76.25 \pm 2.87$ km/s/Mpc, which is $3.1σ$ above the CMB-calibrated $Λ$CDM prediction but consistent with SH0ES ($1.0σ$). Adding the legacy SNIa decreases $H_0$ to $67.98 \pm 0.69$ km/s/Mpc, $4.7σ$ below SH0ES. The $8.27$ km/s/Mpc drop in $H_0$ upon including the legacy SNIa cannot be attributed to expected sample variance: under a fiducial $Λ\text{CDM}$ model, only $1.24\%$ of $10^4$ mock realisations exhibit such a large $H_0$ shift due to the changed redshift sampling. This points to a systematic tension between DES and legacy SNIa datasets. We next add local $H_0$ anchors (the bright siren, SNII, TRGB + SBF) and test a parametrisation where the $Λ$CDM $H(z)$ is enhanced by $ΔH(z) = A \exp (-z/z_1)$. We recover $H_0 = 70.00^{+4.78}_{-0.98}$ km/s/Mpc and $z_1 < 0.086$ at $1σ$ confidence using DES SNIa, but $H_0 = 67.66 \pm 1.01$ km/s/Mpc and $z_1$ is unconstrained using DES + legacy. Late-time solutions are thus viable with the homogeneous DES catalogue, but not with the legacy SNIa as published. Allowing them to be arbitrarily recalibrated relative to the main DES sample yields an inferred offset of $ΔM_\rm{legacy} = 0.0646 \pm 0.0391$~mag, raising the predicted $H_0$ to $71.03 \pm 1.99$ km/s/Mpc. Ongoing SNIa surveys like DEBASS will provide a homogeneous sample of SNIa down to redshifts accessible via the distance ladder, assessing the validity of the current legacy samples and hence the feasibility of late-time solutions to the Hubble tension.

astro-ph.CO↗

The age of the Universe from a large sample of the oldest Galactic stars

We estimate the age of the Universe using the Xiang & Rix sample of 247,103 Milky Way stars with low-resolution spectroscopy from LAMOST DR7 and $Gaia$ eDR3 parallaxes. Stellar ages were estimated using YY isochrones up to 20 Gyr. We require consistency between YY ages and photometric FLAME ages based only on $Gaia$ data. To remove a population of stars with unusually low vertical action for their age, metallicity, and distance, we require old stars to be metal-poor and $α$-enriched. Our final sample of 146,810 stars within 5 kpc provides consistent cosmic age estimates using several techniques of increasing rigour. Our main results use an MCMC reconstruction of the latent age distribution, though our iterative reconstruction is very similar. Applying an innovative approach to our MCMC reconstruction and its uncertainties, we find that the oldest star has an age of $A_\star = 13.80^{+0.20}_{-0.16}$ Gyr. We conservatively estimate a systematic uncertainty of $\pm 0.4$ Gyr due to the choice of quality cuts, isochrone model, and initial helium abundance. Allowing 0.2 Gyr for the first long-lived stars to form, our inferred $A_\star$ is consistent with the 13.6 Gyr expected in CMB-calibrated $Λ$CDM. This agreement casts doubt on solutions to the Hubble tension solely through new physics prior to recombination, which generally imply a cosmic age of $12.9 \pm 0.2$ Gyr to match low redshift probes. It is difficult for stellar modelling uncertainties to reconcile such a low age with our result given the low metallicities of the oldest stars in our sample and independent asteroseismic constraints.

astro-ph.CO↗

Stars Would Remember a Recent Gravitational Transition: Testing G-step Resolutions of the Hubble Tension with Red Dwarfs

We test the hypothesis that Newton's gravitational constant underwent a rapid transition approximately $73$-$130\,{\rm Myr}$ ago, as proposed in some resolutions of the Hubble tension. Fully convective red dwarf stars with masses below $\sim0.35\,M_\odot$ have thermal timescales comparable to the proposed transition look-back times, so an appreciable present-day memory can remain. We show that a rapid step in $G$ induces measurable changes in their radii and confront these predictions with mass-radius measurements of eight red dwarfs in detached eclipsing binaries. We find that the $G$-step models proposed to resolve the Hubble tension, corresponding to a $4\%$ change in $G$ at a look-back time of $73\,{\rm Myr}$ and an $11\%$ change at $130\,{\rm Myr}$, are excluded at significances of $3.1σ$ and $4.0σ$, respectively.

astro-ph.CO↗

Disentangling modified gravity and galaxy bias with field-level inference

We present a field-level inference framework for testing gravity with large-scale structure, exploiting the full information of the galaxy distribution. Traditional analyses based on the power spectrum discard non-Gaussian and Fourier phase information, resulting in strong degeneracies between modified gravity (MG) and galaxy bias. Our approach overcomes this limitation by performing a Bayesian analysis directly on the three-dimensional galaxy number counts, jointly constraining MG and bias parameters using both amplitudes and phases. As an illustrative application, we analyse mock data in real space in the context of $f(R)$ gravity and a non-linear galaxy bias model designed to mimic the real-data 2M++ BORG analysis of Jasche&Lavaux (2019). Non-linear structure formation is modelled using COmoving Lagrangian Acceleration (COLA) under different gravity strengths, parameterised by $f_{R0}$. The resulting dark-matter fields are then mapped to mock galaxy catalogues via a non-linear bias prescription. We demonstrate that, with the initial phases assumed known, including non-Gaussian and phase information yields tighter constraints on both $f_{R0}$ and the primary bias parameter $β$ (corresponding to the linear galaxy bias on large scales), relative to power-spectrum-only analyses. Notably, the field-level approach breaks the degeneracies between MG and galaxy bias inherent to two-point statistics. Through a cosmic-web classification into voids, walls, filaments and clusters, we find that under-dense regions are the primary drivers in distinguishing gravity at the field level. Finally, we establish the robustness of our pipeline against variations in initial conditions, Poisson noise, and galaxy-field thresholding, providing a powerful path forward for field-level tests of gravity.

astro-ph.CO↗

Two-rung ladder: $H_0$ from Tip of the Red Giant Branch and geometric anchors alone

We infer the Hubble constant $H_0$ from the ~400 Extragalactic Distance Database hosts with a tip of the red giant branch (TRGB) distance, calibrated on the geometric anchors of the Large Magellanic Cloud and NGC 4258. We forward-model the tip magnitudes and host redshifts including magnitude-limited selection, sky exposure, and density-dependent galaxy bias, conditioning on each of the 80 Manticore-Local realisations of the local density and peculiar-velocity fields in turn. As most hosts lie at $cz_{\rm CMB} \lesssim 1000$ km/s, where peculiar velocities are important, the inference is limited by the variance of these fields rather than by the TRGB calibration. For a fixed realisation, the redshift likelihood and density smoothing each move $H_0$ by less than 1 km/s/Mpc and a velocity monopole by $5.4 \pm 3.4$ km/s/Mpc, whereas the field-to-field scatter in $H_0$ is 5.2 km/s/Mpc. Weighting the posteriors from each realisation equally gives $H_0 = 68.4 \pm 5.9$ km/s/Mpc, while weighting each by its Bayesian evidence gives $72.3 \pm 2.6$ km/s/Mpc. The latter value is set by a single particularly high-evidence realisation, making it sensitive to the finite sampling of the reconstruction posterior. This ladder therefore cannot yet arbitrate the Hubble tension. Extending the TRGB data beyond the present ~1000 km/s, together with improved reconstructions of the local Universe, would reduce the field-to-field scatter well below the 5.2 km/s/Mpc we find here.

astro-ph.CO↗

A reanalysis of the megamaser Hubble constant: from spot catalogues to peculiar velocities

The Hubble tension motivates careful scrutiny of the redshift-independent distances underpinning the local distance ladder. Water megamasers afford a purely geometric distance from orbital dynamics in an accretion disc. For five of the six Megamaser Cosmology Project (MCP) galaxies, we present a Bayesian forward model that infers the angular-diameter distance and warped Keplerian disc parameters jointly from the very long baseline interferometry quasi-observables: maser spot positions, velocities, and accelerations. We then constrain $H_0$ from these distances and the host redshifts, including a physical distance prior, modelling the galaxy sample selection and testing a linear and a non-linear peculiar-velocity reconstruction. Accounting for peculiar velocities with the non-linear Manticore reconstruction, we infer $H_0 = 71.6 \pm 3.7$ km/s/Mpc, where the uncertainty is statistical. Relative to ignoring peculiar velocities, the linear and non-linear reconstructions lower $H_0$ by $0.8$--$1.4$ and $1.6$--$2.1$ km/s/Mpc, respectively, whereas changing the selection variable alters it by less than $1$ km/s/Mpc. We analyse the sixth galaxy, NGC 4258, separately. Modelling its disc as eccentric and quadratically warped, we infer an angular-diameter distance of $7.558 \pm 0.073$ (stat.) Mpc, within $0.13σ$ of the published maser distance.

astro-ph.CO↗

Constraints on the gravitational potential from DESI DR2 BAO and its implications for the local void scenario

We constrain the difference in gravitational potential between our location and sources at $z \gtrsim 0.3$ using datasets at those redshifts. Our motivation is that the Hubble tension might be caused by a local void, as suggested by galaxy number counts. This would increase the redshift through outflow and gravitational redshift (GR). Only the latter is important at high redshift, where a void contributes a fixed additional GR contribution of $z_0$ due to our location on a potential hill. This $z_0$ model has various subtle effects that were not previously considered, including a hotter CMB and reduced BAO scale $r_{\rm d}$. We test whether $z_0$ can have the previously expected value of 0.84%, which was based on fitting void parameters to galaxy number counts and local $H_0$ measurements. Combining BBN, CMB, BAO, and CC datasets at $z > 0.5$, we find that $z_0 = -0.4^{+0.8}_{-0.9}\%$, which rises to $-0.1 \pm 0.7\%$ when extending our analysis down to $z > 0.29$. Although the results prefer the standard value of $z_0 = 0$, the best-fitting model with $z_0 = 0.84\%$ fits the data almost as well as $Λ$CDM, with $Δχ^2 < 2$. We find that $Λ$CDM faces a $3.07σ$ BAO anomaly in the standard $(H_0 r_{\rm d}, Ω_{\rm m})$ parameter space, where different regions are preferred by BAO and non-BAO datasets from $z > 0.29$. Fixing $z_0 = 0.84\%$ reduces this to $2.79σ$. This suggests that a local void large enough to solve the Hubble tension cannot be ruled out by higher-redshift datasets despite its novel impacts on them.

astro-ph.CO↗

MIGHTEE-HI / LADUMA: Investigating the link between baryons and dynamics with 130 resolved HI-selected galaxies

The baryonic Tully-Fisher relation (bTFR) and the radial acceleration relation (RAR) link the observed dynamics in galaxies to that expected from their baryonic mass distributions. The relations' small intrinsic scatters place strong constraints on galaxy formation models, dark matter properties and theories of modified dynamics, yet detailed measurements beyond the very local Universe remain limited. We use 130 purely HI-selected galaxies with resolved HI kinematics and baryonic mass profiles to measure the bTFR and RAR up to $z\approx0.09$. We measure a tight RAR with an acceleration scale $a_0=(1.50\pm0.05)\times10^{-10},{\rm m,s^{-2}}$ and an intrinsic scatter of $0.096\pm0.006$ dex, consistent with local results. We fit the bTFR in the `inverse' direction, conditioning on $M_{\rm bar}$ to mitigate HI flux-related selection effects, measuring a logarithmic slope of $0.27\pm0.01$ (corresponding to a forward slope of $3.72\pm0.16$), with vertical intrinsic scatter $σ_\perp\approx0.05$ dex. Fitting the general $δ$-family of MOND interpolating functions to the RAR, we infer $δ=4.10^{+1.4}_{-0.68}$, consistent with the value required by Solar System gravitational constraints and a null Wide Binary Test. We find no significant redshift evolution in the RAR acceleration scale for our pure HI-selected sample. However, the bTFR zero-point shows an apparent evolutionary trend that is strongly dependent on the fit direction: the traditional forward fit yields an $8.7σ$ preference for $z$ evolution, while for our fiducial inverse fit, this reduces to $3.4σ$, within $\approx2σ$ of the RAR evolution constraint. This suggests selection effects bias the forward fit; a careful consideration of such effects will be required in future endeavours to robustly measure the redshift evolution of dynamical scaling relations.

astro-ph.GA↗

Forward-modelling Milky Way Cepheids: selection effects and physical priors in the Gaia-HST calibration

The advent of high-precision Gaia parallaxes for Milky Way Cepheids enables per cent-level calibration of the local distance ladder and the Hubble constant $H_0$. We revisit the Milky Way Cepheid calibration from Gaia EDR3 parallaxes using a fully forward-modelled Bayesian framework that simultaneously infers the period--luminosity relation, the Gaia parallax zero-point offset, and individual stellar distances while explicitly incorporating the disc geometry of the Galaxy through the distance prior and the selection functions specified in two HST SH0ES campaigns. We derive an analytic treatment of the detection probability that accounts for magnitude, parallax, period, and extinction cuts and reduces it to a tractable integral over distance and sky position. Posterior predictive checks show that this generative model matches the observed distributions of parallaxes, magnitudes, and periods. Modelling Galactic structure and survey truncation self-consistently in a Bayesian framework yields period--luminosity parameters that agree with the SH0ES maximum-likelihood values at the ${<}0.5\,σ$ level, a consequence of the small intrinsic scatter of the Cepheid period--luminosity relation. Adopting the uniform-in-volume prior recently advocated by Högås & Mörtsell (2026), without simultaneously accounting for selection, leads to a ${\sim}\,0.05~\mathrm{mag}$ bias in the period--luminosity zero-point and posterior predictive distributions incompatible with the observed data; this shift is mostly driven by the omission of the selection model, and produces an apparent and unjustified shift in $H_0$ that reflects this mismodelling. A consistent Bayesian treatment of Galactic structure and selection effects reinforces the local distance-ladder determination of $H_0$, and hence the Hubble tension with early-Universe inferences.

astro-ph.GA↗

Evaluation of Population Initialization Methods for Genetic Programming-based Symbolic Regression

We analyze the effect of optimizing the initial population of genetic programming (GP) for symbolic regression (SR) on the accuracy and complexity of solutions. We compare three well-established random initialization methods as well as initialization with small optimized solutions from exhaustive symbolic regression (ESR) using a GP/SR implementation which is based on the multi-objective evolutionary algorithm NSGA-II. We compare the final Pareto fronts found with each initialization method on twelve synthetic problems of varying complexity and one real-world dataset. We find no significant differences in accuracy or model complexity among the initialization methods. The initial advantage of initialization with ESR disappears after only a few generations. Our results show that, given similar diversity in the initial population, the effect of the initialization method in GP-based symbolic regression on the final Pareto front is negligible.

cs.NE↗

Forward-modelling the Tolman and distance-duality tests with IllustrisTNG

The Tolman surface-brightness test and the angular-size distance-duality test are two complementary probes of the same underlying relation between luminosity and angular-diameter distance, $D_L = (1+z)^2 D_A$, as holds in any metric theory of gravity where photon number is conserved. Both tests have recently delivered a priori surprising signals: JWST/ASTRODEEP measurements yield a surface brightness scaling with redshift much flatter than the expected value, and ultracompact radio sources also appear to follow a flatter $D_L/D_A$ scaling with redshift. These results have been suggested to support non-expanding cosmologies, however they are also sensitive to astrophysical and instrumental effects. We test whether these results indicate genuine departures from standard cosmology by forward-modelling observed surface-brightness evolution in the IllustrisTNG cosmological hydrodynamical simulation, with an empirical mock-spectroscopic selection trained on ASTRODEEP. We show that the astrophysical evolution relevant for both tests may be effectively parametrised as a single power-law exponent for the luminosity density as a function of redshift, for which the simulation gives $γ=2.23\pm0.20$ across realistic aperture conventions. This value is approximately sufficient to explain both the Tolman and distance-duality signals within standard cosmology and galaxy formation physics, with a small discrepancy for the latter suggesting that radio AGN evolve slightly more strongly than bright galaxies.

astro-ph.CO↗

The subtle statistics of the distance ladder: On the distance prior and selection effects

Statistical methodology is rarely considered significant in distance-ladder studies or a potential contributor to the Hubble tension. We suggest it should be, highlighting two appreciable issues. First, astronomical distances are inferred latent parameters, requiring a prior. We show that the (often implicit) uniform priors on distance moduli common to Bayesian distance-ladder analyses bias distances low due to objects being uniformly distributed in volume, which biases the Hubble constant high. Frequentist $χ^2$ methods are unbiased for volume- or redshift-limited samples only if the redshift uncertainty (including peculiar velocities) vanishes, though simulation-based calibration can correct the bias. Second, in a Bayesian framework, selection effects introduce additional posterior factors describing the probability of objects entering the sample under the model. These partly counteract the volume prior, depending on the nature of the selection. After detailed analytic and mock-based studies, we quantify the volume-prior effect in the CosmicFlows-4 and SH0ES samples. Both use frequentist methods, so the effect appears as a potential estimator bias rather than a missing prior. The implied Hubble constant shifts are significant but must not be applied na\"ıvely -- principled selection modelling is also required, as we investigate explicitly for CosmicFlows-4. Both effects should already be captured by the SH0ES pipeline's simulation-based bias corrections. Our work highlights the crucial need to model both distances and selection accurately, either directly in a Bayesian forward model, or via post-hoc simulation-based corrections with realistic source and selection distributions. Such modelling requires samples with known, homogeneous selection criteria, which future surveys should prioritise.

astro-ph.CO↗

Two per cent measurement of $H_0$ from Cepheids alone

One of the most pressing problems in current cosmology is the cause of the Hubble tension. We revisit a two-rung distance ladder, composed only of Cepheid periods and magnitudes, anchor distances in the Milky Way, Large Magellanic Cloud, NGC 4258, and host galaxy redshifts. We adopt the SH0ES data for the most up-to-date and carefully vetted measurements, where the Cepheid hosts were selected to harbour also Type Ia supernovae. We introduce two important improvements: a rigorous selection modelling and a state-of-the-art density and peculiar velocity model using Manticore-Local, based on the Bayesian Origin Reconstruction from Galaxies (BORG) algorithm. We infer $H_0 = 71.1 \pm 1.4~\mathrm{km}\,\mathrm{s}^{-1}\,\mathrm{Mpc}^{-1}$, assuming the Cepheid host sample was selected by supernova magnitudes. However, the actual selection criteria are not clear, and other assumptions can increase $H_0$ by up to one statistical standard deviation. The posterior has a lower central value and a 41 per cent smaller uncertainty than a previous study using the same distance-ladder data. This result is lower than the supernova-based SH0ES inferred value of $H_0 = 73.2 \pm 0.9~\mathrm{km}\,\mathrm{s}^{-1}\,\mathrm{Mpc}^{-1}$ at about $1.3σ$, and is in $2.8σ$ tension with the latest cosmic microwave background results in the standard cosmological model. These results demonstrate that a measurement of $H_0$ of sufficient precision to weigh in on the Hubble tension is achievable using second-rung data alone, underscoring the importance of robust and accurate statistical and velocity-field modelling.

astro-ph.CO↗

CMBolic: Symbolic emulators for the Cosmic Microwave Background. I. Lensing

We present the first installment of CMBolic: a suite of symbolic cosmic microwave background (CMB) emulators. In this instance, we emulate the CMB lensing potential power spectrum $C_\ell^{ϕϕ}$ for the widely used extended $Λ$CDM model which simultaneously includes massive neutrinos and evolving dark energy modelled using the Chevallier-Polarski-Linder (CPL) parameterization. We achieve comparable precision to existing neural network emulators, with the added benefit of simpler handling as our emulators are analytic functions of the model parameters and multipole $\ell$. On independent validation spectra evaluated in the range $2\leq \ell \leq 5500$, CMBolic achieves mean absolute fractional errors of $0.27\%$ in the $Λ$CDM subspace and $0.32\%$ across the full extended parameter space. This emulation error is well below even the most optimistic noise forecasts from CMB Stage 4 experiments. We apply CMBolic to cosmological parameter estimation with Bayesian inference using the lensing-only likelihoods from ACT DR6 and Planck. We show excellent agreement between the posteriors obtained by CMBolic and the Boltzmann code CLASS. This demonstrates the practical use of CMBolic on cosmological parameter estimation, reducing the runtime from 2 weeks to under 3 minutes.

astro-ph.CO↗

Validating Digital Twins of the Local Universe with the Thermal Sunyaev-Zel'dovich Signal

The thermal Sunyaev-Zel'dovich (tSZ) effect provides a powerful probe of the thermal pressure of ionised gas in galaxy clusters and the cosmic web; constrained simulations reconstruct the mass and velocity fields of the local Universe. We explore how these two may be mutually informative: the tSZ signal provides a benchmark for assessing the fidelity of constrained simulations, and constrained simulations contribute information on the positions, total masses and density profiles of cosmic web structures for use in tSZ studies. We focus on cluster predictions in the Bayesian Origin Reconstruction from Galaxies (BORG) paradigm, introducing CSiBORG-Manticore, a new state-of-the-art suite of digital twins -- data-constrained posterior simulations whose initial conditions are inferred via Bayesian forward modelling. We develop a framework for scoring constrained simulations on their ability to match measured Planck Compton-$y$ maps around clusters, and use it to demonstrate improvement from previous BORG reconstructions. We further validate halo masses against weak-lensing-calibrated X-ray masses from eROSITA. We also show how high-fidelity digital twins offer a practical route to extracting additional information from tSZ data through a novel calibration of the mass-observable relation, and provide a complementary framework to purely statistical analyses of Compton-$y$ maps. This paves the way for integrating the large-scale structure information inherent in constrained simulations into the study of CMB secondary anisotropies.

astro-ph.CO↗

Precision constraints on stellar physics from main sequence detached eclipsing binaries

We present a Bayesian framework to constrain {mass ($M$), metallicity ($Z$), stellar age ($τ$), and the convective mixing length parameter ($α_{\rm MLT}$)} in main-sequence (MS) detached eclipsing binaries (DEBs). These systems provide precise values of stellar mass and radius, offering stringent tests of stellar evolution models. We combine these with broadband magnitudes in the $B$ and $V$ bands and Gaussian priors on spectroscopic mass and metallicity, and perform Markov Chain Monte Carlo inference using a fast machine-learning surrogate for one-dimensional stellar evolution models computed with Modules for Experiments in Stellar Astrophysics. To make this approach computationally feasible, we implement an active learning strategy that adaptively selects new stellar models to evaluate, concentrating training data in regions of parameter space where the surrogate is most uncertain. Applying this framework to 38 stars in DEB systems, we recover ages more precise than previous isochrone-based determinations and obtain bounds on $α_{\rm MLT}$ for a subset of lower-mass stars ($M \lesssim 1.5 M_\odot$), where convective envelopes provide sensitivity to the mixing length parameter. For several stars, the inferred $α_{\rm MLT}$ values lie below the Solar-calibrated value, supporting previous indications that a universal mixing length parameter may not adequately describe convection across the main sequence. The active learning methods developed here provide a scalable route to Bayesian inference with stellar evolution models, with clear applications to additional stellar physics parameters and other precisely characterized stellar systems.

astro-ph.SR↗

Guiding Multi-Objective Genetic Programming with Description Length Improves Symbolic Regression Solutions

Symbolic regression with genetic programming (GPSR) may suffer from overfitting and structural bloat, especially when noise is present. In this paper we evaluate description length (DL) and fractional Bayes factor (FBF) criteria as principled, data-efficient alternatives to heuristics for selecting compact expressions that generalise well. We implement DL using a Fisher-information-based parameter encoding and compare it to AIC and BIC across multiple datasets, including noisy synthetic benchmarks and real-world regression problems. We study three search/selection strategies: (i) multi-objective search for accuracy and program length followed by DL/FBF selection; (ii) multi-objective search using DL directly as an objective; and (iii) single-objective optimisation with DL/FBF as the fitness. Across datasets we find that DL/FBF post-selection improves test performance compared to AIC/BIC baseline and that BIC in combination with the same function complexity penalty from DL/FBF produces similar results. In contrast, using DL/FBF directly as a fitness function in single-objective GPSR frequently induces premature convergence to overly simple models. We conclude with practical guidance for using DL/FBF as robust model-selection tools in genetic programming workflows.

cs.NE↗