arXiv subjects

Krešimir Tisanić

Publications and source records attributed to Krešimir Tisanić.

9 recordsLinked to original sources

arXiv2026

Partial-field inferability of galaxy overdensity structure: An information-geometric account of incomplete sky coverage

Structural statistics of the galaxy distribution are routinely measured on only partially observed fields, yet how much structure survives incomplete coverage, and which aspects are lost first, is rarely quantified independently of the statistic being reported. We cast this as inference on a fixed structure vector Theta=(Delta_Omega, f_HD, L_HD): a global density contrast, a high-density fraction, and the connected extent of the high-density set. Densities are assigned to galaxies through their Voronoi cell areas, with connectivity on the dual Delaunay graph. Two mechanisms generate subsets, spatially correlated random-walk coverage and spatially uncorrelated node loss, and recovery is measured with a Fisher-inspired reference metric held fixed across coverage levels, mechanisms and estimators. Applied to a z-sliced COSMOS field of 94043 galaxies, the three components form a strict hierarchy: the connected extent is recovered several times less accurately than the scalar amplitudes at every incomplete coverage, by a margin that widens as coverage improves, and is the last to converge. The two mechanisms trade off in opposite directions, uniform loss preserving the amplitudes and walk coverage the morphology, so inferability depends not only on how much of the field is observed but on how it is observed. A joint single-number criterion can order the mechanisms but is tolerance dependent and can reverse, whereas the component-level contrast is metric-independent and robust; the degree-based inverse-probability weight, though natural for random-walk sampling, is inconsistent for set-valued coverage. Partial-field structure statistics therefore degrade anisotropically in information space, and a single coverage fraction is insufficient to characterise what a partially observed field can support.

astro-ph.IM↗

arXiv2026

Information Geometry meets Functional ANOVA: An Exact Fisher-Information Decomposition, with an Application to Radio Luminosity Functions

Information geometry represents a fitted model as a manifold whose metric is the Fisher information. We show that, for a nonlinear regression model, this metric decomposes exactly across the Hoeffding-Sobol' (functional ANOVA) channels of its covariates. Treating the score contributions of main effects and interactions as vectors in a Hilbert space, their Gram matrix reproduces the (Gauss-Newton) Fisher information metric exactly, and this identity extends to an exact, additive decomposition of the Fisher information matrix into main-effect, interaction, and cross-channel terms: a Sobol'-type accounting for parameter information rather than output variance, and equivalently a channel-resolved pullback of the ambient Fisher metric along the model embedding. We give the exact decomposition theorem with a short proof (deriving both the centered and uncentered Fisher information), define information-theoretic analogues of Sobol' indices, characterize the sign and finite-sample behaviour of the cross-channel terms, and describe two practical estimators. As an astrophysical application we decompose the parameter information of a combined 1.4 GHz radio luminosity-function model spanning active galactic nuclei (AGN) and star-forming galaxies (SFG), fit to 6dFGS-NVSS and VLA-COSMOS 3 GHz data: the decomposition localizes where in luminosity each population and survey constrains the fit, and exposes the compensating ("sloppy") channel combinations near the luminosity-function turnover. A plant CO2-uptake experiment serves as a small, fully categorical companion example.

astro-ph.IM↗

arXiv2026

Graph reconstruction from random-walk co-visitation: Geometric, empirical, and controlled networks

Reconstructing an unknown graph from the trajectory of a random walk arises both for spatial correlation networks in astrophysics and for connectivity inference in network science. We present a reconstruction pipeline whose observable is the random-walk co-visitation matrix, whose model is a pairwise edge-weight basis, and whose fitter is a frame-balanced Levenberg-Marquardt (fbLM) scheme with per-node group weights and a self-calibrated edge readout. Unlike the marginal occupation, the co-visitation retains the ordered pair before the row sum is taken, and the pairwise basis can represent structure that an additive node-potential model cannot; neither change suffices alone. We apply the pipeline to an email communication subgraph, to Delaunay and Voronoi networks built from a COSMOS sky catalogue, and to two controlled 12-vertex test graphs, one unicyclic and one a tree, under both analytic-noise and finite-walk regimes. Reconstructions are scored against the ground-truth adjacency, which enters nowhere in the fit, by true/false positives and the Matthews correlation coefficient (MCC). All test-beds are reconstructed with high fidelity at full graph size: on finite-walk data we recover the COSMOS Delaunay and Voronoi graphs at MCC above 0.98 up to their full extent, N=119 and N=223, the whole graph rather than a cut-out of it, and the empirical email-Eu-core graph at N=240 (417 edges). On the full Delaunay graph a graphical-lasso reference returns MCC 0.540 against 0.988 for fbLM. Each reconstructed edge carries a Fisher-propagated uncertainty, and the residual misses are almost entirely confined to edges the walk never traverses. In the finite-walk regime the limiting factor is therefore walk coverage rather than the fit: essentially every edge the walk visits is recovered, so reconstructibility is governed by the sampling of the graph rather than by the estimator.

astro-ph.IM↗

arXiv2026

Gram-Schmidt correlation bookkeeping for weighted FBET-type frameworks: separating short- and long-range correlation scales in correlated measurements, with an application to radio luminosity functions

Uncertainty quantification for correlated measurements frequently requires a covariance model that mixes several qualitatively different correlation structures: local measurement noise, short-range (e.g. cyclic or seasonal) correlations, and long-range correlations associated with trend, drift, or long-memory behavior. Weighted and generalized least squares can encode such structure only when the covariance matrix is known or credibly specified in advance, which in practice it rarely is. We describe a structured correlation bookkeeping construction intended for use inside a weighted FBET-type (wFBET) uncertainty-quantification framework: candidate correlation basis functions associated with distinct lag variables are orthogonalized with a weighted Gram-Schmidt procedure (under an explicitly declared, statistically motivated metric) before they are allowed to enter the covariance or metric construction. This avoids the double counting inherent in naive additive covariance decompositions of the form $Σ= Σ_{\mathrm{short}} + Σ_{\mathrm{long}}$. We benchmark the construction (in its vectorized form, its operator-valued kernel form, and the finite-rank basis-function realization used by the implementation) on the classic AirPassengers benchmark dataset, and we apply it to the astrophysical case of radio luminosity functions measured in redshift bins. No claim of universal optimality is made; the goal is a transparent and auditable way to organize correlation scales within weighted information-geometric fitting frameworks.

astro-ph.IM↗

arXiv2026

Inferring Unreported Measurement Uncertainties via Information Geometry in Astrophysics

Modern radio and multi-instrument astrophysical datasets are increasingly assembled from surveys with different sensitivities and selection effects. In such heterogeneous datasets, published measurement uncertainties are often incomplete, non-uniform across subsets, or missing cross-correlation information altogether. This limits reliable statistical inference, since underestimated or inconsistently modeled uncertainties can distort fitted spectral shapes, bias parameter estimates, and obscure physically meaningful structure. We introduce the Fisher Information Metric Error Reconstruction (FIMER), an information-geometric framework for reconstructing effective measurement uncertainties directly from heterogeneous astrophysical data. FIMER combines weighted Fisher-information geometry, FBET and an adaptive discrete hyperparameter search, while incorporating prior statistical knowledge of detector behavior into the weighting procedure. The priors used are not chosen as arbitrary tuning prescriptions or uninformative regularizers; they are motivated by statistical properties of the underlying detection process. Poisson priors represent counting-statistics behavior, while extreme-value priors allow tail-dominated fluctuations to be incorporated when rare or asymmetric excursions are expected to influence the inferred uncertainty distribution. We apply FIMER to radio SEDs of RxAGN using COSMOS VLA data at 1.4 and 3 GHz together with GMRT data at 325 and 610 MHz. The results show that FIMER provides a practical route to uncertainty reconstruction in heterogeneous survey combinations, especially when reported uncertainties are unavailable, underestimated, or strongly correlated. The method is particularly relevant for archival and multi-survey astrophysical datasets, where full covariance information is rarely available but reliable statistical inference remains essential.

astro-ph.IM↗

arXiv2023

Detecting long-period variability in SDSS Stripe 82 standards catalog

We report the results of a search for long-period ($100<P<600$ days) periodic variability in SDSS Stripe 82 standards catalog. The SDSS coverage of Stripe 82 enables such a search because there are on average 20 observations per band in $ugriz$ bands for about 1 million sources, collected over about 6 years, with a faint limit of $r\sim22$ mag and precisely calibrated 1-2% photometry. We calculated the periods of candidate variable sources in this sample using the Lomb-Scargle periodogram and considered the three highest periodogram peaks in each of the $gri$ filters as relevant. Only those sources with $gri$ periods consistent within 0.1% were later studied. We use the Kuiper statistic to ensure uniform distribution of data points in phased light curves. We present 5 sources with the spectra consistent with quasar spectra and plausible periodic variability. This SDSS-based search bodes well for future sensitive large-area surveys, such as the Rubin Observatory Legacy Survey of Space and Time, which, due to its larger sky coverage (about a factor of 60) and improved sensitivity ($\sim2$ mag), will be more powerful for finding such sources.

astro-ph.GA↗

arXiv2021

The XXL Survey: XLIII. The quasar radio loudness dichotomy exposed via radio luminosity functions obtained by combining results from COSMOS and XXL-S X-ray selected quasars

We studied a sample of 274 radio and X-ray selected quasars (XQSOs) detected in the COSMOS and XXL-S radio surveys at 3 GHz and 2.1 GHz, respectively. This sample was identified by adopting a conservative threshold in X-ray luminosity, Lx [2-10\ keV] >= 10^44 erg/s, selecting only the most powerful quasars. Using available multiwavelength data, we examined various criteria for the selection of radio-loud (RL) and radio-quiet (RQ) XQSOs, finding that the number of RL/RQ XQSOs changes significantly depending on the chosen criterion. This discrepancy arises due to the different criteria tracing different physical processes and due to our sample being selected from flux-limited radio and X-ray surveys. Another approach to study the origin of radio emission in XQSOs is via their radio luminosity function (RLF). We constructed the XQSO 1.4 GHz radio luminosity functions (RLFs) in six redshift bins at 0.5 <= z <= 3.7. The lower-1.4 GHz luminosity end shows a higher normalization than expected only from AGN contribution in all studied redshift bins. The found "bump" is mostly dominated by emission due to star-forming (SF) processes within the XQSO host galaxies. As expected, AGN-related radio emission dominates at the higher-luminosity end of RLF. The evolution of XQSO RLF was studied via combination of analytic forms from the literature to constrain the lower-luminosity "bump" and the higher-luminosity AGN part of the RLF. We defined two 1.4 GHz luminosity thresholds, L_th,SF and L_th,AGN, below and above which more than 80% of sources contributing to the RLF are dominated by SF and AGN-related activity, respectively. These thresholds evolve with redshift, most likely due to the strong evolution of SFRs of the XQSO host galaxies.

astro-ph.GA↗

arXiv2020

The VLA-COSMOS 3 GHz Large Project: Average radio spectral energy distribution of active galactic nuclei

As the SKA is expected to be operational in the next decade, investigations of the radio sky in the range of 100 MHz to 10 GHz have become important for simulations of the SKA observations. In determining physical properties of galaxies from radio data, the radio SED is often assumed to be described by a simple power law, usually with a spectral index of 0.7 for all sources. Even though radio SEDs have been shown to exhibit deviations from this assumption, both in differing spectral indices and complex spectral shapes, it is often presumed that their individual differences cancel out in large samples. We constructed the average radio SED of radio-excess active galactic nuclei (RxAGN), defined as those that exhibit a 3 $σ$ radio luminosity excess with respect to the value expected only from contribution from star formation, out to z~4. We combined VLA observations of the COSMOS field at 1.4 GHz and 3 GHz with GMRT observations at 325 MHz and 610 MHz. To account for nondetections in the GMRT maps, we employed the survival analysis technique. We selected a sample of RxAGN out to z~4. We find that a sample of RxAGN can be described by a spectral index of $α_1=0.28\pm0.03$ below the break frequency $ν_b=(4.1\pm0.2)$ GHz and $α_2=1.16\pm0.04$ above, while a simple power-law model yields a single spectral index of $α=0.64\pm0.07$. By binning in 1.4 GHz radio luminosity and redshift, we find that the power-law spectral index, as well as broken power-law spectral indices, may increase for larger source sizes, while the power-law spectral index and lower-frequency (<4 GHz) broken power-law spectral index are additionally positively correlated with redshift.

astro-ph.GA↗

arXiv2018

Average radio spectral energy distribution of highly star-forming galaxies

The infrared-radio correlation (IRRC) offers a way to assess star formation from radio emission. Multiple studies found the IRRC to decrease with increasing redshift. This may in part be due to the lack of knowledge about the possible radio spectral energy distributions (SEDs) of star-forming galaxies. We constrain the radio SED of a complete sample of highly star-forming galaxies ($SFR>100\,\mathrm{M_{\odot}}/\,\mathrm{yr}$) based on the VLA-COSMOS $1.4\,\mathrm{GHz}$ Joint and $3\,\mathrm{GHz}$ Large Project catalogs. We reduce archival GMRT $325\,\mathrm{MHz}$ and $610\,\mathrm{MHz}$ observations, broadening the rest-frame frequency range to $0.3-15\,\mathrm{GHz}$. Employing survival analysis and fitting a double power law SED, we find that the slope steepens from a spectral index of $α_1=0.51\pm 0.04$ below $4.5\,\mathrm{GHz}$ to $α_2=0.98\pm0.07$ above $4.5\,\mathrm{GHz}.$ Our results suggest that the use of a K-correction assuming a single power-law radio SED for star forming galaxies is likely not the root cause of the IRRC trend.

astro-ph.GA↗