arXiv subjects

Indira Ocampo

Publications and source records attributed to Indira Ocampo.

6 recordsLinked to original sources

arXiv2026

Explaining Neural Networks on the Sky: Machine Learning Interpretability for Cosmic Microwave Background Maps

We present a framework for cosmological model selection using Neural Networks (NNs) trained directly on simulated Cosmic Microwave Background (CMB) temperature and polarisation maps. We also apply SHAP (Shapley Additive exPlanations) as an interpretability approach to map the network's decision-making process. By introducing a galactic mask as a diagnostic control, we obtain attribution maps that correctly exclude this region and confirm that the classification is driven exclusively by the unmasked sky. Ultimately, these maps verify that the architecture distinguishes between $\Lambda$CDM and the non-standard feature model by evaluating the global statistical variance distributed across the valid cosmological regions. We describe the generation of Planck-like CMB maps, and the implemented hybrid architecture that combines principal component analysis and NNs, optimised for classification tasks. This work serves as a follow-up to previous analyses at the level of summary statistics and as a proof-of-concept for using machine learning and interpretability to evaluate the global statistical properties of masked CMB data. This diagnostic Machine Learning framework has the potential to enhance future detection of nontrivial inflationary signals and improve cosmological model discrimination, while also introduces the Open Science project SkyExplain, providing public access to the full pipeline for simulation, training, and interpretability of CMB map-based neural networks.

astro-ph.CO↗

arXiv2026

Forecast constraints on null tests of the $Λ$CDM model with SPHEREx

In this work we quantify the ability of the upcoming SPHEREx survey to constrain cosmological observables and test the internal consistency of the cosmological constant and cold dark matter ($Λ$CDM) model. Using Fisher matrix forecasting, we assess the expected precision on Baryon Acoustic Oscillations (BAO) observables, such as the angular diameter distance $D_\mathrm{A}(z)$ and the Hubble parameter $H(z)$. We further explore SPHEREx's potential to probe some of the fundamental assumptions of large-scale spatial homogeneity and isotropy, through model-independent reconstructions of several consistency tests of the $Λ$CDM model. In addition, we also examine the effect of the model dependence of the resulting Fisher and covariance matrices, using a neural network (NN) classification approach. We find that, while it is commonly assumed the covariance matrix depends weakly on the model, in fact the NN can very accurately ($\sim 98\%$) detect the underlying fiducial cosmological model based solely on the covariance matrix of the data, thus challenging this assumption. This model dependence, often neglected in standard analyses, can be naturally incorporated within simulation-based inference frameworks, which offer a flexible alternative for capturing such effects.

astro-ph.CO↗

arXiv2025

DESI constraints on two-field quintessence with exponential potentials

We investigate a quintessence model involving two scalar fields with double-exponential potentials. This configuration allows the system as a whole to emulate the dynamics of a single field with a shallower potential, enabling scalar fields that individually cannot drive cosmic acceleration to collectively achieve and sustain it. We assess the viability of this model by performing a fully Bayesian analysis and confronting its predictions with observational data, including the Planck 2018 cosmic microwave background (CMB) shift parameters, the newly released Dark Energy Spectroscopic Instrument (DESI) DR2 baryon acoustic oscillation (BAO) measurements, and the Dark Energy Survey Year 5 (DESY5) type Ia supernova (SnIa) sample. Our analysis shows that the two-field quintessence model yields a log Bayes factor relative to the flat $\Lambda$ cold dark matter model of $\Delta \ln B \sim 4$, indicating moderate evidence against the latter. We also find that the central values of the two slopes of the exponential potentials are both close to 1, whereas the slope of an effective single-field system is constrained to be less than order unity. This property is theoretically desirable from the perspective of higher-dimensional theories. Thus, the two-field quintessence model with exponential potentials provides a physically motivated and compelling mechanism that is consistent with both observational and theoretical requirements.

astro-ph.CO↗

arXiv2025

Enhancing Cosmological Model Selection with Interpretable Machine Learning

We propose a novel approach using neural networks (NNs) to differentiate between cosmological models, and implemented LIME as an interpretability approach to identify the key features influencing our model's decisions. We show the potential of NNs to enhance the extraction of meaningful information from cosmological large-scale structure data, based on current galaxy-clustering survey specifications, for the cosmological constant and cold dark matter ($Λ$CDM) model and the Hu-Sawicki $f(R)$ model. We find that the NN can successfully distinguish between $Λ$CDM and the $f(R)$ models, by predicting the correct model with approximately $97\%$ overall accuracy, thus demonstrating that NNs can maximize the potential of current and next generation surveys to probe for deviations from general relativity.

astro-ph.CO↗

arXiv2024

Non-Linearity-Free prediction of the growth-rate $fσ_8$ using Convolutional Neural Networks

The growth-rate $fσ_8(z)$ of the large-scale structure of the Universe is an important dynamic probe of gravity that can be used to test for deviations from General Relativity. However, for galaxy surveys to extract this key quantity from cosmological observations, two important assumptions have to be made: i) a fiducial cosmological model, typically taken to be the cosmological constant and cold dark matter ($Λ$CDM) model and ii) the modeling of the observed power spectrum, especially at non-linear scales, which is particularly dangerous as most models used in the literature are phenomenological at best. In this work, we propose a novel approach involving convolutional neural networks (CNNs), trained on the Quijote N-body simulations, to predict $fσ_8(z)$ directly and without assuming a model for the non-linear part of the power spectrum, thus avoiding the second of the assumptions above. This could serve as an initial step towards the future development of a method for parameter inference in Stage IV surveys. We find that the predictions for the value of $fσ_8$ from the CNN are in excellent agreement with the fiducial values since they outperform a maximum likelihood analysis and the CNN trained on the power spectrum. Therefore, we find that the CNN reconstructions provide a viable alternative to avoid the theoretical modeling of the non-linearities at small scales when extracting the growth rate.

astro-ph.CO↗

arXiv2019

Spectral index-mass accretion rate correlation and evaluation of black hole masses in AGNs 3C~454.3 and M87

We present the discovery of correlations between the X-ray spectral (photon) index and mass accretion rate observed in active galactic nuclei (AGNs) 3C~454.3 and M87. We analyzed spectral transition episodes observed in these AGNs using Chandra, Swift, Suzaku, BeppoSAX, ASCA and RXTE data. We applied a scaling technique for a black hole (BH) mass evaluation which uses a correlation between the photon index (Gamma) and normalization of the seed component which is proportional to a disk mass accretion rate Mdot. We developed an analytical model that shows that Gamma of the BH emergent spectrum undergoes an evolution from lower to higher values depending on Mdot. To estimate a BH mass in 3C~454.3 we consider extra-galactic SMBHs NGC~4051 and NGC~7469 as well as Galactic BHs Cygnus X--1 and GRO~J1550--564 as reference sources for which distances, inclination angles are known and the BH masses are already evaluated. For M87 on the other hand, we provide the BH mass scaling using extra-galactic sources (IMBHs: ESO 243-49 HLX 1 and M 101 ULX--1) and Galactic sources (stellar mass BHs: XTE J1550-564, 4U 1630-472, GRS 1915+105 and H 1743-322) as reference sources. Application of the scaling technique for the photon index-Mdot correlation provides estimates of the BH masses in 3C 454.3 and M87 to be about 3.4x10^9 and 5.6 x10^7 solar masses, respectively. We also compared our scaling BH mass estimates with a recent BH mass estimate of M_{87}=6.5x 10^9 M_{\odot} made using the {Event Horizon Telescope} which gives an image at 1.3 mm and is based on the angular size of the `BH event horizon'. Our BH mass estimate in M87 is at least two orders of magnitude lower than that made by the EHT team.

astro-ph.GA↗