arXiv subjects

Mahdi Naseri

Publications and source records attributed to Mahdi Naseri.

7 recordsLinked to original sources

arXiv2026

SHAMISA: SHAped Modeling of Implicit Structural Associations for Self-supervised No-Reference Image Quality Assessment

No-Reference Image Quality Assessment (NR-IQA) aims to estimate perceptual quality without access to a reference image of pristine quality. Learning an NR-IQA model faces a fundamental bottleneck: its need for a large number of costly human perceptual labels. We propose SHAMISA, a non-contrastive self-supervised framework that learns from unlabeled distorted images by leveraging explicitly structured relational supervision. Unlike prior methods that impose rigid, binary similarity constraints, SHAMISA introduces implicit structural associations, defined as soft, controllable relations that are both distortion-aware and content-sensitive, inferred from synthetic metadata and intrinsic feature structure. A key innovation is our compositional distortion engine, which generates an uncountable family of degradations from continuous parameter spaces, grouped so that only one distortion factor varies at a time. This enables fine-grained control over representational similarity during training: images with shared distortion patterns are pulled together in the embedding space, while severity variations produce structured, predictable shifts. We integrate these insights via dual-source relation graphs that encode both known degradation profiles and emergent structural affinities to guide the learning process throughout training. A convolutional encoder is trained under this supervision and then frozen for inference, with quality prediction performed by a linear regressor on its features. Extensive experiments on synthetic, authentic, and cross-dataset NR-IQA benchmarks demonstrate that SHAMISA achieves strong overall performance with improved cross-dataset generalization and robustness, all without human quality annotations or contrastive losses.

cs.CV↗

arXiv2024

Exploring pathways to forming twin stars

A viable model for the dense matter equation of state above the nuclear saturation density includes a hadron-to-quark phase transition at densities relevant to compact objects. In this case, stable hybrid hadron-quark stars can arise. An even more interesting scenario is one where the hadron-to-quark phase transition results in the emergence of a third branch of stable compact objects (in addition to white dwarfs and neutron stars). Inherent to the presence of a third family of compact stars is the existence of twin stars - hybrid stars with the same mass as the corresponding neutron stars, but with smaller radii. Interestingly, the neutron star-twin star scenario is consistent with GW170817. If twin stars exist in nature, it raises a question about the mechanism that leads to their formation. Here, we explore gravitational collapse as a pathway to the formation of low-mass twin stars. We perform fully general relativistic simulations of the collapse of a stellar iron core, modeled as a cold degenerate gas, to investigate whether the end product is a neutron star or a twin star. Our simulations show that even with unrealistically large perturbations in the initial conditions, the core bounces well below the hadron-to-quark phase transition density, if the initial total rest mass is in the twin star range. Following cooling, these configurations produce neutron stars. We find that twin stars can potentially form due to mass loss, e.g., through winds, from a slightly more massive hybrid star that was initially produced in the collapse of a more massive core or if the maximum neutron star mass is below the Chandrasekhar mass limit. The challenge in producing twin stars in gravitational collapse, in conjunction with the fine-tuning required because of their narrow mass range, suggests the rarity of twin stars in nature.

astro-ph.HE↗

arXiv2024

ExGRG: Explicitly-Generated Relation Graph for Self-Supervised Representation Learning

Self-supervised Learning (SSL) has emerged as a powerful technique in pre-training deep learning models without relying on expensive annotated labels, instead leveraging embedded signals in unlabeled data. While SSL has shown remarkable success in computer vision tasks through intuitive data augmentation, its application to graph-structured data poses challenges due to the semantic-altering and counter-intuitive nature of graph augmentations. Addressing this limitation, this paper introduces a novel non-contrastive SSL approach to Explicitly Generate a compositional Relation Graph (ExGRG) instead of relying solely on the conventional augmentation-based implicit relation graph. ExGRG offers a framework for incorporating prior domain knowledge and online extracted information into the SSL invariance objective, drawing inspiration from the Laplacian Eigenmap and Expectation-Maximization (EM). Employing an EM perspective on SSL, our E-step involves relation graph generation to identify candidates to guide the SSL invariance objective, and M-step updates the model parameters by integrating the derived relational information. Extensive experimentation on diverse node classification datasets demonstrates the superiority of our method over state-of-the-art techniques, affirming ExGRG as an effective adoption of SSL for graph representation learning.

cs.LG↗

arXiv2024

Generated Contents Enrichment

We study Generated Contents Enrichment (GCE), a conditional image-generation task in which a sparse scene description is first enriched through an explicit scene representation and then rendered into semantically richer visual content. Conventional image-generation systems can produce visually realistic outputs from limited scene descriptions, but the added content is usually implicit in the generator rather than represented as an inspectable intermediate structure. In contrast, GCE seeks to make scene enrichment explicit at the scene-representation level while examining its visual consequences during generation, with the goal of encouraging generated content that is visually plausible, structurally coherent, and semantically richer than the sparse input. To instantiate GCE, we propose a jointly trained adversarial framework that enriches scene graphs by modeling object semantics and inter-object relations. Our approach first represents the input description as a scene graph, where nodes model objects and edges capture inter-object relations. The framework uses graph convolutional networks to predict additional objects and their relations to the existing scene. Finally, the enriched scene graph is passed through the downstream image-generation pipeline to generate the corresponding visual content. We evaluate the framework with proxy scene graph enrichment metrics, image-quality comparisons, qualitative examples, and user studies on the Visual Genome dataset.

cs.CV↗

arXiv2022

Primordial Black Hole Merger Rate in Self-Interacting Dark Matter Halo Models

We study the merger rate of primordial black holes (PBHs) in self-interacting dark matter (SIDM) halo models. To explore a numerical description for the density profile of SIDM halo models, we use the result of a previously performed simulation for SIDM halo models with $σ/m=10~{\rm cm^{2}g^{-1}}$. We also propose a concentration-mass-time relation that can explain the evolution of the halo density profile related to SIDM models. Furthermore, we investigate the encounter condition of PBHs that may have been randomly distributed in the medium of dark matter halos. Under these assumptions, we calculate the merger rate of PBHs within each halo considering SIDM halo models and compare the results with that obtained for cold dark matter (CDM) halo models. To do this, we employ the definition of the time after halo virialization as a function of halo mass. We indicate that SIDM halo models for $f_{\rm PBH}>0.32$ can generate sufficient PBH mergers in such a way that those exceed the one resulted from CDM halo models. By considering the spherical-collapse halo mass function, we obtain similar results for the cumulative merger rate of PBHs. Moreover, we calculate the redshift evolution of the PBH total merger rate. To determine a constraint on the PBH abundance, we study the merger rate of PBHs in terms of their fraction and masses and compare those with the black hole merger rate estimated by the Advanced LIGO (aLIGO)-Advanced Virgo (aVirgo) detectors during the third observing run. The results demonstrate that within the context of SIDM halo models, the merger rate of $10~M_{\odot}-10~M_{\odot}$ events can potentially fall within the aLIGO-aVirgo window. We also estimate a relation between the fraction of PBHs and their masses, which is well consistent with our findings.

astro-ph.CO↗

arXiv2021

Super Interacting Dark Sector: An Improvement on Self-Interacting Dark Matter via Scaling Relations of Galaxy Clusters

Self-interacting dark matter is known as one of the most appropriate candidates for dark matter. Due to its excellent success in removing many astrophysical problems, particularly in small scale structure, studying this model has taken on added significance. In this paper, we focus on the results of two previously performed simulations of cluster sized halos with self-interacting dark matter and introduce a new function for the density profile of galaxy clusters, which can perfectly describe the result of these simulations. This density profile helps to find a velocity dispersion profile and also a relation between cluster mass and concentration parameter. Using these relations, we investigate two scaling relations of galaxy clusters, namely mass-velocity dispersion and mass-temperature relations. The scaling relations reveal that in the self-interacting dark matter model, halos are more massive than what the standard non-interacting model predicts for any fixed temperature. We also study the mass-temperature relation for a hybrid interacting model, which is a combination of self-interacting dark matter idea with another model of the dark sector in which dark matter particle mass is determined according to its interaction with dark energy. This super interacting dark sector (SIDS) model can change the mass-temperature relation to a modified form that has the same result as a non-interacting model. Finally, we provide quantitative expressions which can describe the constants of this interacting model with the value of cross-section per unit mass of dark matter particles.

astro-ph.CO↗

arXiv2020

The Effect of Interacting Dark Energy on Mass-Temperature Relation in Galaxy Clusters

There are a variety of cosmological models for dark matter and dark energy in which a possible interaction is considered between these two significant components of the universe. We focus on five suggested models of interacting dark matter and dark energy and derive the modified virial theorem for them by developing a previous approach. It provides an opportunity to study the evolution of this modified virial theorem with time and interacting constants for different interacting models. Then we use this obtained virial condition to investigate the modified mass-temperature relation in galaxy clusters via three various methods. It reveals that the effect of interaction between dark matter and dark energy merely appears in the normalization factor of $M\propto T^{\frac{3}{2}}$. This relation also leads to a new constraint on the constants of interacting models, which only depends on the concentration parameter and density profile of the cluster. Then we use five observational data sets to check some proposed figures for the constants of interaction which have been resulted from other observational constraints. Finally, by fitting the observational results to the modified mass-temperature relation, we obtain values for interacting constants of three models and four specific cases of the two remained models. In agreement with many other observational outcomes, we find that according to observational data for masses and temperatures of the galaxy clusters, energy transfer occurs from dark matter to dark energy in the seven investigated models.

astro-ph.CO↗