arXiv subjects

Sayan Mandal

Publications and source records attributed to Sayan Mandal.

At least 19 recordsLinked to original sources

arXiv2026

Biomazon: A Multimodal Dataset for 3D Forest Structure and Biomass Modeling in the Amazon Basin

Accurate, spatially explicit characterization of tropical forest structure is essential for carbon accounting and ecosystem monitoring, yet most ML pipelines predict canopy-top height proxies (e.g., RH95/RH98) or AGBD as separate scalar targets, rather than learning the forest vertical structure as an ordered profile. The community lacks a ML-ready multimodal benchmark for predicting the entire GEDI RH profile jointly with AGBD, or for evaluating methods that enforce physically consistent ordering across RH percentiles. We address this with Biomazon, a 20 m multimodal benchmark dataset over the Amazon Basin that pairs GEDI RH and AGBD targets with multi-sensor predictors (Sentinel-1/2, ALOS-2 PALSAR-2, Copernicus DEM, Dynamic World LULC, and AlphaEarth embeddings) under standardized spatial splits and evaluation protocols. Using a shared encoder-decoder with task-specific heads as a baseline framework, we conduct a comprehensive ablation study of (i) backbone/model scale, (ii) modality contributions, and (iii) the use of auxiliary embeddings under standalone and fusion settings, and we report both single-target and joint-target results to quantify tradeoffs under a unified training protocol. Finally, we contextualize baseline performance through regionally aligned comparisons against existing gridded products, including GEDI L4D RH10-RH98 and AGBD, at matching temporal scale. Biomazon, together with the accompanying protocols and baseline results, establishes a reference benchmark for future work on structurally consistent RH-profile prediction and structure-biomass modeling in tropical forests.

cs.CV↗

arXiv2026

Imprints of primordial magnetic fields in gravitational collapse during early structure formation

Context. Primordial magnetic fields (PMFs) generated in the early Universe might have left observable imprints on present-day large-scale structure. However, the spatial scales on which primordial signatures are able to survive the nonlinear processes that accompany structure formation remain unclear. Aims. The aim of this study is to investigate the evolution of the statistical properties of PMFs during the onset of gravitational collapse. Methods. We performed a suite of high-resolution direct numerical simulations of isothermal self-gravitating, magnetized gas clouds. By varying the viscosity, we probed different Reynolds-number regimes and follow the coupled evolution of gravitational collapse and magnetohydrodynamic turbulence. Results. At sufficiently high Reynolds numbers, turbulence generated during collapse triggers the onset of a small-scale dynamo, which amplifies magnetic energy below the Jeans scale and modifies the magnetic energy spectrum significantly. The question of whether dynamo amplification dominates the magnetic field evolution is determined by the competition between the dynamo growth time and the free-fall time. Conclusions. Our results highlight the importance of resolving the Jeans scale and the associated turbulent inertial range in cosmological magnetohydrodynamic (MHD) simulations to accurately capture the interplay between gravitational compression and dynamo amplification and to assess which structures retain memory of primordial fields.

astro-ph.CO↗

arXiv2026

AEG: A Baremetal Framework for AI Acceleration via Direct Hardware Access in Heterogeneous Accelerators

This paper introduces a unified, hardware-independent baremetal runtime architecture designed to enable high-performance machine learning (ML) inference on heterogeneous accelerators, such as AI Engine (AIE) arrays, without the overhead of an underlying real-time or general-purpose operating system. Existing edge-deployment frameworks, such as TinyML, often rely on real-time operating systems (RTOS), which introduce unnecessary complexity and performance bottlenecks. To address this, our solution fundamentally decouples the runtime from hardware specifics by flattening complex control logic into linear, executable Runtime Control Blocks (RCBs). This "Control as Data" paradigm allows high-level models, including Adaptive Data Flow (ADF) graphs, to be executed by a generic engine through a minimal Runtime Hardware Abstraction Layer (RHAL). We further integrate Runtime Platform Management (RTPM) to handle system-level orchestration (including a lightweight network stack) and a Runtime In-Memory File System (RIMFS) to manage data in OS-free environments. We demonstrate the framework's efficacy with a ResNet-18 image classification implementation. Experimental results show 9.2$\times$ higher compute efficiency (throughput per AIE tile) compared to Linux-based Vitis AI deployment, 3--7$\times$ reduction in data movement overhead, and near-zero latency variance (CV~$=0.03\%$). The system achieves 68.78\% Top-1 accuracy on ImageNet using only 28 AIE tiles compared to Vitis AI's 304 tiles, validating both the efficiency and correctness of this unified bare-metal architecture.

cs.DC↗

arXiv2025

SAM2LoRA: Composite Loss-Guided, Parameter-Efficient Finetuning of SAM2 for Retinal Fundus Segmentation

We propose SAM2LoRA, a parameter-efficient fine-tuning strategy that adapts the Segment Anything Model 2 (SAM2) for fundus image segmentation. SAM2 employs a masked autoencoder-pretrained Hierarchical Vision Transformer for multi-scale feature decoding, enabling rapid inference in low-resource settings; however, fine-tuning remains challenging. To address this, SAM2LoRA integrates a low-rank adapter into both the image encoder and mask decoder, requiring fewer than 5\% of the original trainable parameters. Our analysis indicates that for cross-dataset fundus segmentation tasks, a composite loss function combining segmentationBCE, SoftDice, and FocalTversky losses is essential for optimal network tuning. Evaluated on 11 challenging fundus segmentation datasets, SAM2LoRA demonstrates high performance in both blood vessel and optic disc segmentation under cross-dataset training conditions. It achieves Dice scores of up to 0.86 and 0.93 for blood vessel and optic disc segmentation, respectively, and AUC values of up to 0.98 and 0.99, achieving state-of-the-art performance while substantially reducing training overhead.

cs.CV↗

arXiv2025

Grounded AI for Code Review: Resource-Efficient Large-Model Serving in Enterprise Pipelines

Automated code review adoption lags in compliance-heavy settings, where static analyzers produce high-volume, low-rationale outputs, and naive LLM use risks hallucination and incurring cost overhead. We present a production system for grounded, PR-native review that pairs static-analysis findings with AST-guided context extraction and a single-GPU, on-demand serving stack (quantized open-weight model, multi-tier caching) to deliver concise explanations and remediation guidance. Evaluated on safety-oriented C/C++ standards, the approach achieves sub-minute median first-feedback (offline p50 build+LLM 59.8s) while maintaining competitive violation reduction and lower violation rates versus larger proprietary models. The architecture is decoupled: teams can adopt the grounding/prompting layer or the serving layer independently. A small internal survey (n=8) provides directional signals of reduced triage effort and moderate perceived grounding, with participants reporting fewer human review iterations. We outline operational lessons and limitations, emphasizing reproducibility, auditability, and pathways to broader standards and assisted patching.

cs.SE↗

arXiv2025

Forward cascade of large-scale primordial magnetic fields during structure formation

The origin of large scale magnetic fields in the Universe is widely thought to be from early Universe processes, like inflation or phase transitions. These magnetic fields evolve via magnetohydrodynamic processes until the epoch of recombination. When structures begin to form in the later Universe, the conservation of magnetic flux amplifies the magnetic fields via the adiabatic collapse of gravitationally bound gas clouds hosting the magnetic fields and moves them to smaller scales. In this work, we have semi-analytically studied this forward cascade effect, considering simple models of gravitational collapse of structures. We find that this simple model is able to reproduce the general qualitative features of the evolution of the magnetic field spectrum as seen from magnetized cosmological simulations.

astro-ph.CO↗

arXiv2024

Deep Learning to Predict Glaucoma Progression using Structural Changes in the Eye

Glaucoma is a chronic eye disease characterized by optic neuropathy, leading to irreversible vision loss. It progresses gradually, often remaining undiagnosed until advanced stages. Early detection is crucial to monitor atrophy and develop treatment strategies to prevent further vision impairment. Data-centric methods have enabled computer-aided algorithms for precise glaucoma diagnosis. In this study, we use deep learning models to identify complex disease traits and progression criteria, detecting subtle changes indicative of glaucoma. We explore the structure-function relationship in glaucoma progression and predict functional impairment from structural eye deterioration. We analyze statistical and machine learning methods, including deep learning techniques with optical coherence tomography (OCT) scans for accurate progression prediction. Addressing challenges like age variability, data imbalances, and noisy labels, we develop novel semi-supervised time-series algorithms: 1. Weakly-Supervised Time-Series Learning: We create a CNN-LSTM model to encode spatiotemporal features from OCT scans. This approach uses age-related progression and positive-unlabeled data to establish robust pseudo-progression criteria, bypassing gold-standard labels. 2. Semi-Supervised Time-Series Learning: Using labels from Guided Progression Analysis (GPA) in a contrastive learning scheme, the CNN-LSTM architecture learns from potentially mislabeled data to improve prediction accuracy. Our methods outperform conventional and state-of-the-art techniques.

cs.CV↗

arXiv2023

Observational Constraints on Dynamical Dark Energy Models

$\phi$CDM models provide an alternative to the standard $\Lambda$CDM paradigm, while being physically better motivated. These models lead to a time-dependent speed of sound for dark energy that is difficult to replicate by $w$CDM parametrizations. We review the most up-to-date status of observational evidence for the $\phi$CDM models in this paper. We start with an overview of the motivation behind these classes of models, the basic mathematical formalism, and the different classes of models. We then present a compilation of recent results of applying different observational probes to constraining $\phi$CDM model parameters. Over the last twenty years, the precision of observational data has increased immensely, leading to ever tighter constraints. A combination of the recent measurements favors the spatially flat $\Lambda$CDM model, but a large class of $\phi$CDM models is still not ruled out.

astro-ph.CO↗

arXiv2022

Mass-varying Dark Matter from a Phase Transition

We propose a mass-varying dark matter (MVDM) model consisting of a scalar field and a fermionic field interacting via a simple Yukawa coupling, and containing an exponential self-interaction potential for the scalar field. Analyzing the evolution of this coupled scalar-fermion system in an expanding Universe, we find that it initially behaves like radiation but then undergoes a phase transition after which it behaves like pressureless dark matter. The one free parameter of this model is the temperature at which the phase transition occurs; the mass of the dark matter particle, given by the mass of the fermion, is derived from this. For a phase transition temperature between 10 MeV and $10^7$ GeV, the current dark matter relic density is achieved for a fermion mass in the range of 1 GeV to $10^9$ GeV. In this dark matter model, the scalar becomes a sub-dominant unclustered component of dark matter that can lower the amplitude of structure formation by up to a few percent. Another feature is that the mass-varying fermion component can lead to discrepant measurements of the current dark matter density of about ten percent inferred from early and late-time probes assuming LCDM.

hep-ph↗

arXiv2022

Report of the Topical Group on Dark Energy and Cosmic Acceleration: Complementarity of Probes and New Facilities for Snowmass 2021

The mechanism(s) driving the early- and late-time accelerated expansion of the Universe represent one of the most compelling mysteries in fundamental physics today. The path to understanding the causes of early- and late-time acceleration depends on fully leveraging ongoing surveys, developing and demonstrating new technologies, and constructing and operating new instruments. This report presents a multi-faceted vision for the cosmic survey program in the 2030s and beyond that derives from these considerations. Cosmic surveys address a wide range of fundamental physics questions, and are thus a unique and powerful component of the HEP experimental portfolio.

astro-ph.CO↗

arXiv2022

Snowmass2021 CMB-HD White Paper

CMB-HD is a proposed millimeter-wave survey over half the sky that would be ultra-deep (0.5 uK-arcmin) and have unprecedented resolution (15 arcseconds at 150 GHz). Such a survey would answer many outstanding questions about the fundamental physics of the Universe. Major advances would be 1.) the use of gravitational lensing of the primordial microwave background to map the distribution of matter on small scales (k~10 h Mpc^(-1)), which probes dark matter particle properties. It will also allow 2.) measurements of the thermal and kinetic Sunyaev-Zel'dovich effects on small scales to map the gas density and velocity, another probe of cosmic structure. In addition, CMB-HD would allow us to cross critical thresholds: 3.) ruling out or detecting any new, light (< 0.1 eV) particles that were in thermal equilibrium with known particles in the early Universe, 4.) testing a wide class of multi-field models that could explain an epoch of inflation in the early Universe, and 5.) ruling out or detecting inflationary magnetic fields. CMB-HD would also provide world-leading constraints on 6.) axion-like particles, 7.) cosmic birefringence, 8.) the sum of the neutrino masses, and 9.) the dark energy equation of state. The CMB-HD survey would be delivered in 7.5 years of observing 20,000 square degrees of sky, using two new 30-meter-class off-axis crossed Dragone telescopes to be located at Cerro Toco in the Atacama Desert. Each telescope would field 800,000 detectors (200,000 pixels), for a total of 1.6 million detectors.

astro-ph.CO↗

arXiv2022

Finding Evidence for Inflation and the Origin of Galactic Magnetic Fields with CMB Surveys

The origin of the $\mu$G magnetic fields observed in galaxies is unknown. One promising scenario is that magnetic fields generated during inflation, larger than 0.1 nG on Mpc scales, were adiabatically compressed to $\mu$G strengths in galaxies during structure formation. Thus, detecting a scale-invariant primordial magnetic field (PMF) above 0.1 nG on Mpc scales just after recombination would indicate an inflationary origin of galactic magnetic fields. This would also provide compelling evidence that inflation occurred since only an inflationary mechanism could generate such a strong, scale-invariant magnetic field on Mpc scales. In contrast, constraining the scale-invariant PMF strength to be below 0.1 nG would imply an inflationary scenario is not the primary origin, since such weak PMFs cannot be amplified enough via adiabatic compression to produce the strength of the galactic fields we observe today. We find that measurements of anisotropic birefringence by future CMB surveys will be able to improve the sensitivity to Mpc-scale inflationary PMFs by an order of magnitude, and, in particular, that CMB-HD would lower the upper bound to 0.072 nG at the 95% C.L., which is below the critical 0.1 nG threshold for ruling out a purely inflationary origin. If inflationary PMFs exist, we find that a CMB-HD survey would be able to detect them with about $3\sigma$ significance or higher, providing evidence for inflation itself.

astro-ph.CO↗

arXiv2021

Assessing glaucoma in retinal fundus photographs using Deep Feature Consistent Variational Autoencoders

One of the leading causes of blindness is glaucoma, which is challenging to detect since it remains asymptomatic until the symptoms are severe. Thus, diagnosis is usually possible until the markers are easy to identify, i.e., the damage has already occurred. Early identification of glaucoma is generally made based on functional, structural, and clinical assessments. However, due to the nature of the disease, researchers still debate which markers qualify as a consistent glaucoma metric. Deep learning methods have partially solved this dilemma by bypassing the marker identification stage and analyzing high-level information directly to classify the data. Although favorable, these methods make expert analysis difficult as they provide no insight into the model discrimination process. In this paper, we overcome this using deep generative networks, a deep learning model that learns complicated, high-dimensional probability distributions. We train a Deep Feature consistent Variational Autoencoder (DFC-VAE) to reconstruct optic disc images. We show that a small-sized latent space obtained from the DFC-VAE can learn the high-dimensional glaucoma data distribution and provide discriminatory evidence between normal and glaucoma eyes. Latent representations of size as low as 128 from our model got a 0.885 area under the receiver operating characteristic curve when trained with Support Vector Classifier.

eess.IV↗

arXiv2021

Polarization of gravitational waves from helical MHD turbulent sources

We use direct numerical simulations of decaying primordial hydromagnetic turbulence with helicity to compute the resulting gravitational wave (GW) production and its degree of circular polarization. We find a clear dependence of the polarization of the resulting GWs on the fractional helicity of the turbulent source and we show that driven magnetic fields produce GWs more efficiently than magnetic fields that are initially present, leading to larger spectral amplitudes. The helicity does not have a huge impact on the maximum spectral amplitude in any of the two types of turbulence considered. However, the GW spectrum at wave numbers away from the peak becomes smaller for larger values of the magnetic fractional helicity. The degree of circular polarization approaches zero at frequencies below the peak, and reaches its maximum at the peak. At higher frequencies, it stays finite if the magnetic field is initially present, and it approaches zero if it is driven. We predict that the spectral peak of the GW signal can be detected by LISA if the turbulent energy density is at least $\sim 3\%$ of the radiation energy density, and the characteristic scale is a hundredth of the horizon at the electroweak scale. We show that the resulting GW polarization is unlikely to be detectable by the anisotropies induced by our proper motion in the dipole response function of LISA. Such signals can, however, be detectable by cross-correlating data from the LISA-Taiji network for turbulent energy densities of $\sim 5\%$, and fractional helicity of 0.5 to 1. Second-generation space-based GW detectors, such as BBO and DECIGO, would allow for the detection of a larger range of the GW spectrum and smaller amplitudes of the magnetic field.

gr-qc↗

arXiv2021

The scalar, vector, and tensor modes in gravitational wave turbulence simulations

We study the gravitational wave (GW) signal sourced by primordial turbulence that is assumed to be present at cosmological phase transitions like the electroweak and quantum chromodynamics phase transitions. We consider various models of primordial turbulence, such as those with and without helicity, purely hydrodynamical turbulence induced by fluid motions, and magnetohydrodynamic turbulence whose energy can be dominated either by kinetic or magnetic energy, depending on the nature of the turbulence. We also study circularly polarized GWs generated by parity violating sources such as helical turbulence. Our ultimate goal is to determine the efficiency of GW production through different classes of turbulence. We find that the GW energy and strain tend to be large for acoustic or irrotational turbulence, even though its tensor mode amplitude is relatively small at most wave numbers. Only at very small wave numbers is the spectral tensor mode significant, which might explain the efficient GW production in that case.

gr-qc↗

arXiv2021

Circular Polarization of Gravitational Waves from Early-Universe Helical Turbulence

We perform direct numerical simulations to compute the net circular polarization of gravitational waves from helical (chiral) turbulent sources in the early Universe for a variety of initial conditions, including driven (stationary) and decaying turbulence. We investigate the resulting gravitational wave signal assuming different turbulent geneses such as magnetically or kinetically driven cases. Under realistic physical conditions in the early Universe we compute numerically the wave number-dependent polarization degree of the gravitational waves. We find that the spectral polarization degree strongly depends on the initial conditions. The peak of the spectral polarization degree occurs at twice the typical wavenumber of the source, as expected, and for fully helical decaying turbulence, it reaches its maximum of nearly 100\% {\it only} at the peak. We determine the temporal evolution of the turbulent sources as well as the resulting gravitational waves, showing that the dominant contribution to their spectral energy density happens shortly after the activation of the source. Only through an artificially prolonged decay of the turbulence can further increase of the gravitational wave amplitude be achieved. We estimate the detection prospects for the net polarization, arguing that its detection contains {\it clean} information (including the generation mechanisms, time, and strength) about the sources of possible parity violations in the early Universe.

astro-ph.CO↗

arXiv2020

End-to-End Bengali Speech Recognition

Bengali is a prominent language of the Indian subcontinent. However, while many state-of-the-art acoustic models exist for prominent languages spoken in the region, research and resources for Bengali are few and far between. In this work, we apply CTC based CNN-RNN networks, a prominent deep learning based end-to-end automatic speech recognition technique, to the Bengali ASR task. We also propose and evaluate the applicability and efficacy of small 7x3 and 3x3 convolution kernels which are prominently used in the computer vision domain primarily because of their FLOPs and parameter efficient nature. We propose two CNN blocks, 2-layer Block A and 4-layer Block B, with the first layer comprising of 7x3 kernel and the subsequent layers comprising solely of 3x3 kernels. Using the publicly available Large Bengali ASR Training data set, we benchmark and evaluate the performance of seven deep neural network configurations of varying complexities and depth on the Bengali ASR task. Our best model, with Block B, has a WER of 13.67, having an absolute reduction of 1.39% over comparable model with larger convolution kernels of size 41x11 and 21x11.

eess.AS↗

arXiv2020

Primordial magnetic helicity evolution with a homogeneous magnetic field from inflation

Motivated by a scenario of magnetogenesis in which a homogeneous magnetic field is generated during inflation, we study the magnetohydrodynamic evolution of the primordial plasma motions for two kinds of initial conditions -- (i) a spatially homogeneous field with an unlimited correlation length, and (ii) a zero flux scale-invariant statistically homogeneous magnetic field. In both cases, we apply, for a short initial time interval, monochromatic forcing at a certain wave number so that the correlation length is finite, but much smaller than the typical length scale of turbulence. In particular, we investigate the decay of nonhelical and helical hydromagnetic turbulence. We show that, in the presence of a homogeneous magnetic field, the decay of helical and nonhelical small-scale fields can occur rapidly. This is a special property of a system with a perfectly homogeneous magnetic field, which is sometimes considered as a local approximation to a slowly varying background field. It can never change and acts as an imposed magnetic field. This is in a sharp contrast to the case of a statistically homogeneous magnetic field, where we recover familiar decay properties: a much slower decay of magnetic energy and a faster growth of the correlation length, especially in the case with magnetic felicity. The result suggests that a homogeneous magnetic field, if generated during inflation, should persist under the influence of small-scale fields and could be the origin of the large-scale magnetic field in the Universe.

astro-ph.CO↗