arXiv subjects

Ramit Dey

Publications and source records attributed to Ramit Dey.

16 recordsLinked to original sources

arXiv2026

An explicit operator explains end-to-end computation in the modern neural networks used for sequence and language modeling

We establish a mathematical correspondence between state space models, a state-of-the-art architecture for capturing long-range dependencies in data, and an exactly solvable nonlinear oscillator network. As a specific example of this general correspondence, we analyze the diagonal linear time-invariant implementation of the Structured State Space Sequence model (S4). The correspondence embeds S4D, a specific implementation of S4, into a ring network topology, in which recent inputs are encoded, as waves of activity traveling over the one-dimensional spatial layout of the network. We then derive an exact operator expression for the full forward pass of S4D, yielding an analytical characterization of its complete input-output map. This expression reveals that the nonlinear decoder in the system induces interactions between these information-carrying waves that enable classifying real-world sequences. These results generalize across modern SSM architectures, and show that they admit an exact mathematical description with a clear physical interpretation. These insights enable a new level of interpretability for these systems in terms of nonlinear oscillator networks.

cs.NE↗

arXiv2023

GRINN: A Physics-Informed Neural Network for solving hydrodynamic systems in the presence of self-gravity

Modeling self-gravitating gas flows is essential to answering many fundamental questions in astrophysics. This spans many topics including planet-forming disks, star-forming clouds, galaxy formation, and the development of large-scale structures in the Universe. However, the nonlinear interaction between gravity and fluid dynamics offers a formidable challenge to solving the resulting time-dependent partial differential equations (PDEs) in three dimensions (3D). By leveraging the universal approximation capabilities of a neural network within a mesh-free framework, physics informed neural networks (PINNs) offer a new way of addressing this challenge. We introduce the gravity-informed neural network (GRINN), a PINN-based code, to simulate 3D self-gravitating hydrodynamic systems. Here, we specifically study gravitational instability and wave propagation in an isothermal gas. Our results match a linear analytic solution to within 1\% in the linear regime and a conventional grid code solution to within 5\% as the disturbance grows into the nonlinear regime. We find that the computation time of the GRINN does not scale with the number of dimensions. This is in contrast to the scaling of the grid-based code for the hydrodynamic and self-gravity calculations as the number of dimensions is increased. Our results show that the GRINN computation time is longer than the grid code in one- and two- dimensional calculations but is an order of magnitude lesser than the grid code in 3D with similar accuracy. Physics-informed neural networks like GRINN thus show promise for advancing our ability to model 3D astrophysical flows.

cs.LG↗

arXiv2023

Spectrogram correlated stacking: A novel time-frequency domain analysis of the Stochastic Gravitational Wave Background

The astrophysical stochastic gravitational wave background (SGWB) originates from numerous faint sub-threshold gravitational wave (GW) signals arising from the coalescing binary compact objects. This background is expected to be discovered from the current (or next-generation) network of GW detectors by cross-correlating the signal between multiple pairs of GW detectors. However, detecting this signal is challenging and the correlation is only detectable at low frequencies due to the arrival time delay between different detectors. In this work, we propose a novel technique, \texttt{Spectrogram Correlated Stacking} (or \texttt{SpeCs}), which goes beyond the usual cross-correlation (and to higher frequencies) by exploiting the higher-order statistics in the time-frequency domain which accounts for the \textit{chirping} nature of the individual events that comprise SGWB. We show that \texttt{SpeCs} improves the signal-to-noise for the detection of SGWB by a factor close to $8$, compared to standard optimal cross-correlation methods which are tuned to measure only the power spectrum of the SGWB signal.\texttt{SpeCs} can probe beyond the power spectrum and its application to the GW data available from the current and next-generation GW detectors would speed up the SGWB discovery.

gr-qc↗

arXiv2023

Predicting Stellar Mass Accretion: An Optimized Echo State Network Approach in Time Series Modeling

Modeling the dynamics of the formation and evolution of protostellar disks as well as the history of stellar mass accretion typically involve the numerical solution of complex systems of coupled differential equations. The resulting mass accretion history of protostars is known to be highly episodic due to recurrent instabilities and also exhibits short timescale flickering. By leveraging the strong predictive abilities of neural networks, we extract some of the critical temporal dynamics experienced during the mass accretion including periods of instability. Particularly, we utilize a novel form of the Echo-State Neural Network (ESN), which has been shown to efficiently deal with data having inherent nonlinearity. We introduce the use of Optimized-ESN (Opt-ESN) to make model-independent time series forecasting of mass accretion rate in the evolution of protostellar disks. We apply the network to multiple hydrodynamic simulations with different initial conditions and exhibiting a variety of temporal dynamics to demonstrate the predictability of the Opt-ESN model. The model is trained on simulation data of $\sim 1-2$ Myr, and achieves predictions with a low normalized mean square error ($\sim 10^{-5}$ to $10^{-3}$) for forecasts ranging between 100 and 3800 yr. This result shows the promise of the application of machine learning based models to time-domain astronomy.

astro-ph.SR↗

arXiv2022

Using Bayesian Deep Learning to infer Planet Mass from Gaps in Protoplanetary Disks

Planet induced sub-structures, like annular gaps, observed in dust emission from protoplanetary disks provide a unique probe to characterize unseen young planets. While deep learning based model has an edge in characterizing the planet's properties over traditional methods, like customized simulations and empirical relations, it lacks in its ability to quantify the uncertainty associated with its predictions. In this paper, we introduce a Bayesian deep learning network "DPNNet-Bayesian" that can predict planet mass from disk gaps and provides uncertainties associated with the prediction. A unique feature of our approach is that it can distinguish between the uncertainty associated with the deep learning architecture and uncertainty inherent in the input data due to measurement noise. The model is trained on a data set generated from disk-planet simulations using the \textsc{fargo3d} hydrodynamics code with a newly implemented fixed grain size module and improved initial conditions. The Bayesian framework enables estimating a gauge/confidence interval over the validity of the prediction when applied to unknown observations. As a proof-of-concept, we apply DPNNet-Bayesian to dust gaps observed in HL Tau. The network predicts masses of $ 86.0 \pm 5.5 M_{\Earth} $, $ 43.8 \pm 3.3 M_{\Earth} $, and $ 92.2 \pm 5.1 M_{\Earth} $ respectively, which are comparable to other studies based on specialized simulations.

astro-ph.EP↗

arXiv2022

Scalar perturbations around rotating regular black holes and wormholes: quasi-normal modes, ergoregion instability and superradiance

We study scalar test-field perturbations on top of a Kerr--black-bounce background, i.e. a family of rotating regular black holes and/or rotating traversable wormholes that can mimic Kerr black holes. We compute the quasi-normal modes for a massless field in both the regular black holes and wormhole branches, confirming the stability of the former and identifying a set of growing modes that renders the latter unstable. We further compute the superradiance amplification factors, for massless and massive fields, in the regular black hole branch, confirming that these objects superradiate, though to a lesser degree than the corresponding Kerr black holes.

gr-qc↗

arXiv2021

Echoes in the Kerr/CFT correspondence

The Kerr/CFT correspondence is a possible route to gain insight into the quantum theory of gravity in the near-horizon region of a Kerr black hole via a dual holographic conformal field theory (CFT). Predictions of the black hole entropy, scattering cross-section and the quasi normal modes from the dual holographic CFT corroborate this proposed correspondence. More recently, it has been suggested that quantum gravitational effects in the near-horizon region of a black hole may drastically modify the classical general relativistic description, leading to potential observable consequences. In this paper, we study the absorption cross-section and quasi normal modes of a horizonless Kerr-like exotic compact object (ECO) in the dual CFT picture. Our analysis suggests that the near-horizon quantum modifications of the black hole can be understood as finite size and/or finite $N$ effects in the dual CFT. Signature of the near-horizon modification to a black hole geometry manifests itself as delayed echoes in the ringdown (i.e. the postmerger phase) of a binary black hole coalescence. From our dual CFT analysis we show how the length of the circle, on which the dual CFT lives, must be related to the echo time-delay that depends on the position of the near-horizon quantum structure. We further derive the reflectivity of the ECO membrane in terms of the CFT modular parameters, showing that it takes the Boltzmann form.

hep-th↗

arXiv2021

DPNNet-2.0 Part I: Finding hidden planets from simulated images of protoplanetary disk gaps

The observed sub-structures, like annular gaps, in dust emissions from protoplanetary disk, are often interpreted as signatures of embedded planets. Fitting a model of planetary gaps to these observed features using customized simulations or empirical relations can reveal the characteristics of the hidden planets. However, customized fitting is often impractical owing to the increasing sample size and the complexity of disk-planet interaction. In this paper we introduce the architecture of DPNNet-2.0, second in the series after DPNNet \citep{aud20}, designed using a Convolutional Neural Network ( CNN, here specifically ResNet50) for predicting exoplanet masses directly from simulated images of protoplanetary disks hosting a single planet. DPNNet-2.0 additionally consists of a multi-input framework that uses both a CNN and multi-layer perceptron (a class of artificial neural network) for processing image and disk parameters simultaneously. This enables DPNNet-2.0 to be trained using images directly, with the added option of considering disk parameters (disk viscosities, disk temperatures, disk surface density profiles, dust abundances, and particle Stokes numbers) generated from disk-planet hydrodynamic simulations as inputs. This work provides the required framework and is the first step towards the use of computer vision (implementing CNN) to directly extract mass of an exoplanet from planetary gaps observed in dust-surface density maps by telescopes such as the Atacama Large (sub-)Millimeter Array.

astro-ph.EP↗

arXiv2021

Ergoregion instability and echoes for braneworld black holes: Scalar, electromagnetic and gravitational perturbations

In the context of higher dimensional braneworld scenario, we have argued that the occurrence of horizonless exotic compact objects, as an alternative to classical black holes, are more natural. These exotic compact objects carry a distinctive signature of the higher dimension, namely a tidal charge parameter, inherited from the projection of the higher dimensional Weyl tensor onto the four dimensional spacetime we live in. Due to the absence of any horizon, rotating exotic compact objects are often unstable because of superradiance. Interestingly, these higher dimensional exotic compact objects, in the presence of the tidal charge, are more stable than their four dimensional counterpart. A similar inference is drawn by analysing the static modes associated with these exotic compact objects, irrespective of the nature of the perturbation i.e., it holds true for scalar, electromagnetic and also gravitational perturbation. The post-merger ringdown phase of the exotic compact object in the braneworld scenario, which can be described in terms of the quasi-normal modes, holds plethora of information regarding the nature of the higher dimension. In this connection we have discussed the analytical computation of the quasi-normal modes as well as their numerical estimation for perturbations of arbitrary spin, depicting existence of echoes in the ringdown waveform. As we have demonstrated, the echoes in the ringdown waveform depends explicitly on the tidal charge parameter and hence its future detection can provide constraints on the tidal charge parameter, which in turn will enable us to provide a possible bound on the length of the extra dimension.

gr-qc↗

arXiv2020

Echoes from the braneworld black holes

The holographic interpretation of the Randall-Sundrum model based on the adaptation of the AdS/CFT correspondence to the braneworld scenario states that the black holes localized on the brane are {\it quantum corrected}. This can be better understood from the fact that the classical $AdS_{5}$ bulk dynamics is dual to gravity coupled with CFT on the four dimensional brane. Based on the back reaction of the CFT on the classical black hole geometry, localized on the brane, it is expected that there exist possible near horizon modifications. This may result in the black hole horizon becoming partially reflective, thus giving rise to echoes in the ringdown signal in gravitational wave observations. In this paper, we investigate the existence of such echoes in the ringdown phase of a black hole localized on the brane, carrying a \emph{negative} tidal charge, and establish the layout for future investigation of higher dimensional effects in the ringdown signal. Confirmed detections of echoes at the current levels of instrumental sensitivity can constrain the dimensionless value of tidal charge to $|Q| \lesssim M^2$ .

gr-qc↗

arXiv2019

Black hole quantum atmosphere for freely falling observers

We analyze Hawking radiation as perceived by a freely-falling observer and try to draw an inference about the region of origin of the Hawking quanta. To do so, first we calculate the energy density from the stress energy tensor, as perceived by a freely-falling observer. Then we compare this with the energy density computed from an effective temperature functional which depends on the state of the observer. The two ways of computing these quantities show a mismatch at the light ring outside the black hole horizon. To better understand this ambiguity, we show that even taking into account the (minor) breakdown of the adiabatic evolution of the temperature functional which has a peak in the same region of the mismatch, is not enough to remove it. We argue that the appearance of this discrepancy can be traced back to the process of particle creation by showing how the Wentzel--Kramers--Brillouin approximation for the field modes breaks down between the light ring at $3M$ and $4M$, with a peak at $r=3.3M$ exactly where the energy density mismatch is maximized. We hence conclude that these facts strongly support a scenario where the Hawking flux does originate from a "quantum atmosphere" located well outside the black hole horizon.

gr-qc↗

arXiv2018

Noether Current, Black Hole Entropy and Spacetime Torsion

We show that the presence of spacetime torsion, unlike any other non-trivial modifications of the Einstein gravity, does \emph{not} affect black hole entropy. The theory being diffeomorphism invariant leads to a Noether current and hence to a Noether charge, which can be associated to the heat content of the spacetime. Furthermore, the evolution of the spacetime inheriting torsion can be encoded in the difference between suitably defined surface and bulk degrees of freedom. For static spacetimes the surface and bulk degrees of freedom coincides, leading to holographic equipartition. In doing so one can see that the surface degrees of freedom originate from horizon area and it is clear that spacetime torsion never contributes to the surface degrees of freedom and hence neither to the black hole entropy. We further show that the gravitational Hamiltonian in presence of torsion does not inherit any torsion dependence in the boundary term and hence the first law originating from the variation of the Hamiltonian, relates entropy to area. This reconfirms our claim that torsion does not modify the black hole entropy.

gr-qc↗

arXiv2017

The black hole quantum atmosphere

Ever since the discovery of black hole evaporation, the region of origin of the radiated quanta has been a topic of debate. Recently it was argued by Giddings that the Hawking quanta originate from a region well outside the black hole horizon by calculating the effective radius of a radiating body via the Stefan--Boltzmann law. In this paper we try to further explore this issue and end up corroborating this claim, using both a heuristic argument and a detailed study of the stress energy tensor. We show that the Hawking quanta originate from what might be called a quantum atmosphere around the black hole with energy density and fluxes of particles peaked at about $4M$, running contrary to the popular belief that these originate from the ultra high energy excitations very close to the horizon. This long distance origin of Hawking radiation could have a profound impact on our understanding of the information and transplanckian problems.

gr-qc↗

arXiv2017

Spacetime thermodynamics in the presence of torsion

It was shown by Jacobson in 1995 that the Einstein equation can be derived as a local constitutive equation for an equilibrium spacetime thermodynamics. With the aim to understand if such thermodynamical description is an intrinsic property of gravitation, many attempts have been done so far to generalise this treatment to a broader class of gravitational theories. Here we consider the case of the Einstein-Cartan theory as a prototype of theories with non-propagating torsion. In doing so, we study the properties of Killing horizons in the presence of torsion, establish the notion of local causal horizon in Riemann-Cartan spacetimes, and derive the generalised Raychaudhuri equation for this kind of geometries. Then, starting with the entropy that can be associated to these local causal horizons, we derive the Einstein-Cartan equation by implementing the Clausius equation. We outline two ways of proceeding with the derivation depending on whether we take torsion as a geometric field or as a matter field. In both cases we need to add internal entropy production terms to the Clausius equation as the shear and twist cannot be taken to be zero a priori for our setup. This fact implies the necessity of a non-equilibrium thermodynamics treatment for the local causal horizon. Furthermore, it implies that a non-zero twist at the horizon will in general contribute to the Hartle-Hawking tidal heating for black holes with possible implications for future observations.

gr-qc↗

arXiv2016

AdS and dS black hole solutions in analogue gravity: The relativistic and non-relativistic cases

We show that Schwarzschild black hole solutions in asymptotically Anti-de Sitter (AdS) and de Sitter (dS) spaces may, up to a conformal factor, be reproduced in the framework of analogue gravity. The aforementioned derivation is performed using relativistic and non-relativistic Bose-Einstein condensates. In addition, we demonstrate that the (2+1) planar AdS black hole can be mapped into the non-relativistic acoustic metric. Given that AdS black holes are extensively employed in the gauge/gravity duality, we then comment on the possibility to study the AdS/CFT correspondence and gravity/fluid duality from an analogue gravity perspective.

gr-qc↗

arXiv2016

Higher derivative gravity: field equation as the equation of state

One of the striking features of general relativity is that the Einstein equation is implied by the Clausius relation imposed on a small patch of locally constructed causal horizon. Extension of this thermodynamic derivation of the field equation to more general theories of gravity has been attempted many times in the last two decades. In particular, equations of motion for minimally coupled higher curvature theories of gravity, but without the derivatives of curvature, have previously been derived using a thermodynamic reasoning. In that derivation the horizon slices were endowed with an entropy density whose form resembles that of the Noether charge for diffeomorphisms, and was dubbed the Noetheresque entropy. In this paper, we propose a new entropy density, closely related to the Noetheresque form, such that the field equation of any diffeomorphism invariant metric theory of gravity can be derived by imposing the Clausius relation on a small patch of local causal horizon.

gr-qc↗