Niharika Sravan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Niharika Sravan has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Andrew Cantrell
Department of Astrophysics, Yale University, P. O. Box 208101, New Haven, CT 06520, USA
2 shared publications - Charles D. Bailyn
Yale University
2 shared publications - Ilya Mandel
Monash University
2 shared publications - Laura Kreidberg
Department of Astrophysics, Yale University, P. O. Box 208101, New Haven, CT 06520, USA
2 shared publications - Will M. Farr
Center for Interdisciplinary Exploration and Research in Astrophysics Department of Physics and Astronomy Northwestern University 2145 Sheridan Road 60208 Evanston IL
2 shared publications - Matthew J. Graham
California Institute of Technology
1 shared publication
Latest Publications
- Open Access
- 1 October 2024
- Published by American Astronomical Society in The Astrophysical Journal
- Vol. 974 (2) , 214
Machine-directed Gravitational-wave Counterpart Discovery
Joint observations in electromagnetic and gravitational waves shed light on the physics of objects and surrounding environments with extreme gravity that are otherwise unreachable via siloed observations in each messenger. However, such detections remain challenging due to the rapid and faint nature of counterparts. Protocols for discovery and inference still rely on human experts manually inspecting survey alert streams and intuiting optimal usage of limited follow-up resources. Strategizing an optimal follow-up program requires adaptive sequential decision-making given evolving light curve data that (i) maximizes a global objective despite incomplete information and (ii) is robust to stochasticity introduced by detectors/observing conditions. Reinforcement learning (RL) approaches allow agents to implicitly learn the physics or detector dynamics and the behavior policy that maximizes a designated objective through experience. To demonstrate the utility of such an approach for the kilonova follow-up problem, we train a toy RL agent with the goal of maximizing follow-up photometry for the true kilonova among several contaminant transient light curves. In a simulated environment where the agent learns online, it achieves 3× higher accuracy compared to a random strategy. However, it is surpassed by human agents by up to a factor of 2. This is likely because our hypothesis function (Q that is linear in state-action features) is an insufficient representation of the optimal behavior policy. More complex agents could perform at par or surpass human experts. Agents like these could pave the way for machine-directed software infrastructure to efficiently respond to next generation detectors, for conducting science inference and optimally planning expensive follow-up observations, scalably and with demonstrable performance guarantees.
- 1 October 2024
- Published by American Astronomical Society in The Astrophysical Journal
- Vol. 974 (2) , 214
Open AccessMachine-directed Gravitational-wave Counterpart Discovery
Joint observations in electromagnetic and gravitational waves shed light on the physics of objects and surrounding environments with extreme gravity that are otherwise unreachable via siloed observations in each messenger. However, such detections remain challenging due to the rapid and faint nature of counterparts. Protocols for discovery and inference still rely on human experts manually inspecting survey alert streams and intuiting optimal usage of limited follow-up resources. Strategizing an optimal follow-up program requires adaptive sequential decision-making given evolving light curve data that (i) maximizes a global objective despite incomplete information and (ii) is robust to stochasticity introduced by detectors/observing conditions. Reinforcement learning (RL) approaches allow agents to implicitly learn the physics or detector dynamics and the behavior policy that maximizes a designated objective through experience. To demonstrate the utility of such an approach for the kilonova follow-up problem, we train a toy RL agent with the goal of maximizing follow-up photometry for the true kilonova among several contaminant transient light curves. In a simulated environment where the agent learns online, it achieves 3× higher accuracy compared to a random strategy. However, it is surpassed by human agents by up to a factor of 2. This is likely because our hypothesis function (Q that is linear in state-action features) is an insufficient representation of the optimal behavior policy. More complex agents could perform at par or surpass human experts. Agents like these could pave the way for machine-directed software infrastructure to efficiently respond to next generation detectors, for conducting science inference and optimally planning expensive follow-up observations, scalably and with demonstrable performance guarantees.
- Open Access
The Mass Distribution of Stellar-Mass Black Holes
We perform a Bayesian analysis of the mass distribution of stellar-mass black holes using the observed masses of 15 low-mass X-ray binary systems undergoing Roche lobe overflow and five high-mass, wind-fed X-ray binary systems. Using Markov Chain Monte Carlo calculations, we model the mass distribution both parametrically---as a power law, exponential, gaussian, combination of two gaussians, or log-normal distribution---and non-parametrically---as histograms with varying numbers of bins. We provide confidence bounds on the shape of the mass distribution in the context of each model and compare the models with each other by calculating their relative Bayesian evidence as supported by the measurements, taking into account the number of degrees of freedom of each model. The mass distribution of the low-mass systems is best fit by a power-law, while the distribution of the combined sample is best fit by the exponential model. We examine the existence of a "gap" between the most massive neutron stars and the least massive black holes by considering the value, M_1%, of the 1% quantile from each black hole mass distribution as the lower bound of black hole masses. The best model (the power law) fitted to the low-mass systems has a distribution of lower-bounds with M_1% > 4.3 Msun with 90% confidence, while the best model (the exponential) fitted to all 20 systems has M_1% > 4.5 Msun with 90% confidence. We conclude that our sample of black hole masses provides strong evidence of a gap between the maximum neutron star mass and the lower bound on black hole masses. Our results on the low-mass sample are in qualitative agreement with those of Ozel, et al (2010).
Open AccessThe Mass Distribution of Stellar-Mass Black Holes
We perform a Bayesian analysis of the mass distribution of stellar-mass black holes using the observed masses of 15 low-mass X-ray binary systems undergoing Roche lobe overflow and five high-mass, wind-fed X-ray binary systems. Using Markov Chain Monte Carlo calculations, we model the mass distribution both parametrically---as a power law, exponential, gaussian, combination of two gaussians, or log-normal distribution---and non-parametrically---as histograms with varying numbers of bins. We provide confidence bounds on the shape of the mass distribution in the context of each model and compare the models with each other by calculating their relative Bayesian evidence as supported by the measurements, taking into account the number of degrees of freedom of each model. The mass distribution of the low-mass systems is best fit by a power-law, while the distribution of the combined sample is best fit by the exponential model. We examine the existence of a "gap" between the most massive neutron stars and the least massive black holes by considering the value, M_1%, of the 1% quantile from each black hole mass distribution as the lower bound of black hole masses. The best model (the power law) fitted to the low-mass systems has a distribution of lower-bounds with M_1% > 4.3 Msun with 90% confidence, while the best model (the exponential) fitted to all 20 systems has M_1% > 4.5 Msun with 90% confidence. We conclude that our sample of black hole masses provides strong evidence of a gap between the maximum neutron star mass and the lower bound on black hole masses. Our results on the low-mass sample are in qualitative agreement with those of Ozel, et al (2010).
- Top Cited Papers
- 24 October 2011
- Published by American Astronomical Society in The Astrophysical Journal
- Vol. 741 (2) , 103
THE MASS DISTRIBUTION OF STELLAR-MASS BLACK HOLES
We perform a Bayesian analysis of the mass distribution of stellar-mass black holes using the observed masses of 15 low-mass X-ray binary systems undergoing Roche lobe overflow and five high-mass, wind-fed X-ray binary systems. Using MCMC calculations, we model the mass distribution both parametrically---as a power law, exponential, gaussian, combination of two gaussians, or log-normal distribution---and non-parametrically---as histograms with varying numbers of bins. We provide confidence bounds on the shape of the mass distribution in the context of each model and compare the models by calculating their Bayesian evidence. The mass distribution of the low-mass systems is best fit by a power-law, while the distribution of the combined sample is best fit by the exponential model. This difference indicates that the low-mass subsample is not consistent with being drawn from the distribution of the combined population. We examine the existence of a ``gap'' between the most massive neutron stars and the least massive black holes by considering the 1% quantile from each black hole mass distribution, M_1%. The best model (the power law) fitted to the low-mass systems gives a distribution with M_1% > 4.3 MSun with 90% confidence, while the best model (the exponential) fitted to all 20 systems has M_1% > 4.5 MSun with 90% confidence. We conclude that our sample of black hole masses provides strong evidence of a gap between the maximum neutron star mass and the minimum black hole mass. Our results on the low-mass sample are in qualitative agreement with those of Ozel et al (2010), although our broad model-selection analysis more reliably reveals the best-fit underlying mass distribution. The presence of a mass gap remains theoretically unexplained.
- 24 October 2011
- Published by American Astronomical Society in The Astrophysical Journal
- Vol. 741 (2) , 103
Top Cited PapersTHE MASS DISTRIBUTION OF STELLAR-MASS BLACK HOLES
We perform a Bayesian analysis of the mass distribution of stellar-mass black holes using the observed masses of 15 low-mass X-ray binary systems undergoing Roche lobe overflow and five high-mass, wind-fed X-ray binary systems. Using MCMC calculations, we model the mass distribution both parametrically---as a power law, exponential, gaussian, combination of two gaussians, or log-normal distribution---and non-parametrically---as histograms with varying numbers of bins. We provide confidence bounds on the shape of the mass distribution in the context of each model and compare the models by calculating their Bayesian evidence. The mass distribution of the low-mass systems is best fit by a power-law, while the distribution of the combined sample is best fit by the exponential model. This difference indicates that the low-mass subsample is not consistent with being drawn from the distribution of the combined population. We examine the existence of a ``gap'' between the most massive neutron stars and the least massive black holes by considering the 1% quantile from each black hole mass distribution, M_1%. The best model (the power law) fitted to the low-mass systems gives a distribution with M_1% > 4.3 MSun with 90% confidence, while the best model (the exponential) fitted to all 20 systems has M_1% > 4.5 MSun with 90% confidence. We conclude that our sample of black hole masses provides strong evidence of a gap between the maximum neutron star mass and the minimum black hole mass. Our results on the low-mass sample are in qualitative agreement with those of Ozel et al (2010), although our broad model-selection analysis more reliably reveals the best-fit underlying mass distribution. The presence of a mass gap remains theoretically unexplained.