arXiv subjects

Yuting Liu

Publications and source records attributed to Yuting Liu.

76 records · Page 5Linked to original sources

arXiv2019

Phase transitions and entropy force of charged de Sitter black holes with cloud of string and quintessence

In this paper, we investigate the combined effects of the cloud of strings and quintessence on the thermodynamics of a Reissner-Nordström-de Sitter black hole. Based on the equivalent thermodynamic quantities considering the correlation between the black hole horizon and the cosmological horizon, we extensively discuss the phase transitions of the space-time. Our analysis prove that similar to the case in AdS space-time, second-order phase transitions could take place under certain conditions, with the absence of first-order phase transition in the charged de Sitter black holes with cloud of string and quintessence. The effects of different thermodynamic quantities on the phase transitions are also quantitatively discussed, which provides a new approach to study the thermodynamic qualities of unstable dS space-time. Focusing on the entropy force generated by the interaction between the black hole horizon and the cosmological horizon, as well as the Lennard-Jones force between two particles, our results demonstrate the strong degeneracy between the entropy force of the two horizons and the ratio of the horizon positions, which follows the surprisingly similar law given the relation between the Lennard-Jones force and the ratio of two particle positions. Therefore, the study of the entropy force between two horizons, is not only beneficial to the deep exploration of the three modes of cosmic evolution, but also helpful to understand the correlation between the microstates of particles in black holes and those in ordinary thermodynamic systems.

hep-th↗

arXiv2019

Point in, Box out: Beyond Counting Persons in Crowds

Modern crowd counting methods usually employ deep neural networks (DNN) to estimate crowd counts via density regression. Despite their significant improvements, the regression-based methods are incapable of providing the detection of individuals in crowds. The detection-based methods, on the other hand, have not been largely explored in recent trends of crowd counting due to the needs for expensive bounding box annotations. In this work, we instead propose a new deep detection network with only point supervision required. It can simultaneously detect the size and location of human heads and count them in crowds. We first mine useful person size information from point-level annotations and initialize the pseudo ground truth bounding boxes. An online updating scheme is introduced to refine the pseudo ground truth during training; while a locally-constrained regression loss is designed to provide additional constraints on the size of the predicted boxes in a local neighborhood. In the end, we propose a curriculum learning strategy to train the network from images of relatively accurate and easy pseudo ground truth first. Extensive experiments are conducted in both detection and counting tasks on several standard benchmarks, e.g. ShanghaiTech, UCF_CC_50, WiderFace, and TRANCOS datasets, and the results show the superiority of our method over the state-of-the-art.

cs.CV↗

arXiv2019

Implications of the lens redshift distribution of strong lensing systems: cosmological parameters and the global properties of early-type galaxies

In this paper, we assemble a well-defined sample of early-type gravitational lenses extracted from a large collection of 158 systems, and use the redshift distribution of galactic-scale lenses to test the standard cosmological model ($Λ$CDM) and the modified gravity theory (DGP). Two additional sub-samples are also included to account for possible selection effect introduced by the detectability of lens galaxies. Our results show that independent measurement of the matter density parameter ($Ω_m$) could be expected from such strong lensing statistics. Based on future measurements of strong lensing systems from the forthcoming LSST survey, one can expect $Ω_m$ to be estimated at the precision of $ΔΩ_m\sim 0.006$, which provides a better constraint on $Ω_m$ than \textit{Planck} 2015 results. Moreover, use the lens redshift test is also used to constrain the characteristic velocity dispersion of the lensing galaxies, which is well consistent with that derived from the optical spectroscopic observations. A parameter $f_E$ is adopted to quantify the relation between the lensing-based velocity dispersion and the corresponding stellar value. Finally, the accumulation of detectable galactic lenses from future LSST survey would lead to more stringent fits of $Δf_E\sim10^{-3}$, which encourages us to test the global properties of early-type galaxies at much higher accuracy.

astro-ph.CO↗

arXiv2018

Finite Sample Analysis of the GTD Policy Evaluation Algorithms in Markov Setting

In reinforcement learning (RL) , one of the key components is policy evaluation, which aims to estimate the value function (i.e., expected long-term accumulated reward) of a policy. With a good policy evaluation method, the RL algorithms will estimate the value function more accurately and find a better policy. When the state space is large or continuous \emph{Gradient-based Temporal Difference(GTD)} policy evaluation algorithms with linear function approximation are widely used. Considering that the collection of the evaluation data is both time and reward consuming, a clear understanding of the finite sample performance of the policy evaluation algorithms is very important to reinforcement learning. Under the assumption that data are i.i.d. generated, previous work provided the finite sample analysis of the GTD algorithms with constant step size by converting them into convex-concave saddle point problems. However, it is well-known that, the data are generated from Markov processes rather than i.i.d. in RL problems.. In this paper, in the realistic Markov setting, we derive the finite sample bounds for the general convex-concave saddle point problems, and hence for the GTD algorithms. We have the following discussions based on our bounds. (1) With variants of step size, GTD algorithms converge. (2) The convergence rate is determined by the step size, with the mixing time of the Markov process as the coefficient. The faster the Markov processes mix, the faster the convergence. (3) We explain that the experience replay trick is effective by improving the mixing property of the Markov process. To the best of our knowledge, our analysis is the first to provide finite sample bounds for the GTD algorithms in Markov setting.

cs.LG↗