SearcharxivSearch

arXiv subjects

Yangkang Zhang

Publications and source records attributed to Yangkang Zhang.

2 recordsLinked to original sources

The radio detection and accretion properties of the peculiar nuclear transient AT 2019avd

AT 2019avd is a nuclear transient detected from infrared to soft X-rays, though its nature is yet unclear. The source has shown two consecutive flaring episodes in the optical and the infrared bands and its second flare was covered by X-ray monitoring programs. During this flare, the UVOT/Swift photometries revealed two plateaus: one observed after the peak and the other one appeared ~240 days later. Meanwhile, our NICER and XRT/Swift campaigns show two declines in the X-ray emission, one during the first optical plateau and one 70-90 days after the optical/UV decline. The evidence suggests that the optical/UV could not have been primarily originated from X-ray reprocessing. Furthermore, we detected a timelag of ~16-34 days between the optical and UV emission, which indicates the optical likely comes from UV reprocessing by a gas at a distance of 0.01-0.03 pc. We also report the first VLA and VLBA detection of this source at different frequencies and different stages of the second flare. The information obtained in the radio band - namely a steep and a late-time inverted radio spectrum, a high brightness temperature and a radio-loud state at late times - together with the multiwavelength properties of AT 2019avd suggests the launching and evolution of outflows such as disc winds or jets. In conclusion, we propose that after the ignition of black hole activity in the first flare, a super-Eddington flaring accretion disc formed and settled to a sub-Eddington state by the end of the second flare, associated with a compact radio outflow.

astro-ph.HE

DGL-LifeSci: An Open-Source Toolkit for Deep Learning on Graphs in Life Science

Graph neural networks (GNNs) constitute a class of deep learning methods for graph data. They have wide applications in chemistry and biology, such as molecular property prediction, reaction prediction and drug-target interaction prediction. Despite the interest, GNN-based modeling is challenging as it requires graph data pre-processing and modeling in addition to programming and deep learning. Here we present DGL-LifeSci, an open-source package for deep learning on graphs in life science. DGL-LifeSci is a python toolkit based on RDKit, PyTorch and Deep Graph Library (DGL). DGL-LifeSci allows GNN-based modeling on custom datasets for molecular property prediction, reaction prediction and molecule generation. With its command-line interfaces, users can perform modeling without any background in programming and deep learning. We test the command-line interfaces using standard benchmarks MoleculeNet, USPTO, and ZINC. Compared with previous implementations, DGL-LifeSci achieves a speed up by up to 6x. For modeling flexibility, DGL-LifeSci provides well-optimized modules for various stages of the modeling pipeline. In addition, DGL-LifeSci provides pre-trained models for reproducing the test experiment results and applying models without training. The code is distributed under an Apache-2.0 License and is freely accessible at https://github.com/awslabs/dgl-lifesci.

cs.LG