VLASS > VLASS at AAS242 Abstracts

VLASS at AAS242 Abstracts

Explorers Guide to the VLA Sky Survey

Splinter Session

Creating a Neural Network to Classify VLASS Objects

Brian R. Kent (NRAO)

The VLA Sky Survey will generate a suite of data products
for the astronomical community.  These products include high resolution images and a catalog of millions of previously unresolved sources. In order to facilitate basic classification of these objects, we will show users how to build a neural network using Keras and the TensorFlow framework.  This deep learning exercise is common in the field of machine learning.  We will import a subset of VLASS data into a Jupyter notebook in Google Colab, while understanding basic data input requirements and manipulation in order to leverage TensorFlow.  We will build a model with various layers in order to optimize our neural network, understand how to accelerate the fitting process, and analyze our resulting classification predictions for model refinement. 

Talk material can be found here.

An Explorer’s Guide to the VLITE Commensal Sky Survey

Emil Polisensky (NRL)

The VLA Low-band Ionosphere and Transient Experiment (VLITE) is a commensal observing mode leveraging the infrastructure of the VLA to expand its capabilities with a separate scientific data stream. VLITE records data from the 340 MHz prime focus receivers in parallel with VLA observations utilizing the 1-50 GHz secondary focus receivers. The VLITE Commensal Sky Survey (VCSS) is generated from the low frequency commensal data recorded during VLASS. The on-the-fly observing mode and sky sampling method designed for the smaller, 3 GHz field of view produces ~160,000 highly overlapping VCSS snapshot images covering 32,000 square degrees with ~15” resolution, per observing epoch. I will describe the VCSS snapshots and data products, the publicly available bright source catalog, and science highlights from the first two observing epochs.