CLOVER: Convolutional neural network spectra identifier and kinematics predictor

Discussion topics for individual codes
Post Reply
Ada Coda
ASCL Robot
Posts: 1574
Joined: Thu May 08, 2014 5:37 am

CLOVER: Convolutional neural network spectra identifier and kinematics predictor

Post by Ada Coda » Mon Sep 30, 2019 4:01 am

CLOVER: Convolutional neural network spectra identifier and kinematics predictor

Abstract: CLOVER (Convnet Line-fitting Of Velocities in Emission-line Regions) is a convolutional neural network (ConvNet) trained to identify spectra with two velocity components along the line of sight and predict their kinematics. It works with Gaussian emission lines (e.g., CO) and lines with hyperfine structure (e.g., NH3). CLOVER has two prediction steps, classification and parameter prediction. For the first step, CLOVER segments the pixels in an input data cube into one of three classes: noise (i.e., no emission), one-component (emission line with single velocity component), and two-component (emission line with two velocity components). For the pixels identified as two-components in the first step, a second regression ConvNet is used to predict centroid velocity, velocity dispersion, and peak intensity for each velocity component.

Credit: Keown, Jared; Di Francesco, James; Teimoorinia, Hossen; Rosolowsky, Erik; Chen, Michael Chun-Yuan

Site: https://github.com/jakeown/astroclover/
https://ui.adsabs.harvard.edu/abs/2019arXiv190908727K

Bibcode: 2019ascl.soft09009K

ID: ascl:1909.009

Post Reply