Host Galaxy Based Supernova Classifiication
Main Authors: | Chou, Noel, Villar, Ashley, Berger, Edo, Jones, David, Ntampaka, Michelle |
---|---|
Format: | info publication-thesis eJournal |
Terbitan: |
, 2020
|
Subjects: | |
Online Access: |
https://zenodo.org/record/3753220 |
ctrlnum |
3753220 |
---|---|
fullrecord |
<?xml version="1.0"?>
<dc schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ http://www.openarchives.org/OAI/2.0/oai_dc.xsd"><creator>Chou, Noel</creator><creator>Villar, Ashley</creator><creator>Berger, Edo</creator><creator>Jones, David</creator><creator>Ntampaka, Michelle</creator><date>2020-04-15</date><description>Upcoming optical surveys such as the Large Synoptic Survey Telescope will discover supernovae at rates far out-pacing feasible spectroscopic classification. It is therefore critical that we optimize alternate classification methods using all available information. The use of host galaxy data for classification has not been well-developed, despite well-known trends between host galaxy properties and supernova types, and machine learning methods are well suited to this task of identifying complex trends. Using Pan-STARRS1 Medium-Deep Survey (PS1-MDS) images, we trained machine learning algorithms to predict supernovae types using solely contextual information. In particular, we present a random forest classier using known host galaxy properties, and a convolutional neural network using host galaxy images. Classifying between types Ia, Ibc, II, IIn, and superluminous, we find the convolutional neural network performs significantly better than the random forest classier. They achieve average accuracies per class of 63% and 47% respectively. The convolutional neural network performs similarly across most classes with accuracies above 50%, whereas the random forest struggles to classify types IIn and Ibc. Classifying between supernovae Ia and all other types, we achieve similar accuracy with both algorithms, with 70% in the random forest and 74% in the CNN. This is a significant improvement from existing host-galaxy based classification work (Foley and Mandel 2013). Future work includes combining our algorithms with photometric classification pipelines for optimized classification.</description><description>Astro 98</description><identifier>https://zenodo.org/record/3753220</identifier><identifier>10.5281/zenodo.3753220</identifier><identifier>oai:zenodo.org:3753220</identifier><relation>doi:10.5281/zenodo.3753219</relation><relation>url:https://zenodo.org/communities/cfa-theses</relation><rights>info:eu-repo/semantics/openAccess</rights><rights>https://creativecommons.org/licenses/by/4.0/legalcode</rights><subject>Supernovae</subject><subject>Random Forest Classifier</subject><subject>Convolutional Neural Network</subject><subject>Astro 98</subject><subject>Supernovae</subject><subject>Random Forest</subject><subject>Convolutional Neural Networks</subject><title>Host Galaxy Based Supernova Classifiication</title><type>Other:info:eu-repo/semantics/doctoralThesis</type><type>Other:publication-thesis</type><recordID>3753220</recordID></dc>
|
format |
Other:info:eu-repo/semantics/doctoralThesis Other Other:publication-thesis Journal:eJournal Journal |
author |
Chou, Noel Villar, Ashley Berger, Edo Jones, David Ntampaka, Michelle |
title |
Host Galaxy Based Supernova Classifiication |
publishDate |
2020 |
topic |
Supernovae Random Forest Classifier Convolutional Neural Network Astro 98 Random Forest Convolutional Neural Networks |
url |
https://zenodo.org/record/3753220 |
contents |
Upcoming optical surveys such as the Large Synoptic Survey Telescope will discover supernovae at rates far out-pacing feasible spectroscopic classification. It is therefore critical that we optimize alternate classification methods using all available information. The use of host galaxy data for classification has not been well-developed, despite well-known trends between host galaxy properties and supernova types, and machine learning methods are well suited to this task of identifying complex trends. Using Pan-STARRS1 Medium-Deep Survey (PS1-MDS) images, we trained machine learning algorithms to predict supernovae types using solely contextual information. In particular, we present a random forest classier using known host galaxy properties, and a convolutional neural network using host galaxy images. Classifying between types Ia, Ibc, II, IIn, and superluminous, we find the convolutional neural network performs significantly better than the random forest classier. They achieve average accuracies per class of 63% and 47% respectively. The convolutional neural network performs similarly across most classes with accuracies above 50%, whereas the random forest struggles to classify types IIn and Ibc. Classifying between supernovae Ia and all other types, we achieve similar accuracy with both algorithms, with 70% in the random forest and 74% in the CNN. This is a significant improvement from existing host-galaxy based classification work (Foley and Mandel 2013). Future work includes combining our algorithms with photometric classification pipelines for optimized classification. Astro 98 |
id |
IOS17403.3753220 |
institution |
Universitas PGRI Palembang |
institution_id |
189 |
institution_type |
library:university library |
library |
Perpustakaan Universitas PGRI Palembang |
library_id |
587 |
collection |
Marga Life in South Sumatra in the Past: Puyang Concept Sacrificed and Demythosized |
repository_id |
17403 |
city |
KOTA PALEMBANG |
province |
SUMATERA SELATAN |
repoId |
IOS17403 |
first_indexed |
2022-07-26T02:24:36Z |
last_indexed |
2022-07-26T02:24:36Z |
recordtype |
dc |
_version_ |
1739408003914792960 |
score |
17.608969 |