Mission Overview

Classifying Pan-STARRS sources with unsupervised and supervised learning

This entry explores how to classify unresolved sources extracted from the Pan-STARRS survey. The PS1-STRM  team used a convolutional neural network to classify sources into "galaxy", "qso" , "star" and "unsure" classes based on the observed source fluxes. Here we demonstrate how to use simple dimensionality reduction, as well as unsupervised and supervised classification algorithms (k-means and stochastic gradient descent, respectively), to reproduce these classifications. We then compare the performance of these models with the published results. 

Data:  The PS1-STRM HLSP

Notebook: Classifying Pan-STARRS sources with (un)supervised learning

Released: 2022-06-12

Updated: 2022-06-12

Tags: classification, 1d-data, unsupervised, supervised, PCA, tSNE, k-means, clustering

alternate
Pan-STARRS mosaic image (credit: R. White, STScI / PS1 Science Consortium).