| |
 |
| Paper: |
Latent Space Explorer: Unsupervised Data Pattern Discovery on the Cloud |
| Volume: |
535, Astronomical Data Analysis Software and Systems XXXI |
| Page: |
61 |
| Authors: |
Cecconello, T.; Bordiu, C.; Bufano, F.; Puerari, L.; Riggi, S.; Schisano, E.; Sciacca, E.; Maruccia, Y.; Vizzari, G. |
| Abstract: |
The amount of data describing astronomical phenomena is growing at an overwhelming rate: tools supporting a computer assisted analysis of these data is of ever-growing importance. We propose an approach based on unsupervised machine learning as a central element of an overall workflow, involving domain experts and computer scientists; the workflow includes three phases: (i) achieving a compact representation of elements of the dataset by means of representation learning techniques, shifting the analysis from cumbersome representations to compact vectors in a latent space, (ii) visualizing results of this analysis in a 2D or 3D space (further reducing dimensionality of the space) and (iii) potentially clustering points associated to instances to suggest patterns to the domain experts that will evaluate their potential meaning within the domain. This work presents the overall approach within a cloud-based setting, and results on two specific astronomical research topics, namely the study of galactic supernova remnants and star forming clumps. |
|
|
|
|
|
