Characterizing Ext xtragalactic Pre-Main- Sequence Stars wit ith - PowerPoint PPT Presentation

Characterizing Ext xtragalactic Pre-Main- Sequence Stars wit ith Machine and Deep Learnin ing Techniques Victor Francisco Ksoll Understanding the Nearby Star-forming Universe with JWST 27.08.2019

Star Forming Regions in the Large Magellanic Cloud Credit: NASA, ESA, E. Sabbi (STScI) Credit: D. Gouliermis and MYSST Data: Hubble Tarantula Treasury Project HTTP Data: Measuring Young Stars in Space and Time MYSST Region: Tarantula Nebula Region: Star forming Complex Nebula N44 Sources: > 800.000 Sources: > 400.000 Observed Area: 210 pc x 180 pc Observed Area: 135 pc x 190 pc Filters: F275W, F336W, F555W, F658N, F775W, F110W, F160W Filters: F555W, F814W

Supervised ML classification of f stars CMD Spatial Map [mag] [mag]

Supervised ML classification of f stars Training Set: R136 CMD Spatial Map [mag] [mag]

Supervised ML classification of f stars CMD Spatial Map [mag] [mag]

Supervised ML classification of f stars • ML-based classification of PMS stars with RFs and SVM techniques CMD Spatial Map [mag] [mag] Ksoll, et al 2018, MNRAS, Volume 479, Issue 2, p2389-2414

Unsupervised Clustering of f stars • Contour Density Based Clustering Analysis Dendrogram Spatial Density Map Ksoll, Gouliermis et al., in prep

N44 Classification CMD Spatial Map

Goals and Methods Identifying PMS-Stars: Characterizing PMS-Stars • Aim: Separate Young from Old Stars • Aim: Determine Physical Parameters of Stars • Task: Supervised Classification • Task: Regression • Training set: Observations • Training set: Theoretical Models • Photometry of known Clusters • Isochrones from Stellar Evolution Models • Algorithms: • Algorithms: Support Vector Machine (SVM), Invertible Neural Network (INN) Random Forest (RF)

Physical parameter regression with an In Invertible Neural Network (I (INN)

Physical parameter regression with an In Invertible Neural Network (I (INN)

Physical parameter regression with an In Invertible Neural Network (I (INN) Ardizzone, et al 2018, arXiv:1808.04730

Physical parameter regression with an In Invertible Neural Network (I (INN) [Msun] [yr] [yr] [Msun] Current Training Progress

IN INN Application Test Case Westerlund 2 1.5 Myr Isochrone Credit: NASA, ESA, Antonella Nota (ESA/STScI), E. Sabbi (STScI) Photometry in F814W and F160W Cluster Sources: ca. 6200 Gas extinction map: Zeidler et al. 2015, AJ. Age (Zeidler et al. 2015): 0.5 – 2.0 Myr

IN INN Application Test Case Westerlund 2 Credit: NASA, ESA, Antonella Nota (ESA/STScI), E. Sabbi (STScI) Photometry in F814W and F160W Cluster Sources: ca. 6200 Gas extinction map: Zeidler et al. 2015, AJ. Age (Zeidler et al. 2015): 0.5 – 2.0 Myr

IN INN Application Test Case Westerlund 2 Credit: NASA, ESA, Antonella Nota (ESA/STScI), E. Sabbi (STScI) Photometry in F814W and F160W Cluster Sources: ca. 6200 Gas extinction map: Zeidler et al. 2015, AJ. Age (Zeidler et al. 2015): 0.5 – 2.0 Myr

IN INN Application Test Case Westerlund 2 Credit: NASA, ESA, Antonella Nota (ESA/STScI), E. Sabbi (STScI)

Thank you! INN Framework: github.com/VLL-HD/FrEIA The entire Field of View of the MYSST survey