Title Text

 

University of Delaware

Department of Physics and Astronomy

 

federica bianco

Biden School of Public Policy and Administration

Data  Science Institute

 

Vera C. Rubin Observatory

FASTLab

Federica's Astrostattistics Lab

Former UD GS

Current UD GS

The immutable skies

 Bartolomeu Velho, 1568 (Bibliothèque Nationale, Paris)

1549 Oronce Fine, France

From Flammarion's Astronomie Populaire (1880) Denmark

Workshop of Diebold Lauber unknown artist, ca.1450

explosions in the sky

how we study SNe

@fedhere

The Rubin Observatory Legacy Survey of Space and Time will start soon:

the largest astronomical survey ever attempted

400 hours of 4k HD videos or 1,000,000 typical TikTok videos every night

400 hours of 4k HD videos or  2years of TikTok videos every night

Input

x

y

output

data

prediction

physics

Machine Learning

Machine Learning

Input

x

y

output

function

f(x)

Machine Learning

Input

x

y

output

f(x)
f(x) = mx + b

b

m

m: slope 

b: intercept

Machine Learning

Input

x

y

output

f(x)
f(x) = mx + b

b

m

m: slope 

b: intercept

parameters

x

y

learn

goal: find the right m and b that turn x into y

goal: find the right m and b that turn x into y

Machine Learning

p(class)

 pixel values tensor

f(x)

GPT-3

175 Billion Parameters

3,640 PetaFLOPs days

Kaplan+ 2020

Will we discover new physics?

A survey of Machine Learning features and methods to discover unique objects and rare classes in regularly and irregularly sampled time series

Sparse, unevenly sampled Kepler time series

2D T-SNE projection of feature space

Weirdness score

Evenly sampled Kepler time series

When a new instrument opens a new window in the observable space we may discover new phenomena that challenge or understanding of the Universe

What survey strategy shoudl LSST adopt to maximize the chance of serendipitous discovery?

Xiaolong Li et al. 2022

now LSSTC Catalyst Fellow, J. Hopkins

Xiaolong Li et al. 2022

LSSTC Catalyst Fellow, J. Hopkins

AILE: the first AI-based platform for the detection and study of Light Echoes

NSF Award #2108841

P.I. Bianco

Light Ecoes are rare strophsyical pheonomena and a near-pessimal problem for AI, but with as much data as LSST AI is a necessity

 

  • imbalance classes
  • diverse morphology
  • low SNR
  • small training data
  • inaccurate labels

Dr. Somayeh Khakpash

LSSTC Catalyst Fellow, Rutgers

Rare classes will become common, but how do we know what we are looking at and classify different objects for sample studies?

Data-Driven Templates for stripped SESN

 

Dr. Somayeh Khakpash

now LSSTC Catalyst Fellow, Rutgers

Can AI replace forward modeling of physical processes?

Many problems in astrophysics require expensive simulations.

Can AI replace computationally expensive image processing now necessary for astrophysical discovery?

search

template

difference

-

=

Tatiana Acero-Cuellar et al. 2023

UNIDEL fellow

Saliency maps: what pixels matter? 

search

template

difference

Tatiana Acero-Cuellar in prep

UNIDEL fellow

search

template

&

Can AI replace computationally expensive image processing now necessary for astrophysical discovery?

Why did the AI made the decision it made?

template

search

NO DIFFERENCE

Building interpetable classifiers for sparse, multimodal astrophysical time series

Sid Chiaini et al. 2024, submitted!

Astronomy and Computing

Stars that flare

Magnitude -> Flare energy

Star displacement -> color -> flare temperature

NSF Award #2308016

P.I. Bianco

Riley Clarke et al. submitted!

AstroPhysical Journl Supplements

Willow Fox Fortino+...  2024 

Do we really need a spectrograph??

new transformer models!

Continuous readout astronomical images for anomaly detection

Shar Daniels, PhD candidate

new transformer models!

Testing the performancde os SAM on astronomical objects

Rodiat Ayinde

NSF Award #2123264

P.I. Bianco

7 bands

sparse data

 As the nation’s first degree-granting Historically Black College and University (HBCU), The Lincoln University paved a path to higher education for African American males previously unavailable to them.

NSF  Data Science Corps Award 2123264

to create an equitable and accessible data science pedagogical program and build data science pedagogy capacity at HBCUs

Time (days)

Brightness (flux)

The tragic (and beautiful) complexity of the LSST data

Time (days)

Brightness (flux)

The tragic (and beautiful) complexity of the LSST data

7 bands

Time (days)

Brightness (flux)

The tragic (and beautiful) complexity of the LSST data

7 bands

sparse data

Time (days)

Brightness (flux)

The tragic (and beautiful) complexity of the LSST data

7 bands

sparse data

environmental info

Time (days)

Brightness (flux)

The tragic (and beautiful) complexity of the LSST data

7 bands

sparse data

low SNR

environmental info

DRONE:

the sustained tone branch of minimalism

Research Inclusion: sonification of LSST lightcurves

Presented at 2021 Rubin Project Community Workshop (300 ppl) 

Started a new Rubin Working Group

Rubin Rhapsodies

thank you!

 

University of Delaware

Department of Physics and Astronomy

 

Biden School of Public Policy and Administration

Data  Science Institute

@fedhere

federica bianco

fbianco@udel.edu

Prospective Students lightening talk

By federica bianco

Prospective Students lightening talk

  • 11