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
Machine Learning
Input
x
y
output
b
m
m: slope
b: intercept
Machine Learning
Input
x
y
output
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
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