2019 09-06 skunkworks q&a information session v2.1 dist
1. Q&A Information Session
Dane Morgan
University of Wisconsin, Madison
ddmorgan@wisc.edu, W: 608-265-5879, C: 608-234-2906
UW Madison
ECB 1045, September 6, 2019 1
To Join: Send me email at ddmorgan@wisc.edu with your
name, email, major (intended if not set), and any relevant
facts/interests (e.g., have project already, strong machine
learning skills, know python, want only solar energy, …)
2. What do These Have in Common?
• Chess, Jeopardy, Go, Poker
• Language translation
• MRI based diagnosis
• Driving
• “will profoundly change human society and life and
change the world” (China’s State Council, 7/20/2017)
• “leader in this sphere will become the ruler of the
world” (Vladimir Putin, "science lesson" to start off the
Russian school year, 9/4/17)
• “most likely cause of WW3” (Elon Musk, Twitter,
9/4/17)
2
4. What is the Informatics Skunkworks?
The “Informatics Skunkworks” is a group
dedicated to realizing the potential of
informatics for science and engineering.
4
Vision: Transform science and
engineering with informatics
5. Why Form the Informations
Skunkworks?
Incredible opportunity for young creative
researchers
5
Massive Data New HardwareTransformative Tools
6. How the Informatics Skunkworks
Works – Big Picture
• You email/talk to me if you
are interested and we meet.
• We find you a project with a
mentor (me, another faculty,
industry representative) –
you can bring a project.
• You work on the project for
either credit (most common)
or pay (if available) and get
cool results.
6
7. How the Informatics Skunkworks
Works – Details
• ~10h/wk during the year (3
credits), possibly full time
over summer if adequate
funds and interest.
• Weekly (or every 2 weeks)
progress meetings with
mentors.
• Final all-hands meeting at
end of semester
7
8. Some Stuff the Skunkworks Has/Does
• Excellent web page to highlight our
accomplishments
(skunkworks.wisc.edu)
– Always looking for people to help
develop this
• Experienced members who know
powerful informatics tools (python,
matlab, SciKitLearn, tensorflow,
Citrine/Lolo, MASTML, etc.)
• Exciting problems and cool data sets
you can explore (mostly in materials)
• Many opportunities for posters, talks,
papers, etc.
8
9. Some Recent Skunkworks
Accomplishments
• Highlighted as one of the 32
accomplishments of the first five
years of the $500m Materials
Genome Initiative
• Finalist in the 2017 Wisconsin
Innovation Awards
• Papers: 3 published, 1 submitted.
• High-profile fellowships (Hilldale,
Univ. Boosksotre, Welton,
Wiscience, NextGen, …)
• Post-graduate success: Google,
Tesla, MIT, Duke, …
• Dozens of presentations at
conferences
9
10. Possible Projects
• Predicting nuclear materials degradation
• Predicting molecular properties from basic structure
• Detecting defects in microscopy images
• Predicting properties of photovoltaic and catalytic
materials
• Aiding medical diagnosis from medical images
• New unitless feature methods
• Developing automated machine learning tools
• Educational group – learn the basics!
• Distributed Computing in Advanced Instrumentation
(with Paul Evans, pgevans@wisc.edu)
And maybe others …
10
11. Example: Machine Learning for Impurity Diffusion
• Diffusion of element X in
host H is a key materials
property
• Machine-learning models
trained on with high-
throughput calculated data
can extend it by orders of
magnitude
• We have extended our
diffusion data by ~5x with
machine learning model,
saving years and ~$1m
http://diffusiondata.materialshub.org/
0.8
1.2
1.6
2.0
SoluteDiffusionBarrier[eV]
Sc
Y
La
Ti
Zr
Hf
V
Nb
Ta
Cr
Mo
W
Mn
Tc
Re
Fe
Ru
Os
Co
Rh
Ir
Ni
Pd
Pt
Cu
Ag
Au
Zn
Cd
Hg
Ga
In
Tl
Ge
Sn
Pb
As
Sb
Bi
Ca
Sr
Ba
K
Rb
Cs
Pb - GKRR
H. Wu, et al., Scientific Data ‘16; H. Wu, et al., Comp. Mat. Sci ’17; H. Lu, et al., Comp. Mat. Sci ’19
1.0
1.5
Diffusio
2.0
2.5
Diffusio
2.0
2.5
3.0
3.5
4.0
DiffusionBarrier[eV]
Sc
Y
La
Ti
Zr
Hf
V
Nb
Ta
Cr
Mo
W
Mn
Tc
Re
Fe
Ru
Os
Co
Rh
Ir
Ni
Pd
Pt
Cu
Ag
Au
Zn
Cd
Hg
2.0
2.5
3.0
3.5
4.0
DiffusionBarrier[eV]
Sc
Y
La
Ti
Zr
Hf
V
Nb
Ta
Cr
Mo
W
Mn
Tc
Re
Fe
Ru
Os
Co
Rh
Ir
Ni
Pd
Pt
Cu
Ag
Au
Zn
Cd
Hg
Pd Pt
12. Automating Defect Analysis in Irradiated Materials
12
Database of labeled images
Trained
Deep Learning
Neural Network
New image Auto-labeled
image
14. Conclusions
Informatics is a transformative technology for
nearly everything – come join us!
Some experienced skunkwork leaders to talk to
14
Benjamin
Afflerbach
Lane
Schultz
Mingren
Shen
Ryan
Jacobs
Greg
Palmer