T. Lucas Makinen: creating a neural network to study the cosmos

Monday, Aug 17, 2020
by Sharon Adarlo

T. Lucas Makinen, 22, Class of 2020


Makinen majored in astrophysics and earned two certificates: the Undergraduate Certificate Program in Statistics and Machine Learning from the Center for Statistics and Machine Learning (CSML), and the Program in Applied and Computational Mathematics.


Makinen’s independent project for his CSML certificate explored the intersection between observational cosmology and machine learning, specifically in the task of separating contaminating signals from the Milky Way galaxy from signals in the wider universe.

“Cosmologists are interested in studying how matter is distributed on the universe’s largest scales. A good tracer of this distribution is a faint radio emission produced by neutral hydrogen, whose evolution can be traced over cosmic time using the cosmological redshift—in theory at least,” said Makinen. “Much stronger radio signals, mostly from our galaxy, dump a huge amount of noise on top of that interesting, weaker signal beyond. These noisy foregrounds have been forecast as a big problem for the radio cosmology community for a while now.”

Older methods to separate these contaminants are not physically precise and unreliable, and may over-subtract from the cosmological signal underneath, Makinen said.

“Current proposals for removing foreground contaminants rely on blind statistical methods which draw no reference to physical models and suffer from a large variance in accuracy at different redshifts and angular scales,” said Makinen.

To address this problem, Makinen developed a neural network that would learn how to separate foreground patterns (the Milky Way’s signals) from cosmological data that researchers are trying to study. This neural network was trained on simulated data and learned to pick out foreground signals that older methods have missed, even when exposed to variable observational noise.

“The analysis demonstrates the utility of robust, simulation-driven deep learning in the observational sciences, and paves the way for realizable cosmological inference from raw intensity maps for upcoming radio experiments,” Makinen said.

During his studies at Princeton, Makinen found his CSML classes to be helpful, especially in their interdisciplinary bent. In SML 515 – Statistical Data Analysis taught by Peter Melchior – an assistant professor jointly appointed in Astrophysical Sciences and CSML - Makinen encountered assignments that involved economics and public health, among other subjects.

“It was really cool,” said Makinen, whose thesis advisor was also Melchior. “We took the same tools I was using in astrophysics and applied them to a whole range of different problems in other fields, including our final project on pandemic response.”

During his time at Princeton, Makinen served as an intern at the Simons Foundation Flatiron Institute in New York City. He was also a 2019 Streicker International Fellowship recipient, which he used to complete his junior paper at Imperial College London and Cambridge University last summer. In this role, he used Bayesian statistical methods to study supernova data to measure cosmic expansion and Dark Energy.

In addition, Makinen attended the Royal College of Music (RCM) from 2018 to 2019 as an exchange student studying the trumpet. “This is actually where my interest in astrostatistics began—I took music courses at RCM but did my junior paper just down the block with my astrophysics advisor at Imperial College. I’m really thankful I had the opportunity to study both of my interests in such a vibrant city.”

After graduation, Makinen served as a summer intern in computational cosmology research at the University of Copenhagen. There, Makinen worked on a Bayesian inference project attempting to resolve tension in the measurement of cosmological expansion.

Then in the fall, Makinen will enroll at Sorbonne Université in Paris to obtain a master’s degree in physics. In the future, Makinen said he is interested in doing academic research in the intersection of statistics and astrophysics.

Extracurricular Activities:

Makinen was the principal trumpet player in the Princeton University Orchestra and also played in the University’s Jazz Ensemble. He was also involved as a DJ for the campus radio station, WPRB 103.3 FM. He also worked as a student Global Ambassador for the Office of International Programs to promote work and study abroad programs on campus.

For Fun:

Makinen enjoys listening to and recording music, collecting records, traveling, and learning new languages. Besides English, he grew up speaking Finnish and knows French and a little Italian.