To learn more about Sarah's research, click the image above to see a two-minute video.
Joining such luminaries as Alan Turing, Bill Joy, and One Direction, the work of Sarah Sachs, an undergraduate in Brown University's Department of Computer Science (Brown CS), has been featured in Wired, the Today Show, and other publications including Wonk Blog/Washington Post, MIT Technology Review, and PetaPixel. They've all taken an interest in her recent summer research ("A Century of Portraits: A Visual Historical Record of American High School Yearbooks"), a collaboration with Shiry Ginosar, Kate Rakelly, Brian Yin, and Alexei A. Efros of University of California, Berkeley.
In their paper, the researchers turn to a dataset of 37,021 frontal-facing American high school yearbook photos in the hope of capturing "visual culture" not often addressed by historians because of its mundane or abstract nature. By using computation, they look at a historical visual record too large to be evaluated manually, using portraits as a constant visual frame of reference with varying content. From this, they can consider issues such as fashion trends and social norms and how they've changed over time.
"We demonstrate," they explain, "that our historical image dataset may be used together with weakly-supervised datadriven techniques to perform scalable historical analysis of large image corpora with minimal human effort, much in the same way that large text corpora together with natural language processing revolutionized historians’ workflow. Furthermore, we demonstrate the use of our dataset in dating grayscale portraits using deep learning methods."
"Sarah was amazing," says Shiry Ginosar, the paper's first author. "She took charge of collecting our data and making sure we had a wide range of demographics covered. She did a comprehensive literature review on how others have analyzed smiling in photos. She suggested looking at smiles based on what she had seen in previous research. She noticed this was great place for computer vision to start tackling some of the inefficiencies in previous social science, such as manual smile annotation. We would be lost without her!"
Sachs credits her involvement in machine learning to the positive interactions that she had with her APMA 1650 Professor, Caroline Klivans. "She really opened my eyes to power of big data," says Sarah. "She taught me how rewarding it is to verify claims and presumptions with math and computer science. When she became my CS022 professor, Prof. Klivans encouraged me to work somewhere where I could contribute and learn. She helped me sift through all opportunities and find a good fit. I had grown up in the suburbs of Berkeley, so a position at UC Berkeley was ideal because I could live at home and still work at a great institution."
What was it like to have an opportunity like this as an undergraduate?
"I started CS later in my college career," Sarah says, "so this summer opportunity came directly after my intro-sequence. It was intimidating at first to collaborate with PhD candidates when I knew so little in comparison. But Brown taught me how to ask the right questions. I had experience going to TAs and professors to ask for help, so I quickly gained my footing by treating it like any CS class at Brown: stay confident in your progress and never be afraid to ask for help. By the middle of the summer, I felt like an equal contributor and I was so excited by the progress our team was making. Our project was exciting because I got research other social sciences and use the skills I've learned outside CS to do research on many different topics. At the same time, I got to think critically about statistical concepts and computational complexity. I felt like I was finally embracing my Brown education. I was taking my experience in many different fields to do research in a unique and untouched intersection."
After her summer work, Sarah explains that she realized how much more there was to learn about machine learning. She took additional Brown CS classes and even served as a TA for CS 142 Machine Learning and Pattern Recognition. She's currently conducting research with Associate Professor of Engineering and Computer Science Pedro Felzenszwalb.
Other interesting opportunities, this time in the industry, lie just ahead: "After graduation, I'm going to work at Google to do location-based machine learning. I certainly plan on going to graduate school eventually, but I'm excited to learn more about how people use ML outside of academia."