So today, we attended a big data conference for PIs in Alexandria, which is an upscale town outside of DC. The conference was held in a pretty new hotel, with its colorful, plush carpets and fancy bathroom stalls that close you off in your own private space. The event consisted of “lightning talks” where a bunch of researchers come up individually and give a quick rundown of their projects for a hot two minutes before the next researcher comes up. After these “lightning talks,” there is a poster session where you can visit any of the PI’s posters and learn more about their projects.
The PI conference was coordinated by Virginia Tech and my school, Johns Hopkins, which was exciting! There were three Hopkins faculty attending, one of who’s guest lecture I attended this past fall semester. His name is Daniel Robinson, and he talked about his work analyzing medical data organized in matrices to identify certain subspaces, which can yield insight in predicting when a patient might experience septic shock while in a hospital.
Besides meeting faculty from my school, I was really interested in a project involving machine learning and deep learning to identify bird species through audio recordings. This was a joint project involving Cornell and NYU. The Cornell PI I spoke to described the challenges that he faced when attempting to implement bird recognition invariance across factors that contribute to the recorded audio. In other words, it was difficult for him to “teach” the computer how to recognize birds regardless of the varying characteristics of the audio scene across different samples. He was able to solve this to a certain extent with deep learning, which involves a series of non-linear transformations, a linear algebra term. Additionally, a broad implication of this research is that it could widely expand field biologists’ ability to track bird migrations and populations by simply sampling audio in the wild.
I really enjoyed hearing about this project because the PI was engaging and his team’s application of machine learning/deep learning to solve a biological challenge was really cool to me. I highly recommend learning about other people’s work, even if it isn’t exactly what you’re trying to do. Especially while you’re in college, there are tons of opportunities to do this. You never know what new connections you might make, or what new interests you might uncover :).
Until next time, keep it real.