Niklas Smedemark-Margulies
PhD Student
Education
- BA in Neuroscience, Amherst College
- MS in Immunology, Harvard Medical School
Biography
About Me
- Hometown: Boston, MA
- Field of Study: Computer Vision, Representation Learning
- PhD Advisor: Jan-Willem van de Meent and Deniz Erdogmus
What are the specifics of your graduate education (thus far)?
I received my BA in Neuroscience from Amherst College, where I studied the use of light-sensitive proteins to measure and stimulate neurons. I then worked as a Research Fellow at Harvard University designing a protein sensor of membrane voltage. I received my MMSc in Immunology from Harvard Medical School where I learned how to evaluate genetic mutations in clinical care settings. After that, I spent several years as a software engineer working on the analysis of genomic data.
What’s one problem you’d like to solve with your research/work?
Given a previously learned skill or behavior, humans can easily adapt to a new domain using very few examples. For example, learning a third language is much easier than learning a second. Neural networks usually lack this adaptive capability and must be trained from scratch on each task. In my research, I would like to design neural net architectures that learn abstract representations and can perform this type of transfer learning.
What are your research/career goals, going forward?
I would like to work on modeling and understanding information processing in the brain at the level of single neurons. This low-level modeling may help us design learning algorithms that are as data efficient and robust to noise as biological systems and may one day help us understand human cognition. In addition, I want to continue making all my research open-source and easily reproducible.