The Art of Parameterization

One of the most remarkable aspects of data science is the ability to quantify observations of the world around us and, by using analytic and machine learning approaches, infer physical laws and make predictions about the future. In this talk, I will draw on my experiences, both from my time in industry as well as from running a neuroscience research lab at UC San Diego, to examine the art of parameterization: turning large databases of observations into human-level understanding and knowledge discovery. I will specifically focus on lessons learned in industry, and how we are adopting those in our project to aggregate multiple large, heterogeneous neuroscience databases. These data include textual data mined from the millions of peer-reviewed neuroscience publications, functional and structural brain imaging data, neural electrophysiology, and gene expression profiles. Through such aggregation and integration we can mine the integrated data to find missing links and gaps in our knowledge; we can algorithmically generate plausible hypotheses critical to clinical neurology and mental health.