Explaining the "Black Box"
As we build more complex so-called "black box" models and architectures, we run the risk of not understanding how our systems might behave. Aside from the obvious ethical implications, this might also have legal repercussions we want to avoid. In this talk, we will review old and new techniques for introspecting the logic and structure of machine learning models and architectures. We will discuss possible tradeoffs between transparency and accuracy and offer suggestions for building accountable, reproducible models you can interpret and explain.
Katharine Jarmul is a pythonista and founder of Kjamistan, a data science and machine learning consulting company in Berlin, Germany. She's been using Python since 2008 to solve and create problems. She helped form the first PyLadies chapter in Los Angeles in 2010, and co-authored an O'Reilly book along with several video courses on Python and data. She enjoys following the latest developments in machine learning, natural language processing and workflow automation infrastructure and is generally chatty and crabby on Twitter, where you can keep up with her latest shenanigans (@kjam).