STA 290 Seminar: John Miles White (Facebook)

Statistics Colloquium: STA 290

Thursday, February 27th at 4:10pm, MSB 1147 (Colloquium Room)
Refreshments at 3:30pm in MSB 4110 (4th floor lounge)

Speaker: John Miles White, Facebook Inc.

Title: "Julia and statistical computing"

Abstract: Julia is a new language for technical computing. The language is designed to solve the "two language problem", in which scientists prototype code in a higher-level language like R and then rewrite parts (or all) of their code in a lower-level language like C. Julia strives to expose a set of basic abstractions that allow programmers to transition easily between quick-and-dirty prototype code and production-quality code.

In my talk, I'll describe Julia's current suitability for statistical computing. I'll provide a survey of both Julia's strengths and weaknesses, including a review of Julia's emerging statistical libraries. For advanced users with an interest in implementing models from scratch, Julia is already a useful tool. For users who depend upon mature statistical libraries to get their work done, Julia is another year or two away from being ready-for-use.

I will also review several core concepts from programming language theory that provide the necessary tools for reasoning about the ways in which Julia differs from R: this includes a review of type systems, static analysis and scoping rules. Understanding these concepts enables programmers coming from R to understand both why Julia may feel quite unfamiliar and why Julia is able to provide large performance improvements over traditional higher-level languages.