A Semantical Journey
Thursday, November 30, 2023, 11 AM
Bahen Centre for Information Technology, BA 3200
This lecture is open to the public.
No registration is required but space is limited.
Abstract:
In recent joint work with Ningning Xie, we propose an approach to choice-based learning via a combination of algebraic effect handlers and the selection monad. Choice operations are used in training programs written as effectful computations using choose and loss operations; optimization algorithms are implemented separately as choice operation effect handlers making use of losses.
In this talk, I look at the semantical journey which led me to this work. The journey began with Eugenio Moggi's 1989 notions of computation as monads. This inspired my work with John Power and others on the algebraic theory of effects. There was a problem in that theory: how to incorporate exception handlers. This led Matija Pretnar and myself (and now with many others) to algebraic effect handlers.
In a different direction, Carbune et al's Smart Choices approach to making machine learning a first-class citizen in programming languages gave Martín Abadi the idea that Martín Escardó and Paulo Oliva's selection monad could provide the right semantical framework for programs making smart choices. He and I worked on selection monad semantics for languages making optimal choices. However, optimization algorithms certainly need not make optimal choices, so one needed a way to make optimizing choices. Given the above history it then seemed natural, as hinted at above, to combine selection monad ideas and algebraic effect handlers, thus bringing my semantic journey to the present, with Ningning's and my joint research program on languages for choice-based learning.
Bio:
Professor Gordon Plotkin has been a faculty member at Edinburgh University since 1973. His main interests are in the connections between programming languages and logic, with particular interest in the semantics of programming languages. He has also worked a little on natural language and, more extensively, on systems biology. He may be best known for his invention of structural operational semantics. He is a fellow of the Royal Society and, recently, a member of the American Academy of Arts and Sciences; he is a recipient of the ACM Programming Languages Achievement Award and the EATCS Distinguished Achievements Award. He currently remains on the Edinburgh faculty and is also a Senior Staff Research Scientist at Google DeepMind.