WG211/M15Wiklicky

On Framework for Quantitative Program Synthesis

Arguably most work on the problem of program synthesis is based on various models based in discrete structures, e.g. related to model checking, game theoretic models, combinatorial optimisation, etc. In this talk we aim in recasting program synthesis as a non-linear, continuous optimisation problem. This allows among other things for a smoother integration of non-functional constraints. Initial experiments demonstrate that, maybe surprisingly, it is possible to avoid algebraic reasoning for algebraic problems and replace it entirely by continuous optimisation constraints.

———

Dr. Herbert Wiklicky holds Master (MSc) degrees in theoretical physics, mathematics, and computer science and a PhD in computer science from the University and TU Vienna. He held positions in Vienna, Amsterdam, Tokyo, Mannheim and London where his work concentrated on various topics in theoretical computer science in particular in semantics (of concurrency) and program analysis. Since 2001 he is at the Department of Computing at Imperial College London (most recently as Reader in Computer Science). His main research interests are in models, in particular, of probabilistic computation, expressiveness of languages, computer security and general quantitative approaches in computation. Together with Dr. Alessandra Di Pierro (Pisa, Verona) his research focused on quantitative program analysis, which lead e.g to the development of Probabilistic Abstract Interpretation (PAI) based on Linear Operator Semantics (LOS).