Ekaterina Komendantskaya

"Continuous Verification of Machine Learning: a Declarative Programming Approach"

I will discuss state of the art in neural network verification, and in particular I will focus on the trend known as "continuous verification". Continuous verification is a family of methods that explore the ways to adapt (usually discrete) verification methods to machine learning algorithms that work with continuous functions. I will argue that continuous verification must rely on robust programming language infrastructure (refinement types, automated proving, type-driven program synthesis). This provides a major opportunity for the declarative programming language community to join in the AI verification game.

Time: Thursday, 11.03.2021, 13:00

Termin als iCAL Datei downloaden und in den Kalender importieren.