AI@MH seminar

In daily life, randomness is commonly perceived as something negative that should be avoided at all costs. In machine learning, however, randomness is frequently embraced as a way to form predictive models with unique competitive advantages (e.g., minimal training costs). In this talk, we will look at a family of computing models that combine randomness and algebraic operations to impose structure upon the randomized representations. Despite sounding surreal there are solid theoretical arguments why combining randomness and structure could lead to interesting outcomes. While looking at the illustrative examples, we will try to consider possible use cases of such an approach for processing biological data (in a broad sense).