Cascade Processes for Sparse Machine Learning

Abstract

What could seemingly distinct phenomena like a financial crisis, a global pandemic, or the sudden increase of wheat prizes have in common? On an abstract level, they all can emerge as result of a cascade process that evolves on a complex network of interdependent actors. Another well known example of a cascade process, yet, largely disregarded in the context of systemic risk analyses, is the evaluation of a deep neural network, i.e., the work horse of the ongoing revolution by modern machine learning and AI. We will discuss how fundamental insights into the nature of cascade dynamics cannot only guide the design of more robust systems but also inspire us to build efficient AI.