PhD Candidacy Exam: The Demise and Resurrection of Analog Computing

Until the 1970s, analog computers were anticipated to coexist with digital computers, participating in computations that were suited to the continuous model of computation, compared to the discrete one of digital computers. The past 50 years have proven those predictions wrong: the digital von Neumann architecture has become synonymous with computing machinery. Recently, however, voltage scaling has ended in digital computers, and new workloads have reignited interest in analog computers. A rigorous approach to understanding how analog computers may support modern workloads, and where analog computers excel in relation to existing machines, is needed. In this candidacy syllabus, I begin by examining techniques for characterizing workloads, necessary for identifying problems suitable for analog computation. I will examine some of the seminal work in parallel CPUs, GPUs, and accelerators that have given digital computers high scalability and programability. Finally I will study contemporary approaches to analog computing, in particular from the perspective of neural networks.