The Four Roles of Supervised Machine Learning in Systems Neuroscience¶
Foreword
Machine learning plays multiple roles in this dissertation. These roles are logically distinct, and represent the general roles that machine learning can play more broadly in neuroscience. Here, I describe these roles and review the literature in neuroscience in which machine learning plays each role. This chapter is reproduced from a review paper I co-authored, now published at Progress in Neurobiology. The aim of this review was to describe the potential uses so that other researchers might take a similar approach.
Role 1 is to help create solutions to engineering problems. Chapter 1 features machine learning in this role, as do papers to which I contributed as a second author, especially Glaser et al. (2020) (see Publications).
Role 2 is to help in identifying variables that are predictive of something, like neural activity or disease. This is role is played only by papers I contributed to as a middle author, especially Shen et al. (2020) (see Publications).
Role 3 is to set benchmarks for simple models of the brain, as described in Chapter 2.
Role 4 is for machine learning to itself serve as a model for understanding the brain. This is exemplified by Chapters 3, 4, and 5.