Theoretical Neuroscience 2.0: Filling the data/model gap
"Theoretical Neuroscience is the modelling-oriented approach to understanding brain and behaviour. After presenting a utilitarian definition of models, I argue that theoretical and conventional neuroscience developed concurrently, and that the distinction is at best superficial. I then discuss the two extremes in neuroscientific practice: quantitative data analysis, mainly backed up by statistical tools, and qualitative computational modelling, primarily relying on dynamical systems theory. I then present a general framework to bridge the statistical and dynamic approaches, which allows for a data-driven view on computational modelling. I finally demonstrate the advantages of this framework through a model-driven method for the identification of computationally relevant events in multivariate spike count time series. These time series are often highly nonstationary, where statistical moments, such as firing rates, vary to potentially reflect information regarding neural computations, such as learning or the accumulation of sensory evidence."