The Nonlinear Input Model (NIM) -- updated information Sept 2019

The Nonlinear Input Model (NIM) is a statistical model for describing nonlinear computation in sensory neurons. It is in form of an LNLN (linear-nonlinear x2) cascade, with its predicted firing rate given as a sum over nonlinear inputs followed by a 'spiking nonlinearity':

NIM Schematic

Full Details

The Rectified Latent Variable Model (RLVM)

The Rectified Latent Variable Model (RLVM) is a probabilistic model for describing the activity of a large population of neurons based on a much smaller set of inputs (i.e., latent variables). Key elements of the model that distinguish it from other approaches are the constraint that the latent variables be non-negative (e.g., like neural activity), and a lack of other constraints (i.e., need not be uncorrelated, independent, Gaussian-distributed, etc.):

[Preprint] [Code on Github]

The Separable Nonlinear Input Model (sNIM) for detailed spatiotemporal characterization in retina

The sNIM is a version of the NIM (see above) that uses spatiotemporal filters comprised of space-time-separable elements. This allows for detailed spatiotemporal characterization of spatial and temporal sensitivity for neurons with receptive fields where this approximation is valid, such as in the retina:

[Paper] [Code on Github]

Model-Based Eye Tracking

An algorithm that uses probabilistic models (see NIM) of the stimulus processing of visual cortical neurons to infer an animal's eye position from the spiking activity of a recorded neural population.

Full Details