latent manifold tuning model

The old name of the model is Poisson-Gaussian process latent variable model (P-GPLVM).

Description: Estimate latent variables and tuning curves from spike train using P-GPLVM.

As shown in Gaussian process based nonlinear latent structure discovery in multivariate spike train data Wu et al 2017.

• Launch matlab and cd into the directory containing the code (e.g. cd code/poisson-gplvm/).

• Examine the demo scripts for annotated example analyses of simulated datasets:

  • demo1_1DGP.m - Tutorial script illustrating P-GPLVM for 1-dimensional latent variable with tuning curves generated from 1D Gaussian Process.
  • demo2_1DBump.m - Tutorial script illustrating P-GPLVM for 1-dimensional latent variable with tuning curves generated from 1D Gaussian bumps.
  • demo3_2DBump.m - Tutorial script illustrating P-GPLVM for 2-dimensional latent variable with tuning curves generated from 2D Gaussian bumps.

demo_*_ref.m are the new demos with a reference implementation. The new implementation works for multi-trial data with aligned time points.