julia-point-process

The generation/estimation/validation of spike sequences with point process. All codes were written in Julia. These codes were used in a workshop on the point process of Student Association for Brain Science's training camp. We made them by referring to Dr. Shimazaki's implementation in Matlab. Detailed explanation is summarized in ganow's blog (written in Japanese only).

Setup

This repository uses the following dependencies:

• Julia v0.5.0
• Plots + PlotlyJS or PyPlot
• IJulia

After installing Julia, run the install script:

julia /path/to/install_dependencies.jl

Description

All implementation of functions are stored in julia/. Example scripts to execute these functions can be found in example/. These example scripts use sample dataset stored in data/, and the dataset can be generated by make_data.jl. images/ stores the results of the example scripts. The original Matlab codes written by Dr. Shimazaki is in matlab/ directory.

Generation

• julia/pprocess_inhomopoisson.jl: spike generation with inhomogeneous Poisson process.
• julia/pprocess_gamma.jl: spike generation with renewal gamma process.

Estimation

• julia/intensity_inhomopoisson.jl: estimation of the time-varying firing rate by Gaussian kernel regression.
• julia/sskernel.jl: estimation of the time-varying firing rate by Gaussian kernel regression with bandwidth optimization.
• julia/fitgamma.jl: maximum likelihood estimation of the parameters in gamma process.
• julia/intensity_gamma.jl: maximum likelihood estimation of the time-varying firing rate of gamma process.

Validation

• julia/QQplot.jl: validation of estimated firing rate by Quantile-Quantile plot.

Contributors

• Masanori Kawabata: mainly responsible for the generation part.
• Yoshihiro Nagano: mainly responsible for the estimation part.
• Akihiko Akao: mainly responsible for the validation part.