BrainRBMjulia

BrainRBMjulia builds on the RestrictedBoltzmannMachines.jl implementation and adapts it for whole-brain zebrafish calcium imaging data. The package wraps core RBM routines, adds domain-specific utilities, and provides custom visualizations for analyzing neural activity.

Repository structure

The codebase is organized into two main folders:

• src/ – Core RBM functionality and utilities.
  • rbms.jl defines model types and constructors.
  • train.jl, generate.jl, and state_sampling.jl handle model training and data generation.
  • maps.jl and statistics.jl implement evaluation metrics and map-building routines.
  • saving.jl and extendables.jl offer serialization helpers and callbacks.
  • utils/ contains helper modules for binary encodings, coupling approximations, dataset handling, error metrics, reordering, and voxel manipulation.
• ext/ – Makie extensions and plotting recipes.
  • registers Makie recipes for brain projections, correlation heatmaps, RBM diagrams, and other visual diagnostics.

An example/ directory illustrates how to train models and visualize results.

Usage

This package was developped for Julia 1.11.

Add the package in a Julia environment and load it with:

using Pkg
Pkg.add(path="/path/to/BrainRBMjulia")
using BrainRBMjulia

Most workflows revolve around building BrainRBM models, training on calcium imaging datasets, and using the Makie extensions for visual exploration.