Learning Paths

Learning Paths#

brainmass serves three kinds of users. Each persona below is a signposted route through the documentation — follow the one that matches your goal. The paths overlap by design: a beginner who collects data becomes a researcher; a researcher who extends a model becomes a modeler.

Beginner#

New to neural mass models and/or brainmass. You want to run a first simulation and build the mental model.

  1. Installation — install brainmass for CPU, GPU, or TPU.

  2. Quickstart — a five-minute first simulation with the Simulator and a one-line Fitter teaser.

  3. Key Concepts — the mental model: Step-model, Simulator, Network, Fitter, and units.

  4. Your First Simulation, Models and Dynamics, Noise and Stochastic Runs, Building a Network — the core tutorial sequence.

  5. Gallery — browse the model zoo to see each model family in action.

Researcher#

You have empirical data (EEG / MEG / fMRI) and want to fit models to it, map activity to signals, and analyze the results.

  1. Quickstart — get oriented with the orchestration layer.

  2. Forward Models — turn neural activity into BOLD / EEG / MEG signals.

  3. Fitting with Gradients and Gradient-Free Fitting — fit models to data with gradient-based and gradient-free backends.

  4. Analyze Results (FC / FCD / spectra) — functional connectivity, FCD, and power spectra.

  5. Run Parameter Sweeps — sweep parameters efficiently with vmap.

  6. Gallery — work through the case studies for end-to-end examples.

Modeler#

You build and extend models, couplings, and objectives, and run differentiable / data-driven workflows.

  1. Why Differentiable? and Architecture Overview — the narrative and the package architecture.

  2. Custom Coupling, Compose a Custom Objective, and Batch and Accelerate — extend the building blocks and make them fast.

  3. Developer Guide — the developer guides, including the data-driven workflow extension playbook.