Specify the experiment

Step 2 of the TVB-O spine: declare dynamics, network, coupling, observations, figures and analyses as one typed document.

A TVB-O experiment is a document, not a script. You declare what the system is — its dynamics, the network it runs on, how nodes couple, what is observed, what is fitted, what is plotted — and nothing about how to integrate it or which library to call. That separation is the whole point: the same document runs on JAX, TVB, Julia, C or Brian2, and it stays readable years after the run.

This is the longest step, because it is where all the modelling decisions live. Nothing else in the spine asks you to make a choice.

The object itself

The running example

One experiment, grown field by field across the whole guide.

The SimulationExperiment

Every top-level field, and which ones are required.

Writing TVB-O YAML

The YAML dialect: IRIs, includes, anchors and the tvbo_class envelope.

What the nodes do

Dynamical systems

Reuse a curated model, or write your own equations.

Events & stimulation

Drive the system with a stimulus that is itself part of the spec.

Solving a network of DDEs

Integrators, step size, and what delays cost.

What connects them

Networks & connectomes

Where the connectome comes from, how regions couple, and how they differ.

What comes out

Observation models

BOLD, EEG, MEG — the forward model that turns states into a signal.

Declared post-processing that travels with the study.

Specify a figure

A figure as data: a mosaic of panels bound to result fields.

Beyond one run

Parameter exploration

Sweep an axis instead of running one point.

Fitting, inference & optimization

Fit parameters to data, with the loss declared alongside them.

Linked experiments in a SimulationStudy

Many experiments, one study, shared figures.

Fan-out over subjects and phenotypes

One recipe across a cohort.

Calling your own code

The escape hatch, when the grammar cannot say it.

Next

A finished spec is compiled, not interpreted: ③ Generate.