maedoc/autotvb/small-model-essentials

TVB Essentials for Small Models

Critical API Facts — GET THESE RIGHT OR CRASH

1. `sim.run()` returns a LIST, not a generator. Always unpack as:

``python (t1, d1), (t2, d2) = sim.run(simulation_length=T) ``

1. Conduction speed goes on Connectivity, NOT Simulator:

``python conn.speed = numpy.array([4.0]) conn.configure() ` Never pass conduction_speed to Simulator()`.

1. Always `configure()` objects before using them:

``python conn.configure() model = models.Generic2dOscillator() sim.configure() ``

1. Multi-monitor unpacking: one (time, data) pair per monitor:

``python mon_raw = monitors.Raw() mon_tavg = monitors.TemporalAverage() (t_raw, d_raw), (t_tavg, d_tavg) = sim.run(...) ``

Minimum Simulation Durations

Notebook Serialization

In .ipynb JSON, code cell sources use array format with real newlines:

"source": ["x = 1\n", "y = 2\n"]

Never use literal \n in a single string: "x = 1\ny = 2" — this causes SyntaxError.

Additional Rules

• Use deterministic integrators (HeunDeterministic) unless noise requested
• nsig shape must match model's state variables: numpy.array([v, v]) for 2-variable models
• Stimulus weights: use explicit indexing stim_weights[[35, 36], 0] = numpy.array([v1, v2])
• Burn-in must not swallow stimulus: max 100 ms pre-stimulus discard
• Focused analysis: only plots/data the goal explicitly asks for