Model Fitting

braintools.fit fits any JAX or brainstate model to data. You give it a pure simulate(params) function or a module with brainstate.nn.Param parameters, an objective built from fit.losses, and a search space. A Fitter then minimises the objective from many starting points at once, by gradient (all starts vmapped), SciPy or Nevergrad.

What you’ll find here

• Fitting models: simulating trajectories with quad.odeint and quad.solve_ivp; composing objectives; a multi-start fit of a Hopf oscillator; why several starts are needed; mapping the loss landscape with fit.sweep; confidence intervals from the Hessian; fitting the couplings of a Wilson–Cowan brainstate module; differentiable losses for spiking data; and saving results to HDF5.

For the full list of functions and parameters, see the API reference.

• Fitting models to data