traveler.steps.simple_choice.analytic_share_loss_grad#

simple_choice.analytic_share_loss_grad(store, *, choosers=None, param=None, target_shares, scale=1.0)[source]#

Differentiate analytic-share loss with respect to parameters.

Parameters:

  • store (AbstractStore) – Store containing chooser and supporting model data.

  • choosers (WorkingTable or None, optional) – Choosers to evaluate. When omitted, use the configured table.

  • param (dict[str, float] or None, optional) – Utility coefficients. When omitted, use the model’s parameters.

  • target_shares (dict or jax.Array) – Target share for each configured alternative.

  • scale (float, optional) – Multiplier applied to the summed squared error.

Returns:

Gradient pytree with the same structure as param.

Return type:

Any