torus_solver.optimize

class torus_solver.optimize.SourceParams(theta_src, phi_src, currents_raw)

Bases: object

Electrode locations and injected currents (unconstrained).

Parameters:

• theta_src (Array)

• phi_src (Array)

• currents_raw (Array)

currents_raw: Array

phi_src: Array

theta_src: Array

tree_flatten()

classmethod tree_unflatten(aux, children)

torus_solver.optimize.enforce_net_zero(currents_raw)

Project currents onto the Σ I_i = 0 subspace.

Return type:

Array

Parameters:

currents_raw (Array)

torus_solver.optimize.forward_B(surface, params, *, eval_points, sigma_theta, sigma_phi, sigma_s=1.0, current_scale=1.0, tol=1e-10, maxiter=2000, biot_savart_eps=1e-09)

Compute B at eval_points from electrode params on the torus surface.

Return type:

Array

Parameters:

• surface (TorusSurface)

• params (SourceParams)

• eval_points (Array)

• sigma_theta (float)

• sigma_phi (float)

• sigma_s (float)

• current_scale (float)

• tol (float)

• maxiter (int)

• biot_savart_eps (float)

torus_solver.optimize.make_helical_axis_points(*, R_axis, n_points, dtype=<class 'jax.numpy.float64'>)

Convenience: points on the torus magnetic axis (major circle) + basis.

Return type:

Tuple[Array, Array, Array, Array]

Parameters:

• R_axis (float)

• n_points (int)

torus_solver.optimize.optimize_sources(surface, *, init, eval_points, B_target, B_scale=1.0, sigma_theta, sigma_phi, n_steps, lr, reg_currents=1e-06, reg_positions=0.0, callback=None, return_history=False)

Optimize electrode params to match a target B field at given points.

Return type:

SourceParams | tuple[SourceParams, Dict[str, list[float]]]

Parameters:

• surface (TorusSurface)

• init (SourceParams)

• eval_points (Array)

• B_target (Array)

• B_scale (float)

• sigma_theta (float)

• sigma_phi (float)

• n_steps (int)

• lr (float)

• reg_currents (float)

• reg_positions (float)

• callback (Callable[[int, Dict[str, float]], None] | None)

• return_history (bool)

torus_solver.optimize.optimize_sources_lbfgs(surface, *, init, eval_points, B_target, B_scale=1.0, sigma_theta, sigma_phi, maxiter, tol=1e-09, reg_currents=1e-06, reg_positions=0.0, callback=None, return_history=False)

Optimize electrode params using L-BFGS (often faster than Adam for small problems).

Return type:

SourceParams | tuple[SourceParams, Dict[str, list[float]]]

Parameters:

• surface (TorusSurface)

• init (SourceParams)

• eval_points (Array)

• B_target (Array)

• B_scale (float)

• sigma_theta (float)

• sigma_phi (float)

• maxiter (int)

• tol (float)

• reg_currents (float)

• reg_positions (float)

• callback (Callable[[int, Dict[str, float]], None] | None)

• return_history (bool)

torus_solver.optimize.surface_solution(surface, params, *, sigma_theta, sigma_phi, sigma_s=1.0, current_scale=1.0, tol=1e-10, maxiter=2000, use_preconditioner=False)

Return (currents, source_density, potential, surface_current).

Return type:

tuple[Array, Array, Array, Array]

Parameters:

• surface (TorusSurface)

• params (SourceParams)

• sigma_theta (float)

• sigma_phi (float)

• sigma_s (float)

• current_scale (float)

• tol (float)

• maxiter (int)

• use_preconditioner (bool)