aequilibrae.paths.linear_approximation#

Classes

LinearApproximation(assig_spec, algorithm[, ...])

class aequilibrae.paths.linear_approximation.LinearApproximation(assig_spec, algorithm, project=None)[source]#

calculate_stepsize()[source]#: Calculate optimal stepsize in descent direction

check_convergence()[source]#

Calculate relative gap and return True if it is smaller than desired precision.

Two relative gaps are computed and stored on the instance:

self.rgap - the AequilibraE convention, |Σ flow·cost − Σ AON·cost| / Σ flow·cost. This is the only quantity used for the stopping criterion (compared against self.rgap_target). (Σ flow·cost − Σ direction·cost) / Σ flow·cost, where direction is the BFW combined step direction (self.step_direction_flow). Reported alongside self.rgap in the iteration log and the convergence report so the two measures can be compared. NOT used for stopping.