Skip to content

The project contains implementations of several primal-dual subgradient methods for searching traffic equilibria in Stable Dynamic and Beckmann models.

Notifications You must be signed in to change notification settings

ZiggerZZ/TransportNet

 
 

Repository files navigation

Stable Dynamic & Beckmann models

The project contains implementations of several primal-dual subgradient methods for searching traffic equilibria in Stable Dynamic and Beckmann models. Results of experiments on Anaheim transport network are included.

The following methods are implemented:

  1. Subgradient method with adaptive step size [arXiv:1604.08183]
  2. Universal gradient method [ref]
  3. Universal method of similar triangles [arXiv:1701.02473].

More information about models can be found in [Nesterov-de Palma] and [Beckmann]. Anaheim_Experiments.ipynb contains code of experiments on comparison of the above methods and Frank-Wolfe algorithm (only for Beckmann model).

Convergence process on 10 000 iterations for Stable Dynamic model:

Convergence process on 8000 iterations for Beckmann model (+ Frank-Wolfe algorithm):

About

The project contains implementations of several primal-dual subgradient methods for searching traffic equilibria in Stable Dynamic and Beckmann models.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 98.7%
  • Python 1.3%