Synopsis

Uses a gradient descent (via a fake dynamic equation) to move nodes to minimize the mesh-energy E, without relying on the random motions of /adaptive_mesh/anneal. The nodes move in the direction that minimizes E, and unlike /adaptive_mesh/anneal, they all move at once.

On each iteration, ppm2oof numerically computes the gradient of E with respect to each node's position. It computes the time step that will make the fastest node move by a prescribed amount. Then it updates each node's positions by taking one Euler step of the equation ∂r_i/∂t = -∇_iE, where r_i is the position of node i and ∇_i is the gradient with respect to r_i.

This is an experimental technique that doesn't work very well. It can be slow, unreliable, and vindictive.

Arguments