The Elastic Templates algorithm uses simulated mean-field annealing and gradient descent to find near-optimal solutions of problems which are both combinatorial and continuous. For instance, in high-energy physics experiments it has been applied to the tracking problem, which is to minimize the distances between track templates and the hits belonging on each track. I will explain recent attempts to extend and improve this approach for reconstructing simulated data from the D0 experiment. Vertex templates have been added, and the algorithm simultaneously minimizes the distances between the track templates and the vertex assigned to each track.

A weight for each template is used in the algorithm, which represents the certainty that the template is really needed. Error matrices for the hits are used properly. Also, dramatic performance increases have been achieved by modifying the mean-field algorithm to take advantage of STL vectors.