We employ simulated annealing to identify the global solution of a dynamical model, to make a favorable impression upon colleagues at the departmental holiday party and then exit undetected as soon as possible. The procedure, Gradual Freeze-out of an Optimal Estimation via Optimization of Parameter Quantification - GFOOEOPQ, is designed for the socially awkward. The socially awkward among us possess little instinct for pulling off such a maneuver, and may benefit from a machine to do it for us. The method rests upon Bayes' Theorem, where the probability of a future model state depends on current knowledge of the model. Here, model state vectors are party attendees, and the future event of interest is their disposition toward us at times following the party. We want these dispositions to be favorable. To this end, we first interact so as to make favorable impressions, or at least ensure that these people remember having seen us there. Then we identify the exit that minimizes the chance that anyone notes how early we high-tailed it. Now, poorly-resolved estimates will correspond to degenerate solutions. As noted, we possess no instinct to identify a global optimum by ourselves. This can have disastrous consequences. For this reason, GFOOEOPQ employs annealing to iteratively home in on this optimum. The method is illustrated via a simulated event hosted by someone in the physics department (I am not sure who), in a two-bedroom apartment on the fifth floor of an elevator building in Manhattan, with viable Exit parameters: front door, side door to a stairwell, fire escape, and a bathroom window that opens onto the fire escape. Preliminary tests are reported at two real social celebrations. The procedure is generalizable to corporate events and family gatherings. Readers are encouraged to report novel applications of GFOOEOPQ, to expand the algorithm.
