Tracking and distinguishing slime mold solutions to traveling salesperson problems through synchronized amplification in the non-equilibrium steady state
The plasmodium of the true slime mold Physarum polycephalum-an ancient, unicellular, aneural organism-serves as a platform for studying the information-processing capacities of active matter. Previous experiments used Physarum's intricate morphological dynamics and photoavoidance in stellate chips to solve $N$-city traveling salesperson problems (TSPs) of up to eight cities, scaling linearly in time with TSP size. Optical feedback controlled by a modified Hopfield network illuminated specific lanes at regular intervals, prompting Physarum to elongate or retract selected branches. When the illumination pattern stabilized in a non-equilibrium steady state, branches bifurcated reproducibly into solution and non-solution groups, with the former exhibiting lower-frequency, higher-amplitude, and more synchronized oscillations than the latter across 41 trials with valid TSP solutions. Physarum's synchronization dynamics efficiently predict 100% of selected solutions by the midpoint of the optical-feedback interval, achieving statistically significant (paired t-test, $p<0.005$) discrimination from alternate tours well before the non-equilibrium steady state. Observed frequency downconversions and synchronized power amplifications scale linearly and quadratically, respectively, for small-to-moderate TSP size, as captured by a toy model of energy redistribution with saturating optical absorption. Tuning these features in native biomolecular chromophore networks may thus improve both the quality and efficiency of TSP solutions from Physarum-based biocomputers, which exploit the effects of organismal-scale coherence.