The GenKL algorithm (α = 2.42) outperformed the
Gaussian, and performed similarly well as the J algorithm.
Performance of each algorithm was indicated by the similarity between the intact
and deafferented muscle synergies, quantified by the best-match scalar product
averaged across the synergy set of each frog (3 synergies for jump, and 4
synergies for swim). A, For jump, in all four frogs performance of the
GenKL algorithm was comparable to that of the
J in finding synergies that persisted after
deafferentation, one of the best-performing algorithms in our previous study
(Devarajan & Cheung, 2014).
B, For swim, in three out of four frogs GenKL performed
comparably well or even better than J.