In perturbation based nonlinearity compensation (PB-NLC) quantization of perturbation coefficients is employed for reduction of computational and implementation complexity. In this work, we propose and experimentally verified adoption of decision directed least mean square (DD-LMS) algorithm for optimization and complexity reduction of PB-NLC. We show that for 32 GBaud dual polarization (DP) 16 QAM after 2560 km of standard single mode fiber (SMF) propagation, proposed scheme reduce the computational term by 46% compared to that of conventional PB-NLC with uniform quantization of perturbation coefficients for the same Q improvement.