Table 5.

Comparison of convergence trajectories.

Training round	Dynamic weight loss values	Fixed weight loss value	Key event
0	2.35	2.35	Initial state
200	1.07	1.52	First weight adjustment (4 nodes down)
600	0.62	0.89	Channel interference events (15% packet loss)
1000	0.33	0.65	High quality node discovery (α > 0.8)
1500	0.09	0.41	The system enters a steady state
2000	0.03	0.38	–