Factorization -Learning Initialization for Parameter Optimization in Cellular Networks

Table 2

Comparison of performance.

Algorithms	Factorization	Convergence episodes		Average episode reward
		-Greedy	Boltzmann	-Greedy	Boltzmann
-learning	Original	54.6	367.8	-12.14	-0.590
	Factorized	480.7	774.1	8.514	11.91
Dyna -learning [12]	Original	64.9	357.4	-10.26	-1.806
	Factorized	471.5	702.6	9.246	13.16
-learning [13]	Original	57.1	374	-11.90	0.018
	Factorized	464.1	718.6	7.253	10.42
Double -learning [14]	Original	56.8	340.6	-12.60	-0.510
	Factorized	479	824.8	9.393	14.25
Speedy -learning [15]	Original	68.7	365.6	-8.884	1.151
	Factorized	522	699.1	14.39	14.11