An Accelerated Fixed-Point Algorithm Applied to Quadratic Convex Separable Knapsack Problems

Table 3

Correlated test for state-of-the-art algorithms.

Dimension	Iterations			Time (sec)			Error
	Avg	Max	Min	Avg	Max	Min	—
FPA2
500000	4.9	8	3	0.010	0.017	0.008	0
1000000	5.2	8	4	0.020	0.036	0.018	0
10000000	5.2	10	4	0.356	0.421	0.329	0
50000000	5.0	9	4	1.730	1.968	1.617	0
Newton
500000	4.9	8	3	0.017	0.021	0.013	0
1000000	5.1	8	4	0.036	0.045	0.031	0
10000000	5.0	10	4	0.407	0.494	0.359	0
50000000	4.9	9	4	1.961	2.345	1.757	0
Variable fixing
500000	8.2	10	7	0.021	0.024	0.019	0
1000000	8.4	10	7	0.045	0.049	0.040	0
10000000	8.9	12	8	0.517	0.573	0.482	0
50000000	9.0	12	8	2.542	2.695	2.338	0
FPA
500000	4.9	8	3	0.011	0.020	0.009	0
1000000	5.2	8	4	0.023	0.041	0.020	0
10000000	5.2	10	4	0.418	0.507	0.383	0
50000000	5.0	9	4	2.016	2.305	1.883	0
Secant
500000	8.1	12	6	0.018	0.032	0.012	0
1000000	8.5	11	7	0.037	0.064	0.025	0
10000000	8.3	12	6	0.382	0.637	0.239	0
50000000	7.9	11	6	1.927	3.082	1.016	0
Median search
500000	19.9	20	19	0.037	0.040	0.034	0
1000000	21.0	21	20	0.075	0.082	0.068	0
10000000	24.3	25	24	0.896	1.042	0.790	0
50000000	26.7	27	26	4.375	4.570	3.893	0